Andrew Chen Archives

Subscribe · Featured · Recent · The Cold Start Problem 📘
Dear readers, I have moved to Substack and I will be writing here from now on:
In the meantime, I will leave up for posterity. Enjoy!

Author Archive

Psych’d: A new user psychology framework for increasing funnel conversion (Guest Post)

without comments

[Hi readers, my good friend Darius Contractor (currently growth eng at Dropbox) has a brilliant new framework how user psychology has driven growth at companies like Bebo, Tickle, PhotoSugar and of course, Dropbox. Thanks to Darius and the folks at Reforge for putting this together. Hope you enjoy the writeup here! -Andrew]

Increase your funnel conversion by getting users Psych’d – by Darius Contractor

Have you ever wondered why people are bouncing from your nearly-frictionless onboarding flow? Why the same change can result in a lift on one page and cause drop-off on another? Or why people who find you via search bounce away after a few moments?

Having spent years focused on building experiences that got millions of users sharing, onboarding and inviting their friends, I’ve learned 2 things:

1. Every element on the page adds or subtracts emotional energy
2. Inspiring users is as important as reducing friction

A secret of the top growth experts in tech is to think about every UX interaction as an emotional event. But far from being random or beyond our control, emotion-driven interactions can be broken down into components, optimized at each step and replicated to get better results for onboarding and conversion.

The Psych Framework

Today I’m sharing the Psych Framework I’ve used to help grow companies like Tickle, Bebo and Dropbox. It is a systematic way to detect and improve the way an experience affects user emotional energy, which we call “psych”.

User Psych Framework by Darius Contractor

Every UX interaction increases or decreases Psych, the unit of measure for motivation to complete an action.

Every element of a webpage either inspires us by giving us more units of Psych or overwhelms us by depleting our existing store of Psych.

Once you understand what elements are adding to or depleting users’ energy, you can then start to manage that energy: adding inspiration and minimizing overwhelm to help users take your core actions.

Psych Units

We measure user energy in units of “Psych”, from 0 to 100.

User psych fuel gauge measures unitsA user at 100 Psych is maximally committed to their current experience, does not need further motivation, and will overcome most challenges. For example, a person who needs to file their taxes tonight will do whatever it takes to download their W2 from their company’s payroll site. They’ll complete a forgotten password step, suffer through a poor interface they’ve never used before, and read through confusing numbers in order to get their taxes done in time.

A user at 0 Psych is exhausted and disinterested, to the point of abandoning their current experience. For example, a person accidentally clicking on an ad who realizes they’ve ended up on a scam site will have no motivation to continue and will bounce.

Psych Elements

Being aware of +Psych in your UI can massively drive user excitement/growth:

Tinder: “Discover new and interesting people nearby” → Yes, let’s!

Likewise, being unaware of the -Psych in your product can massively decrease success:

Global Entry site: “Fill in 40 form fields about yourself” → Ugh, maybe later… <closes browser>

We call elements that inspire users and add to their emotional energy +Psych.

If a car rental site pitched “Get your car for $15/day” that might be a +Psych, inspiring users to try to get this good deal. Inversely, -Psych are items that tire or overwhelm users, such as long sign up forms, unclear UX, too much text, and unclear next actions.

Let’s test drive this concept with Match’s home page. After that, we’ll talk about what to look for in your flows and examine what Airbnb gets right in their host sign-up experience.

Example 1: Match’s homepage

Match homepage user psychology
How do we evaluate the Psych score of this page?

1. Determine your starting Psych

To understand how much Psych a user has when arriving to a site, consider how they got there.

For example, users who arrive on Match through a Google Search are high-intent and have intrinsic motivation since they’re explicitly searching for dating. So they’re around a 60 Psych.

By contrast, visits from banner clicks would likely be low-intent since their clickthrough came in response to an external trigger. They’re perhaps a 30 Psych.

A referral from a friend might result in a clickthrough that’s low intent but has high social validation. So, they might be at 50 Psych.

2. Follow the user’s attention from top-left to bottom-right

In left-to-right languages like English, we consume content from top-left to bottom-right. As we follow our natural path across the page, our Psych will either go up or (more likely) down as we encounter elements that excite us or elements that are obstacles.

On the Match homepage above, these are the elements we encounter from top-left to bottom-right:

  • Logo
  • Photos of singles
  • “#1 in dates, relationships and marriages”
  • Demographic form
  • “View Photos »” button

As you can see, right after the logo assuring us it’s a real company entity that we can trust, we see appealing photos of smiling people. Then we see the byline “#1 in dates, relationships and marriages,” which assures us that we’re going with the best site and that it’s there to help us achieve dates, relationships, or even marriage.

Next, there’s a form, which requires action that potentially depletes Psych. But we’re spurred on by the “View Photos” button — which is exactly the thing a user interested in “dates, relationships, and marriage” wants to do at this point.

3. Which elements are +Psych for you? Which are -Psych?

Let’s run through the Match homepage again and tally up Psych, element by element.

Match homepage user psychology each step animated
These are the + Psych elements:

  • “Ooohh, these people look attractive!” → +10 Psych
  • “They’re #1? And I can get dates/relationships/marriages?” → +3 Psych
  • “I like that it defaults to Woman seeking Man.” → +3 Psych
  • “Nice, can’t wait to View Photos” → +8 Psych

These are the – Psych elements:

  • “Hrm, what age am I looking for?” → -5 Psych
  • “Why do they need my zipcode? Argh, keyboard…” → -10 Psych

4. Sum it up!

Tally up all the Psych elements to see where users are by the time they get to your call-to-action.
Greater-than-zero Psych means the user got through the flow.

  • Starting from a Search: 50 Psych
  • +Psych: 10+3+3+8 = +24 Psych
  • -Psych: -5-10 = -15 Psych
  • Result: 59 Psych → They made it!

Next, we’ll go through some of the top +Psych and -Psych factors across common pages.

Maximizing Psych on each of your pages

1. Assess initial Psych

People come to your site with an initial quantity of Psych.

If you’re hungry at noon and haven’t eaten all day, then your Psych level for a sandwich will be very high. That will help you power through the friction of standing in line, deciding between options, and pulling out your wallet to be saved by an $8 hero.

Therefore, the first step to evaluating Psych is to look at factors determining how much Psych people have when they enter your funnel:

2. Psych on the landing page

Once a visitor hits your landing page — great! You now have multiple chances to increase their Psych to get them to continue to signup or, if you’re not careful, decrease their Psych and cause them to bounce.

3. Enter personal info

At some point, you’re going to have to ask your visitor to enter some personal information, even if that’s just their email address.

Asking for personal information usually creates a negative Psych moment, whether it’s because people are wary to share their information or because they’re simply feeling lazy and don’t want to complete an input action.

4. Interact with product

If you have a freemium or free trial model, your user will get a chance to interact with the product before it’s time to pay. This is a chance to increase Psych before the user gets to the Psych-depleting payment action.

5. Enter payment information

For most businesses, eventually it’ll be time for users to enter their payment information and complete a transaction.

This is the ultimate Psych-depleting action because of the psychological phenomenon of the pain of spending money.

While we might think that the purchase action should be Psych-increasing because of the anticipated pleasure of acquiring something we want, it actually triggers the same area of the brain as for physical pain. So, spending money = pain.

But, MIT researchers found that credit cards increase detachment from purchases. In other words, credit cards help decrease the pain we feel from spending (cash) money.

Aside from the inherent pain and negative Psych of spending, there are still things you can do to improve Psych at the payment step.

Decreasing cognitive load is more than just short signup forms

The above is a basic framework anyone can go through for figuring out the ups and downs of Psych within an onboarding flow.

But optimizing Psych isn’t just a matter of removing clicks and reducing text. In some cases, more detailed forms or copy can help by decreasing the cognitive load on making the decision — even though it’s technically more effort. For example, including more information about security and money-back guarantees can overcome trust barriers and alleviate fears for big purchases.

Now let’s look at another live example.

Example 2: Airbnb’s hosting flow

Let’s go through a more complex example — becoming a host on Airbnb.

How do we evaluate the Psych score of this page?

1. Determine your starting Psych.

To figure out the ups and downs of Psych for someone who’s thinking about hosting on Airbnb, let’s start with their mindset before they hit the landing page.

  • They might have heard some stories about making a lot of money on Airbnb → +20 Psych
  • They might have heard about it being lots of work or a negative hosting story → -10 Psych

But ultimately their positive starting Psych is greater than their doubts and they are motivated enough to check it out.

2. Follow the user’s attention from top-left to bottom-right

Airbnb offers three different ways for becoming a “host” on the platform:

  1. Renting a house or apartment
  2. Helping neighbors with their Airbnb listings
  3. Leading a tour or other travel experience

To keep it simple for this post, we’ll just look at the core Airbnb “host” action — renting a house or apartment.

As you follow the user’s attention from top-left to bottom-right, it’s pretty clear that Airbnb knows what its hosting users care most about — making extra money by renting out their places.

Here are the immediate elements we encounter above the fold, from top-left to bottom-right:

  • Logo
  • “Earn money” value proposition
  • An interactive calculator to see how much you could earn
  • A “Weekly Average” box with an impressive earnings estimate and a dollar sign

From there, the page displays more content aimed at increasing Psych and reducing doubt:

  • Tactical instructions that demystify how Airbnb works
  • Reassuring details about safety and security, like their host protection insurance and their $1M guarantee for hosts
  • Social proof in the form of a video featuring a diverse handful of hosts talking about how income from Airbnb has helped them to stay in their homes, pay for medical bills, or fund a retirement.

Because users have to log in to continue the host flow, Airbnb’s goal on this first page is to drive up Psych high enough to carry them through the subsequent, more tedious steps in the flow — logging in or creating an account and setting up their first listing.

3. Which elements are +Psych for you? Which are -Psych?

Now let’s run through and tally up Psych, element by element.

Airbnb does a good job of loading up the page with +Psych elements because they know they’re asking a lot of the user:

  • “Earn money? Yes please.” → +10 Psych
  • “I can rent out an entire place, or a room, or just a couch? Then there probably is something for me.” → +5 Psych
  • “$741 weekly average… higher than I thought.” → +10 Psych
  • “Host insurance, and a $1M guarantee that protects my stuff… ok phew.” → +4 Psych

4. Sum it up!

Once again, you can tally up all the Psych elements to see where users are by the time they get to a call-to-action. Remember, greater-than-zero Psych means the user clicked the “Sign me up” button.

This is a simple tally of Psych on just one of the pages of the Airbnb “become a host” flow. There are obviously many more steps before a user becomes a successful host that’s contributing to the marketplace. With that in mind, you can run through the Psych Framework at each step of your onboarding or conversion flow to find opportunities to reverse -Psych and increase +Psych.

So that’s the psych framework in a nutshell. Tell me what you think in the comments and if you’d like to hear more.


Disclaimer: I’m a Dropbox employee, but I’m not posting on behalf of Dropbox or in an official capacity as a Dropbox employee. This post was originally published here.

Written by Andrew Chen

June 12th, 2017 at 10:00 am

Posted in Uncategorized

Startups and big cos should approach growth differently (Video)

without comments

I recently did a video interview on the topic of how your growth strategy changes from being a small startup versus becoming a larger company. It’s hard to compete when you’re launching a new product and you have to think asymmetrically for your growth efforts to work. I walk through each stage, step by step, and talk through some of the strategic dynamics to think about. (Thanks to the Reforge folks for setting this up!)

You can watch the full video here, and check out the notes below.



New startups have to focus on underrated acquisition channels for early growth efforts [0:00]

  • Look for channels that are too small for the big companies to worry about — this is how smaller companies can gain an asymmetric advantage
  • Examples: niche communities, sub reddits, mailing lists, offline, blogs, linkedin groups, facebook groups
  • The best underrated channels are all small but high-intent

Find underrated channels that directly match your product’s target market [1:45]

  • Look at what are the channels that match your product the most.
  • It has everything to do with finding lots of little channels with high relevance.
  • Regardless of where you start, you need to quickly be able to cobble together a bunch of small channels to get going — not just one or two.

Learn with very small channels by focusing on qualitative feedback to start [2:48]

  • Any small channels that allow interaction with the audience, will allow you to do testing.
  • Even small groups for customer development are worth exploring
  • Test for the responsiveness of your audience

Scale up to the next, bigger channels, by starting with tests to optimize your performance [3:58]

  • First path: Go after the bigger channels that the bigger co’s are going after, siphon off a small amount of traffic, at minimum it’s a means of testing
  • Second path: trying to find a new channel / platform, that others haven’t considered that’s unique to your product. Example: Dropbox or Slack integration if you’re doing workplace productivity.
  • Those channels are “medium sized” right now but have the chance to scale up bigger as their APIs develop.

Big companies approach acquisition by building a portfolio of channels that can scale [5:32]

  • Once you’re big, it’s about building a portfolio
  • All the little things will hit their ceiling and won’t scale
  • You’ll end up with a few established, large channels, and your strategy will be about aggregation. Grow your portfolio, not replace channels
  • Big opportunity is around attacking existing channels (or new ones) in a unique way that’s tied to your product
  • Product-channel fit is key (Pinterest + workplace productivity tool = dissonance)
  • Examples for b2b: the calendar, the browser
  • For consumer, there’s YouTube

The most exciting channels “right now” need to tie uniquely/directly to your product [9:02]

  • “Right now” is the wrong way to think about it; it’s about the trajectory
  • Stage 1: Map the user lifecycle:
    • What are all the other tools, apps, offline experiences that your users are also coming into contact with?
    • Find out what your product is most adjacent to
  • Stage 2: Sizing / understanding the trajectory of all these things that are out there (at an MAU level).
    • What kind of integration can you get?
    • Some platforms are more conducive to virality or being used as a growth channel (instagram doesn’t give you a way to link out, so it’s less good of a channel VS youtube, which does allow cross linking and linking out)
  • Channels have to match what people are trying to do with your product

You have to pick the right social channels for virality [11:30]

  • Social isn’t just digital experiences; it’s also offline experiences
  • So, “social channels” are any ways that your users / customers talk to each other to convey that the other person should try a product
  • Direct recommendation or invite: “I’ve invited you to Facebook, you should use this.”
  • Indirect recommendation or invite: “I’ve taken this cool Instagram photo, do you like it?”
  • Same applies to B2B; Dropbox is an example of indirect.
  • Digital is important because it’s attributable, but the broader takeaway is to create a product with a bunch of touchpoints in a user’s life; those touchpoints trigger natural recommendation or invite opportunities

How do products spread on social channels? [16:06]

  • Extrinsic vs intrinsic motivation, and their accompanying rewards systems
  • Extrinsic — get a direct reward for referring someone
  • Intrinsic — my friends using this makes the experience better for me
  • They’re not mutually exclusive, and can work well together
  • Extrinsic is easier to bolt on after the fact, but intrinsic (built in to the product early on) creates deeper defensibility by creating network effects

“Going viral” doesn’t mean just building something cool [21:19]

  • “I’ll just make something really cool, and people will talk about it” is not as sustainable as acquisition channels that are based on combined intrinsic and extrinsic motivations
  • Example: Slack going viral

Not every platform is created equal (for virality). How do you build for the right one? [22:48]

  • Anything that’s spreading from user to user is spreading on a platform that already exists.
  • Platforms are built on top of each other (Facebook on top of .edu, many companies built on top of the Facebook platform) and certain ones are more suitable than others.
  • Key questions to ask when evaluating platform potential:
    • How easy is it for customers to communicate with one another?
    • How much control do you have (via APIs or anything else) to customize the invite experience?
    • Do you have the ability to add a link?
    • Can you get ahold of an address book or social graph, in order to generate invites?
  • You have to assess not only size and trajectory, but also how open is that platform? What are the hooks the it supplies you to build growth?

Written by Andrew Chen

June 6th, 2017 at 10:00 am

Posted in Uncategorized

BBS door games: Social Gaming innovation from the 1980s

without comments


I was always more of a Barren Realms Elite fan, but this picture was cooler!

And now you learn how I spent my childhood…
Some of the fondest memories from my childhood are playing BBS door games. Back before the web existed, I was a 10 year old kid in Seattle dialing into 3 or 4 different  different BBSes using a pirated version of Procomm Plus. There, I found that you could download all sorts of awesome products (though in 20 different parts, which you had to put back together), and more importantly, they had the ability to launch “door games” like Tradewars 2002, Legend of the Red Dragon, Barren Realms Elite, and a number of other games I grew to love. I spent so many hours tying up the phone line after getting back from school that eventually my parents banned me from playing these games – but that only convinced me to set an alarm for 3am to wake up in the middle of the night to play!

Furthermore, it became a critical thing in my life to get all my friends from school to also play these games with me. Together, we’d team up into a powerful, coordinated unit, and dominate the other players on the BBS and regional network. (Or at least that was the plan) Eventually years passed, I learned the internet was more than gopher, and I moved on to better and more powerful massively multiplayer games.

It’s obvious, in retrospect, that a lot of the door games that existed 20 years ago pioneered a lot of the same techniques that social games use today. Let’s drill into a bunch of these similarities, covering the following topics:

  • Door games are the “apps” to the BBS platform
  • Social gameplay with friends and neighbors
  • Turn-based, RPG-like gameplay
  • Low-tech graphical experiences, delivered as a persistent social experience

If I’m missing anything, please leave a comment! Anyway, let’s get started…

Door games are the “apps” to the BBS platform
First, let me describe what a BBS actually is – you can read a more official version on Wikipedia here. Anyone with a phone line, modem, and computer running the right software could start up a BBS. The software would tell the computer, when it received a call, to automatically pick up the line and start talking to the computer on the other end. On the other side, anyone could dial into the BBS with the right terminal software and once the connection was established, you’d get a screen that looked something like this:

On these BBSes, you’d typically find a couple different sections:

  • Private communication (reading and writing messages)
  • Public communication (message boards)
  • File sharing (downloading and uploading)
  • External applications, including Door games

The external applications were integrated with the BBS as described by Wikipedia:

The BBS software starts the external program, and the door system passes data back and forth between the door program, the BBS, and the remote user. To supply the door program with the user’s information (such as the user’s alias and the amount of time they had spent online), the BBS software creates a dropfile containing information for the program to read.

This “dropfile” typically contained all the user information, so rather than the standard API where the app asks for that information, instead it was provided in one big file. The door game would then parse this data and use it for the game. For the extra nerdtastic detail, you can go here for the dropfile specs.

Now of course this process of extending the BBS’s functionality isn’t as flexible as things are now – after all, you couldn’t just upload a new Door game to a BBS and get it running. You also couldn’t update your game and instantly propagate the new version out to all the users. But the central idea is still the same.

Social gameplay with friends and neighbors
One of the interesting properties of BBSes is that because they are all accessed by phone, and you don’t want to pay long-distance bills, you end up dialing into BBSes in your own area code. For me, that was 206, and it means that I was mostly gaming with people in my same regional area. Similarly, I convinced all my friends to also dial into the same BBSes and play games with me. For the games that had leaderboards and user-to-user interaction, it was easy to feel the same fun game-like motivations that make social gaming work today.

As an aside, while doing research for this blog post, I found this funny ASCII based dating site with Myers-Briggs based writing, and a Wall!

You can tell from the number of dudes on the screen above that the world of lonely nerds has not changed much over the last 20 years.

Turn-based, RPG-like gameplay
One of the big design challenges for any BBS game is that you want to encourage social gameplay, yet it can’t be real time. This is because, of course, people can’t be logged in to the same BBS simultaneously unless the BBS had a ton of different phone lines (not likely). As a result, each of the Door games had to support an social, asynchronous play style that allowed people to dial in one after the other and still engage.

The way that was done depended on the game, of course, but usually combined a mix of computer players (aka NPCs) and “slow” real-time action where each loop of action lasted a day. Then on any given day, you are given a number of turns which you can expend. Once you play these turns, then you need to wait until the next day to get more turns. This made it so that for a game like Tradewars 2002, you can log in, do your trading/mining/attacking, and interact with computer-controlled characters. If you encounter another player’s ship, then you can interact with them with the computer taking control of the other player, so when attacking them, they will automatically defend.

Some of these games played very much like RPGs, with levels, currency, monsters, swords, quests, and the usual mechanics. One of the most popular games was called The Legend of the Red Dragon:

Having the combination of social gameplay with the traditional RPG mechanics created a rich world that allowed for months of play time.

Low-tech graphical experiences, delivered as a persistent social experience
You may notice from all of these screenshots that these games are very, very basic. Many of my friends at the time criticized the simplicity of the gameplay, only to be  sucked in for social reasons :-) Similar to the current incarnation of social gaming, the emphasis is not on graphics. The primary advantages of a persistent, continually updated world with social gameplay far outweighed the fact that downloadable single player games had much better graphics. Of course I still downloaded and played Wolfenstein and Doom when it first came out, but throughout that time, I stilled played BBS games.

If there’s one thing to be learned from the BBS games and their related cousins, MUDs, is that great social interactions can trump pretty much everything else. Of course the products that can deliver higher production values and great social experiences are even better off.

It was fun to write this article! Such a blast from the past. Hope you share your memories of BBSes in the comments :-)

Want more?
If you liked this post, please subscribe or follow me on Twitter. You can also find more essays here.

(This blog was republished by the excellent blog Inside Social Gaming)

Written by Andrew Chen

August 25th, 2009 at 8:30 am

Posted in Uncategorized

Go-to-market strategies for vertical social products

without comments

Turns out it’s harder than you think to build a vertical social network for Star Wars fans

Are you trying to build a vertical social product?
One of the most common questions I get on viral marketing is structured as the following:

I’m trying to build a social network targeted at X. How do I make successful with viral marketing?

Typically, the product in question will be some sort of subset of a general social network, such as:

  • parents
  • sales professionals
  • fans of a TV show
  • car lovers
  • etc

and in fact, to generalize, a vertical social product might also include products like:

  • a sci-fi themed MMORPG
  • social products aimed at the Twitter audience
  • products for a certain demographic
  • people with webcams
  • etc

Catch my drift? Well, let’s get even more specific about what it means to be vertical versus horizontal.

Horizontal versus vertical products
Basically, the definition how vertical or horizontal something is depends on what % of people are likely to use it in some form. The most horizontal products become core features of the internet, such as:

  • browsers
  • search
  • email
  • mobile
  • video
  • etc.

For this reason, products centered around communication or content are often considered very horizontal because a lot of people are likely to use them. These end up being the most viral products, for a number of reasons we’ll discuss later.

Compare this to more specific products or features, such as content on cars, or a sci-fi themed video game. These are things where if you were to select a random group of 100 internet users, you’d find many of them have no interest in using that site.

The place that becomes important is the “branching factor” of your viral invite strategy.

How vertical products drive down branching factor
In tree data structures, the branching factor is determined by the number of children that a node has. It can be represented as the variable “num_friends” in the following equation:

users * num_friends * conversion % = new users

If you think of a viral loop as generations upon generations of users inviting each other, it’s very important that each user has a big branching factor, thus inviting enough of their friends to propagate the viral process. (With the usual caveat that you must balance your viral incentives against your user retention goals!)

The problem with vertical products is that it automatically impedes the # of invitees that a typical inviter might reach out to. The % of those invitees that get excited about the product upon their first use of it might be quite low, and if they leave prematurely without inviting their group of friends, then your viral process may peter out. It adds an extra term here, % targeted, which can significantly drive down this process:

users * % targeted * num_friends * conversion % = new users

So as a result, the more vertical a product is, and the lower the % of invitees that may be interested, the harder it will become viral. And in a world of shrinking possibilities on Facebook/Open Social, overcoming a big quantitative hurdle.

But don’t worry, it might still be OK!
I’ll close out this blog with two thoughts on this subject, first on scenarios where vertical products can still work, and also situations where you don’t care about being completely viral.

Vertical products can still work because of the fact that some target audiences are very concentrated amongst each other. The business world is often like this, or teenage girls, also! If you build a product targeted at them, the networks are dense enough and interconnected enough that you’ll get some degree of virality that eventually peters out as you hit this network’s edges, but it works in the meantime. I think that Linkedin is a great example of a site that provides great value, and because peoples’ Outlook contact lists are populated with large target-rich lists of like-minded professionals, this can work quite well.

As a related note, products like Ning, mailing lists, Yahoo Groups, etc. also do an interesting variation of this where maybe you came in via a “for Moms” invitation but then you end up making your own “for pet fanatics” group. This basically converts vertical traffic into horizontal traffic, neat!

Similarly, it’s not required that a site become viral to succeed. Oftentimes, a more specific audience is able to generate key context which provides high advertising revenues and/or leadgen/CPA opportunities. Because of this, these sites can often buy significant amounts of traffic which are then amplified by some degree of viral marketing but is not completely driven by it. In this way, the viral marketing is just a “kicker” that drives down the cost per acquisition without pushing it down to zero.

Both cases can be successes – I’d encourage everyone to check out Dogster, Linkedin, IMVU, and Puzzle Pirates as three examples of vertical products that worked out their acquisition strategies.

If you liked this article, you can check out my related essays here.

Written by Andrew Chen

October 20th, 2008 at 8:00 am

Posted in Uncategorized

Ad targeting talk from Community Next: People Not Pages (updated x2)

without comments

I updated the Community Next sldies from last time with an audio track, so I thought I would publish these slides again. Pardon the intrusion if you feel like you’ve already grokked all the details!

CommunityNext: People not Pages

View SlideShare presentation or Upload your own. (tags: advertising sns)

Hi everyone. My name is Andrew Chen and I’m a blogger and entrepreneur, focused on consumer internet products here in San Francisco.

I recently gave a talk called People Not Pages at the Community Next, which I’ll share today.

This year, the online advertising industry has reached roughly 20 billion dollars, that’s a big number. And a big chunk of those dollars are here because of the fact that advertisers can target consumers in ways that are simply not possible in traditional media.

Talking about ad targeting
So today we’ll talk about the business of ad targeting, and in particular 3 things:

First off, how can publishers generate more revenue from ad targeting? How does this actually happen?

Secondly, how do advertisers evaluate the effectiveness of their targeted campaigns? By understanding how they think about the question “does it work?” publishers can serve them better as customers.

And finally, we’ll discuss a topic that’s near and dear to my heart, which is: Why do social networks suck so much at monetization?

Note that this talk will mostly focus on the publisher-centric view, since the startups here in San Francisco tend to be creators of ad inventory rather than buyers of it.

About me
So quickly, before we begin, a quick bio about me. Again, my name is Andrew Chen and I’ve spent the bulk of my career in online advertising. I was most recently an Entrepreneur-in-residence at Mohr Davidow Ventures, a time I greatly enjoyed, evaluating deals in ad infrastructure as well as companies that are ad-supported. Before that I spent 4 years running product at Revenue Science, a behavioral targeting ad company in Seattle, and before that in the automotive vertical at Cobalt Group.

About Revenue Science
Most my experience will be drawn from my time at Revenue Science, so I’ll talk briefly about the company. Basically, as a behavioral targeting ad provider, they work with some of the largest brands online – Reuters, NY Times, ESPN, AOL, and dozens of others – to help them increase their revenues via targeting. While I was there I worked on their brand publisher products as well as getting their ad network off the ground.

Generating more revenue as a publisher
On to our first topic: How does targeting generate more revenues for publishers?

To understand this, I think you have to go all the way back to the newspaper days. This is an article from the early 1900s, costing one cent, documenting when the Titanic struck an iceberg. You have to think about how advertising was bought and sold at that time to understand why things work the way they do online today.

Now imagine this: You’re a newspaper company, which means you actually have to print physical copies of your text. That means that if you’re going to sell advertising, you must also print your ads on that physical copy, next to the editorial that’s relevant. That means that as an ad sales professional, your job is to take an audience that an agency specifies, let’s take “small investors” as an example, and translate that into a particular section in your paper. And so you sell them the Finance section. This means the agency asks for people, but you sell them pages.

The modern day online publishers are not much different. For a publisher like NY Times, you’ll have a bunch of sections on your site. You have the homepage, which makes a good CPM because it gets the most traffic and is the most prestigious placement on the site. Then you have high-value sections like Technology, and Finance. Then you have low value sections like Opinion and World News, which are harder to sell because what advertiser wants to be next to a story about the War in Iraq? So all of the low value stuff gets classified as “remnant ad inventory” and is lumped together.

So, this means that you have a bunch of different sections, and you can sell each for different prices. Again, the agency asks for a specific kind of person, let’s say IT decisionmakers, and you give them back the Tech section.

But wait a minute here, the IT decisionmaker reads the tech section, but also browses around the site. Is it right to value them highly on one pageview, and then drop them on the next? This is really a vestige of the old newspaper mindset.

If the agency is fundamentally targeting an audience, and you can deliver this audience regardless of where they are on the NY Times. or potentially even across the internet, you can deliver them more value. And lucky for the publisher, they are then converting their low-CPM remnant inventory into high-value targeted inventory.

This process of finding the qualified audiences in the remnant inventory is really what a publisher strives for.

Targeting techniques
So all the different targeting techniques that are out there, be it Geo, SIC, Demographic, Content, Daypart, or Behavioral, is all about finding these valuable audiences.

Once you have your targeting toolset in place, then the value of your ad inventory is increased. You find the IT decisionmakers in the opinion section, and convert low-value ads into high-value ones.

Evaluating targeted ad campaigns
Great, now let’s move to the second topic. How do advertisers evaluate whether or not targeting campaigns are working?

To understand this topic, you have to understand that there are really two major groups of advertisers out there: Brand advertisers, and direct response. Let’s talk about each in succession.

Brand advertising metrics
Brand advertisers are the guys you see all over the place, like Coke, Budweiser, Honda, United Airlines, and others. For many of them, their biggest goal is to drive awareness of their brand messages regardless of strict ROI. Obviously the ads are meant to eventually sell product, but it’s often not accountable from a strict ROI basis.

This brand spend means big dollars. Take a Budweiser NASCAR vehicle, for example. These cars can cost up to $20MM to sponsor. This year, Facebook is rumored to gross $300MM in revenue, which only translates to 15 NASCAR sponsorships.

So what kind of metrics are they looking for?

Well, here’s a case study that I worked on many years back. American Airlines and WSJ worked together on a campaign which was evaluated using Dynamic Logic, an ad effectiveness company. They went and surveyed users who had seen the campaign and who hadn’t, and asked them questions around message awareness and purchase intent. And it led to some positive results, which I’ll discuss.

First, there was a metric called Audience Composition. The campaign was targeted at frequent business travelers, so the survey qualified who fit the bill. The targeted campaign had a higher % of business travelers than the untargeted one, in fact by more than 2X.  Same for the definition of heavy travelers, people who took more than 5 trips a year, where it was 145% better.

The other set of metrics were brand and persuasion metrics, like Aided Brand Awareness and Message association. For example, those questions might revolve around things like, “Of the following brands, which do you associate with higher comfort” and would name a series of airline brands. The ones in the campaign were correctly associating the brand message.

Direct response advertising metrics
Now let’s go to direct response advertising. I used a picture of a cheesy infomercial because that’s the kind of advertiser that really cares about their results. Every dollar they spend must generate more dollars. They care a lot about ROI, as a result.

The example case study here, which I did not work on, is from Yahoo. In their example, they created an audience segment of mobile users who searched for ringtones, specific mobile providers, and browsed relevant pages. For the advertiser targeting this audience, the campaign generated higher CTRs and conversion rates, which are good ROI metrics. The better the increase in conversions, the more you as the advertiser can spend on this audience to get them to purchase your products.

Social network monetization (and why it’s hard)
OK, and the final topic for today, a discussion of why social networks suck so much at generating revenue. This is a topic I’ve written a lot about on my blog.

First off, this slide shows what you get when you search for “myspace addict” in Google images. The fact is, people find MySpace insanely addicting, which is great for them on a pageviews and engagement perspective. Problem is, this same engagement is what makes MySpace users pretty uninterested in what you’re trying to sell them. They are simply not in the mood to buy stuff, and when social network audiences are in their “flow state” clicking around, the CTRs can be quite low. This is exactly what you see, with CTRs often 0.1% and below, whereas search CTRs are often in the 10s of percentage points.

Social network data = sucks for targeting
The next issue is the type of data social networks are amassing. I included a real profile from a friend of mine, Andrea, which states “I have one interest: Dane Cook. I’m kidding I have more: Movies, music, soccer” etc. This tells you that Andrea’s interested in these things, but not that she’s in the transactional window to buy something. This is completely different than the kind of data you get from users when they are searching for dane cook movies on Google, or searching for dane cook on Amazon.

This issue is really the distinction of Interest versus Intent, and it’s a really hard area to break through. There’s a lot of social advertising companies working on this problem, and good luck to them, because it’s very hard.

Don’t assume that all the data you’re collecting is useful, but it’s likely that it’s not, unless the user is very close to making a transaction.

Wrap up
OK great, so we’ve covered the main topics. We talked about how publishers make money, and the process of turning remnant into premium. We also talked about brand versus direct response metrics in evaluating targeting campaigns. And finally, we talked about why social networks are so bad at monetization.

If anyone has questions, please feel free to e-mail me or check out my blog. Thanks!

Written by Andrew Chen

September 12th, 2008 at 8:00 am

Posted in Uncategorized

Remnant ads and the advertisers who love them

without comments

For social networks, your customers are often remnant ad networks
In previous posts, I’ve discussed the reasons why social networks often resort to ad networks to monetize their sites. As a result, it could be said that remnant ad networks are often the largest customers for these sites.

Let’s spend some time in this post discussing these customers of social network ads, and how things are typically structured.

An example remnant ad:

What’s remnant advertising?
First off, let’s define some terms – feel free to skip this if you’re already familiar with online ads. For any site with a bunch of ad impressions, the entire “inventory” of all your impressions is broken up into different pieces. Quoting from one of my earlier essays:

  • The first US impression in a session has the most value ($10)
  • Then impressions 2-5 have some level of brand value or high CTR value ($3-5)
  • Then after that, you’re hitting ad networks selling on category ($1)
  • Then eventually, you hit remnant ad networks ($0.50)
  • Finally, you hit pure CPA remnant networks ($0.10)

Usually at the high end, the advertiser gets to be pretty picky about what kinds of ads they get. They might be able to specify a bunch of targeting, how many impressions they want from each user, and most importantly, where the user will see the ad. So if you have a music section on the site, then a big record label is likely to spend brand dollars only on the first couple impressions, and only in the music section.

So after your ad sales team has sold all the premium impressions they can, what happens to the rest? Well, as my friend Jay Weintraub eloquently describes, the rest of the “remnant” impressions are typically offloaded to ad networks, CPA offer advertisers, leadgen companies, and other folks who don’t mind the fact that you as a publisher aren’t guaranteeing placement on their site.

What it means to be a “blind” ad network
Often times, one interesting characteristic of remnant ad networks is that they are generally described as “blind” networks – this means that the advertiser, in many cases, doesn’t know where their ads are going to appear. Unlike doing an ad deal with ESPN, where you know exactly what page your ads will show up on (the Golf section, let’s say), instead the ad network could put you across any of the publishers they work with. There isn’t much guaranteed, other than the fact it won’t be next to porn or other completely inappropriate content.

It’s not always true that ad networks will sell their ads blind, but it’s an important class of inventory since it often makes up a majority % of the ads that a network will touch. In the case of AdSense, I’d guess that number if >95%, for example. When it’s not blind, often publisher sites are classified into categories like “pets” or “cars” or other broad channels, but let’s focus on the blind inventory for now.

Advertisers who prefer blind ad inventory
Most of the advertisers that are willing to buy this blind inventory tend to be advertisers with very clear monetization strategies in the backend. This is often why you often see ads from companies like:

  • Ringtones/wallpapers/etc (monthly mobile subscription)
  • Mobile jokes/crushes/etc (monthly mobile subscription)
  • Toolbars, like IAC’s Smiley Central (search toolbars = predictable ongoing revenues)
  • Free ipods after you fill out an offer (leadgen where each lead as a concrete value)
  • Dating sites (with a recurring subscription fee)
  • etc.

And inversely, this is also why you don’t often see large brand advertisers or folks with very specific advertising goals (for example, if you did lawsuits for cancer caused by asbestos) advertising on blind inventory. Without control of the context, the dollars put into these campaigns can be wasted.

In fact, there are certain characteristics that advertisers who routinely buy remnant ads often share. Here they are below:

  • Broad, horizontal offering
  • Clear LTV in the backend
  • More transactional than not

First off, you need a broad offering because you often don’t know where the ads are going to be placed. The broader the offering, in fact, the better, since you’ll also be buying primarily on inventory that other advertisers find low-value. A good example of this is ad inventory focused around communication (email/social networks/etc) or media/entertainment. So if your product or service is very broad, then you can maximize the # of likely people you’ll hit, and also be able to purchase on the cheapest ad inventory.

As discussed above, a clear LTV (that’s good!) is important since it lets you focus on the simple equation that your spending costs must be lower than the profit from each user you buy. In many cases, a higher LTV might even drive the advertiser to purchase more broadly, just because the value on the user is so good – I’d consider many “in-market” audiences like home buyers and such as a good example of this.

And finally, a transactional mindset is often important since it correlates with LTV. You want to drive the user down a fairly narrow funnel that gets them to act on a high-value action. A good example of this is compelling the user to download a toolbar, or fill out a lead form, or similar. That way the advertiser is within a very small steps of making money from the user they paid for, and can balance it all out in a “portfolio” basis of buying large number of clicks to their landing pages.

The future of remnant ads on social networks
Right now, most social networks and social network apps have a great degree of reliance on these remnant ad networks. Because there isn’t clear context in writing on someone’s wall or playing a game, these low-value impressions are disposed of fairly cheaply. It’s interesting to see some of these guys “going native” on social networks, like Super Rewards and Offerpal, particularly in regards to virtual currency. I have to imagine that this will only continue.

If you liked this blog post, you can get updates by email or RSS here.

Written by Andrew Chen

September 2nd, 2008 at 8:01 am

Posted in Uncategorized

Counting your big pile of Benjamins: 5 startup tips for maximizing ad revenue

without comments

(this is a cross-post of the guest blog I wrote over at the Startonomics blog yesterday)

So you want to make some money, eh?

Every startup gets to the point where they decide to bring in some ad revenue, and it's worthwhile to consider exactly how much money will be generated through advertising. So in the parlance of the ad industry, your "ad inventory" is the set of impressions that you have to sell to people. Let's talk about modeling that goes beyond the simple reports into thinking about what drivers influence your CPMs and impression counts.

This blog post will go over some of the key milestones that any ad-supported startup will have to face. In particular, we'll discuss 5 stages:

  • Stage 1: Just getting started
  • Stage 2: Layering in new sections
  • Stage 3: Thinking about growth
  • Stage 4: Adding in multiple ad networks
  • Stage 5: Building up brand sales (rep, vertical, direct, or otherwise)

Let's get started with some basic metrics around CPMs and pageviews…

Stage 1: Just getting started
When you first get started, it makes sense to sign up to a self-serve, automated CPC ad network like Google AdSense. At the end of the month, you'll get a report for how much money you've made and what your CPM is, which will look something like this:

For many folks, just looking at this kind of number is enough, although it doesn't provide much detail beyond the core revenue numbers.

As I mentioned in my previous blog on 5 factors that determine your CPM rates, there are some rules of thumb for what kinds of CPMs you should expect on your site. Here are some very rough numbers:

  • Social sites (forums/chat/etc) without direct ad sales teams: <$0.25 CPM
  • Largely international sites: <$0.50 CPM
  • Medium-sized sites that use banner ad networks: <$1 CPM
  • Reference sites in a specific category: >$5 CPM or sometimes
    much higher, depending on category – we ran into home improvement
    reference sites that did $20 CPMs

Much of that is determined by the frequency of the pageviews, how specific it is, how much international advertising there is, etc.

Stage 2: Layering in new sections
For the sites out there that grow beyond AdSense, you need to have more granularity for thinking about and optimizing your ad inventory. You have to break down your site into the pieces that influence the performance the most. In general, that tends to be factors like:

  • site section (or function), like Homepage, Profile, Media, Inbox, etc
  • ad unit size, like 300×250 or 468×60 (full IAB ad units listed here)
  • type
  • targeting
  • etc.

The point is, you should have a good sense of where your ad units and impressions are coming from, and how revenues overlay that. And the more impressions you get, the more granular you'll want to track your advertising.

Let's take a social networking site like MySpace as the example here. They might break down their inventory, at an early stage, into something like below:

Section Impressions Avg CPM Revenue
homepage 100,000 $1.00 $100.00
forum 500,000 $0.50 $250.00
profile 200,000 $0.50 $100.00
message 100,000 $1.00 $100.00
Total 900,000 $550.00

(disclaimer: note that these numbers are for purely illustrative purposes only – your mileage may vary)

Once you have an inventory breakdown like above, you can start rolling up your aggregate revenues, impressions, and CPMs.

Stage 3: Thinking about growth
The next phase is scaling up your site and thinking about where you're likely to add to your ad inventory. In general, static areas like the homepage aren't necessarily going to grow very much. But other portions, like forums or profile or other social areas are likely to experience network effects that drive non-linear growth.

As a result, you'll end up adding traffic disproportionately to the social areas of the site, and your aggregate CPMs are likely to be brought lower by the higher-traffic, lower-value communication parts of your site. In the cases where the higher pageviews are also brought on by higher engagement, you might also expect the CPMs to decrease as you hit frequency caps imposed by the various ad networks upon your userbase.

Section Impressions Avg CPM Revenue
homepage 200,000 $1.00 $200.00
forum 2,000,000 $0.40 $800.00
profile 1,000,000 $0.40 $400.00
message 500,000 $0.75 $375.00
Total 3,700,000 $1,775.00

(again, disclaimer: note that these numbers are for purely illustrative purposes only – your mileage may vary)

The point is that even though the traffic grew over 4x, the revenue only grew 3x. That's to be expected based on the reasons discussed above, primarily driven by lower-value inventory tending to grow like weeds.

Stage 4: Adding in multiple ad networks
It's also often the case that having many different ad networks can dramatically affect your CPM rates as well. Because of frequency capping issues and the availability of high-value campaigns for each ad network, a lot of times high-CPM opportunties are spread lightly across many sites and you don't get much of a benefit for having deep engagement. One way to solve that is by serving the ads of many different networks and optimizing between them to get the best prices at each point.

There's a great post by Mike Nolet of MikeOnAds on this subject, where he talks about how his previous company, RightMedia, handles the ad-network-to-ad-network optimization of revenue.

The key image from Mike's post is below:

What you're seeing here is that as user frequency increases, the monetization between the 3 networks (small/medium/large) is different. Ideally you'd start out on the small network, then go to the medium one, and then go to the large one in order to monetize for that CPM.

My personal experience (while at Revenue Science) working with the MySpace ad ops team led by the fantastically smart David Dipilato was that they juggled more than a dozen very large networks in order to maximize their revenues. This level of expertise in ad ops is a very hard skill to train and hire for, by the way.

Stage 5: Building up brand sales (rep, vertical, direct, or otherwise)
Once your site hits a certain level of ad inventory – and that number's certainly over 50M/month and maybe even several multiples of that – it's likely that you'd want to go after brand dollars. The reason is simple: remnant ad networks bring in CPMs in the dollars, whereas brand can take you into the tens of dollars. The problem is that the process of doing brand ad sales is very foreign and specialized, especially to folks who have thus far focused on technology, automation, and scalability. Brand sales is like enterprise sales – it's slow, custom, and not technology-intensive at all. To win you focus on excellence in sales and marketing operations.

Here are some of the common options for getting brand dollars:

Old school ad networks
Generally most folks get started through their existing ad networks, which start to rep their inventory as part of a brand category. This can happen if you're a tech blog and they sell you as part of their "Technology channel." You'll end up seeing these dollars mixed into your overall ad network CPMs, and hopefully they boost your aggregate numbers into something good!

Vertical ad networks
Another way that's getting much more common now is to get into some sort of vertical ad network like or Jumpstart or whatever is appropriate for your site. (Glam = women and Jumpstart = cars, btw). These guys are far more sophisticated and specialized, and commonly move a ton of ad inventory as part of their business models.

Direct sales teams
Similarly, many companies go out and hire their own sales teams, or engage ad reps who already have those relationships. This is a hugely expensive process, but that's how the big money comes in.

Once you get your brand sales going, the site you're running will need a dedicated team to manage the active campaigns. The different ads coming through the system need to be carefully and constantly checked to make sure they are delivering in a timely manner, that they are running in the right sections, and that whatever performance metrics the advertiser is using goes well.

On the revenue side, you can no longer make little tables of inventory and use that to predict revenue. It's common that a very small % of ad impressions generates a sizeable % of the revenue. I've heard that some sites sell 5-10% of their impressions as brand ads which account for 30-40% of the revenue, for example. The right way to think of this is that your ad inventory ends up being a pyramid with the best stuff on the top and the worst stuff forming the base. 

So ultimately, you track the ad campaigns that the ad sales team is selling and your ad ops team is running exactly like an enterprise system. You have a sales pipeline with different campaigns going through based on RFPs your team is getting, and eventually once those flight, the revenue comes into your company.

In summary
There are many complexities in squeezing out the most dollars from your ad inventory – it's something that even big media companies struggle to master as they grow. Much of the process of doing this smoothly involves making sure you don't try to solve hard problems until they really become core to your business – for example, don't hire a big ad sales team before you have enough pageviews. Similarly, it's important to understand the implications of how your product ultimately impacts the monetization – social products, for various reasons, tend to monetize poorly and that needs to be built into your business model from Day One.

For more essays on advertising, social network monetization, and more, check out my directory of essays here.

Written by Andrew Chen

August 26th, 2008 at 8:00 am

Posted in Uncategorized

Quotes from entrepreneurs and friends about Futuristic Play

without comments

Organizing a blog
I've slowly been working on the blog infrastructure I have set up, including at some point moving to WordPress. Recently I've just focused on organizing the essays and information I've written so far, including making a big directory of all my essays in one place.

Anyway, I wanted to work on making it easier for new users to figure out what this blog was about, so I added a "Are you new?" section with a couple links to my best essays, quotes about this blog, and an about me / contact info section.

I e-mailed a couple folks to throw together the quotes page and got a great response. I wanted to include a snapshot of the list below – thanks to everyone to contributed a quote!

The sad truth about blogging is that 99% of blogs are not worth
reading.  Andrew Chen's blog on online marketing is one of the few
exceptions to this rule.  I often find myself referring people to posts
from Andrew's blog because it actually contains useful research and
insight, instead of the usual noise.
Jonathan Abrams, ceo of Socializr and founder of Friendster

Marc read Andrew
good blog
— a half haiku by Marc Andreessen

Andrew Chen's blog is one of the few I make sure I read every day — both thoughtful and thought provoking.
Lauren Bigelow, svp marketing and gm of Weeworld

Andrew’s incredibly analytical mind, understanding of web
development, and expansive network of entrepreneurs, venture
capitalists and veterans, affords him the unique opportunity to jump
into the trenches on case studies for an intimate micro-level
understanding and to learn from experts a wealth of research and wisdom
of the macro forces at play.
Jason Feffer, ceo of Sodahead, previously vp operations at MySpace

I'd call Andrew's blog 'unmissable.'
Jared Kopf, ceo of Adroll and Paypal alum

Andrew's blog is an inexhaustible source of numbers and ideas on concepts that matter.
Max Levchin, ceo of Slide and cto/co-founder of PayPal

Andrew is one of the smartest geeks in Silicon Valley, and a
brilliant and comprehensive virtuoso on a broad array of tech and
non-tech topics including internet advertising, viral marketing, web
analytics, social networks, online games, and all kinds of web
glitter.  He is a veritable Glenn Gould of startup blogging, and every
chance i get i rip off his ideas and try to pass them off as my own.
Dave McClure, angel investor and Paypal alum

I subscribe to Andrew Chen’s Futuristic Play blog because I value his insights into virtual worlds and gaming.
Cary Rosenzweig, ceo of IMVU

My brother and I built JibJab by instinctually understanding and
leveraging viral marketing.  When I want to think about it in a
quantitative way, there is no better source of information (and
inspiration) than Andrew Chen’s Futuristic Play.
Gregg Spiridellis, ceo of JibJab

Andrew is writing the textbook on web marketing.
— Jim Young, ex-cto/co-founder of HotOrNot

Written by Andrew Chen

August 25th, 2008 at 8:42 am

Posted in Uncategorized

Speaking at Startonomics and guest blog post: 5 startup tips for maximizing ad revenues

without comments


I will be giving a talk in October at Startonomics, run by Dave McClure, which is a conference described below:

Startonomics is a one-day workshop designed by entrepreneurs
for entrepreneurs on how to create simple, actionable metrics for
internet startups, and use them to make better product and marketing
decisions for long-term growth and success.

I’ll be doing a session on making money:

Revenue: The Internet Wants to Be Free, but You Need to Get Paid
Learn how to generate revenue using a variety of business models &
strategies including advertising, digital goods, subscriptions, lead
generation, & e-commerce.

Dave asked me to make a contribution to the Startonomics blog, which I’ve cross-posted a preview for.


Counting your big pile of Benjamins:
5 startup tips for maximizing ad revenue

So you want to make some money, eh?

startup gets to the point where they decide to bring in some ad
revenue, and it’s worthwhile to consider exactly how much money will be
generated through advertising. So in the parlance of the ad industry,
your “ad inventory” is the set of impressions that you have to sell to
people. Let’s talk about modeling that goes beyond the simple reports
into thinking about what drivers influence your CPMs and impression

This blog post will go over some of the key milestones
that any ad-supported startup will have to face. In particular, we’ll
discuss 5 stages:

  • Stage 1: Just getting started
  • Stage 2: Layering in new sections
  • Stage 3: Thinking about growth
  • Stage 4: Adding in multiple ad networks
  • Stage 5: Building up brand sales (rep, vertical, direct, or otherwise)

Let’s get started with some basic metrics around CPMs and pageviews…

Read the rest over at the Startonomics blog »

Written by Andrew Chen

August 25th, 2008 at 8:00 am

Posted in Uncategorized

Weekend links from my Twitter

without comments

Just an update to the random links I've been posting to my Twitter account over the last
week or two – no guarantee that they have anything to do with work!
Mostly a grab bag of nerdy/biz/news stuff.

If you want to get these in real time, you can follow me here.

Here they are:

Have a good weekend!

Written by Andrew Chen

August 24th, 2008 at 10:40 am

Posted in Uncategorized

Giving a talk at CommunityNext in 2 days (free spots available for publishers)

without comments

I'm giving a talk this Thursday at CommunityNext::Monetize to cover an overview of the ad targeting landscape. Tickets are currently free for publishers with 10 open spots, use the code ANDREW.

You can register here

The outline of my talk will be…

People not pages:
An introduction to ad targeting

  • Targeting = $
    How publishers can generate higher CPMs using targeting techniques from demographics, behavioral, geo, etc.
  • Does it work?
    Comparing the performance metrics for brand versus direct response advertisers using studies from Revenue Science and Yahoo campaigns
  • Social networks suck at monetization
    Discussing why social networks traditionally have had low CPMs and driving into issues like 1) low CTRs and what forces generate them, 2) interest versus intent, and 3) ad sales team dynamics

The talk will be about 30 min and should start at 3:30pm.

See you guys there!

Written by Andrew Chen

August 19th, 2008 at 8:23 pm

Posted in Uncategorized

Recent random links from my Twitter feed

without comments

I have been posting random links to my Twitter account over the last week or two – no guarantee that they have anything to do with work! Mostly a grab bag of nerdy/biz/news stuff.

If you want to get these in real time, you can follow me here.

Written by Andrew Chen

August 18th, 2008 at 8:00 am

Posted in Uncategorized

Internet Advertising Bureau and Bain on pricing in online ad markets

without comments

For folks who are working in the online advertising space, see below for a great summary on the dynamics between ad networks and publishers. There's a full document from the IAB and Bain.

Here's the the Executive Summary: Overall, 2007 was a strong year for the seven participating publishers

  • Average revenue growth of 32%, with CPM increases for several participants
  • Growth in ad impressions served and in sell-out (after secondary channels)
  • High demand for premium video inventory trading at 2-3X display CPMs

At the same time, use of ad networks increased dramatically, from 5% of sold inventory in 2006 to 30% in 2007

  • Ad networks were used to monetize significant unsold display inventory
  • Publishers under considerable pressure to realize all revenue opportunity

    Average realized CPMs on ad networks ranged from $0.60-$1.10, versus $10-$20 in direct-sold display inventory, or only 6-11% of direct pricing

Importantly, the study revealed significant publisher challenges in managing pricing and yield

  • Lack of longitudinal sales data to measure trends – overall, by account and by channel
  • Limited staff resources and tools in place to optimize CPMs and inventory yield
  • Several participants lacked data on ad network volumes and pricing

It is still too early in the game to measure the full impact of ad networks on online pricing and revenue share

  • Large marketers still rapidly shifting budgets to online – “all boats rising”
  • Publisher use of ad networks still too recent to see “cause and effect”

    However, growth in marketer use of ad networks will likely lead to erosion of premium CPMs if publishers maintain current behavior

For publishers, two key implications:

  • Need to better support the value of premium inventory – through more innovative offerings and/or reducing units available
  • Need to actively manage secondary channels, both to maximize yield and to safeguard strategic position

The ability of ad networks to increase CPMs and share gains with publishers also appears critical to creating win-win relationships

  • Enhanced ad network targeting and inventory management resulting in higher price realization on premium inventory from publishers
  • Further scale-up and (potentially) consolidation of networks should enable higher margins

Bain IAB Digital Pricing ResearchUpload a Document to Scribd
Read this document on Scribd: Bain IAB Digital Pricing Research

Written by Andrew Chen

August 13th, 2008 at 3:48 pm

Posted in Uncategorized

YouTube vs Webkinz: Case studies for new product adoption

without comments

Time slicing Google Insight queries
Yesterday I wrote about the use of Google Insight to see how widely products had reached in the US. Inspired by a recent post by Erik from Seattle
on the growth of Twitter from month-to-month, I explored the fact that
the Insight product actually has data from as far back as 2004! This
means you can try out timeslices of data on different websites, before
they were successful, to see if there were common growth patterns as they grew in userbase.
For example, you can look for the search term "youtube"
and see what states were active in that search in Nov 2005, Dec 2005,
and so on, until the entire US is saturated.

Interestingly enough, the data actually shows there are many different patterns of growth – I show two below, based on YouTube and Webkinz. Everyone knows what YouTube is, but if you need a refresher on Webkinz, here's the Wikipedia entry.

Before diving into this though, I have to repeat the caveat from my
previous post. First off, this is not traffic from different states,
it's searches. You have to make the assumption that more searches
correlates to more traffic, which sounds reasonable to me but might be
completely wrong. Also, obviously these searches skew toward Google,
which has its own bias as a search engine. And finally, I don't know
how the search index really gets computed, but I'm assuming higher
search index means more searches.

Now that we have that lawyerly disclaimer out of the way, here's the data…


Nov 2005: As expected, YouTube starts out in
California and New York, which are two of the more digitally inclined
states, and also the most populous. The site was launched to the public
in late 2005, and as you can see from these next couple pictures, it
grew very very quickly.

Dec 2005: Next, it spreads to other populous states, including
Florida and Texas. One observation would be that because YouTube was
generating lots of traffic from MySpace and search, it'd be logical to
think that they'd have fairly "general" audience growth and follow
whichever states had the most people.

Jan 2006

Feb 2006

Mar 2006: Starting to fill in more to the central states. Note that
it's light in the south even though there's a TON of people there –
perhaps those were the YouTube late adopters?

Apr 2006

Sep 2006: The final month I'm showing is about 2 years ago, where it
was able to achieve widespread adoption throughtout the US. Note that
Hawaii is very dense, along with California. If you look at the Insight
data now, it basically has not changed much since Sep 2006.


Mar 2006: Webkinz, unlike YouTube, started out in Massachusetts. Did
you know that the kid-centric MMOG shared the same birthplace as
Facebook? The Wikipedia entry for Webkinz
lists the starting date for the company as Apr 2005, so it took them a
year to register enough searches to show up on Google's tool. Note that
since the MMOG is tied with a plush doll, much of the expansion is
probably related to actual product distribution and how they rolled the
product out. Similarly, it's been noted that "viral channels" don't
really exist for kids – they primarily learn from each other via word
of mouth, since few are reading blogs or twittering ;-)

Apr 2006: Unlike the YouTube example, Webkinz doesn't immediately
take off in Texas or California. Instead, it makes a slow expansion
into New York and Florida. (In this set of pics I'm skipping more
months since the growth is so much slower than YouTube, btw)

May 2006

Sep 2006: The growth continues westward, with Illinois, Michigan,
Ohio, and a couple East coast states. Note that California finally
shows up with some searches.

Dec 2006

Jun 2007: Much later, you see that the Webkinz phenomenon is still
mostly East Coast focused, and doesn't involve California much even
though it's a very populous state. Strange!

I think there are a couple interesting observations to be taken out of the two, in concert with each other:

  • Different products get adopted differently, and follow different patterns
  • Horizonal product like YouTube and grow quickly across the US, and hit the most populous states first
  • Products like Webkinz, on the other hand, have very distinct signatures and might have to do with the fact that it's
    tied to a offline product and/or the fact that the customer base of
    kids doesn't have many consolidated online marketing channels
  • Similarly, the actual underlying growth of some products can be
    very fast, with state-by-state growth happening over weeks or single
    months, or sometimes the distribution doesn't change much over many
  • Webkinz is also an interesting example of a non-California
    centric product that is successful elsewhere. Perhaps MySpace is
    non-Silicon Valley, as another observation.

As always, suggestions and comments welcome!

If you end up
doing this analysis for other sites and find other patterns, please
comment and I'll link you to the bottom of this post via UPDATEs.

Written by Andrew Chen

August 13th, 2008 at 8:00 am

Posted in Uncategorized

What web product does this Google Insight map represent??

without comments

I was looking for a site where it'd be popular in Montana, Alaska, Dakotas, but not California. I hit the jackpot. I will post the answer as an update to this post…

Discuss ;-)

UPDATE: Too easy! Randal correctly guessed that it was Adultfriendfinder. I imagine the reason is that unlike California, there's fewer people in Alaska and Montana, and they use more dating/adult sites. That's the theory at least ;-)

Written by Andrew Chen

August 12th, 2008 at 7:03 pm

Posted in Uncategorized

Early adopters vs the Mainstream: Google Insights points out websites only used by Silicon Valley nerds

without comments

Intro to Google Insights
I have recently been playing around with the insanely useful Google Insights for Search product. You should definitely try it out if you haven't. It's basically Google Trends on steroids, and shows you a ton of data on any search you try. An SEO wizard's dream, basically. It's described as:

With Google Insights for Search, you can compare search volume patterns across specific regions, categories, and time frames.

Basically you put in keywords and it give you pretty charts.

Navigation searches and geo-location
One useful query to try is to search for your favorite website – like "" and specifically target it towards the US. It shows you a neat state-by-state breakdown of who is doing those searches.

Although unscientific, it tells you a bit about the location of the people who use the website, since logically the folks in states where the product is popular would tend to search for it quite a bit. Interestingly enough, Nevada, Hawaii, Oregon, and Washington are some of the top searchers for Gaia Online. Fascinating.

Now, another use for this is to figure out what websites are being used mostly by early adopters versus products that have broken into the mainstream. One could just go and search a whole bunch of domains and see what kinds of graphs are produced.

This is exactly what I've done below…

The graphs for Digg, Facebook, MySpace, Netvibes, Skype, Techcrunch, Twitter, and YouTube

Here they are…

Digg is pretty evenly used throughout the US, although there's a big hole in Montana, North and South Dakota, and Wyoming. Weird.

MySpace is mainstream and used throughout the US, although popular in California, Florida, and Vegas. Those are all places known for awesome club scenes, is there any connection? ;-) (er, SoCal)

Netvibes is California only – perhaps this is a good candidate for an early-adopter-only crowd?

Skype is interestingly close to Digg's geographic profile, actually. I'm sure the worldwide chart looks very different, but at least within the US there doesn't seem to be a crazy amount of penetration. Would be interesting to see this graphed against % of population that are immigrant populations, who are talking to their relatives overseas.


Twitter has a surprising profile – it's very strong within the states where it has any presence, but is basically dead outside of it. If you would conclude anything, you'd say that Twitter is in a transitionary period where it's certainly grown to be a larger-than Silicon Valley phenomenon, but is still mostly dominated by early adopters in specific states.

Facebook is mainstream, and unsurprisingly focused towards the east coast rather than the west.

YouTube is also quite mainstream, and looks like the MySpace profile.

Next steps
Basically this tool is a very interesting part of any internet analyst's arsenal, alongside Quantcast, Alexa, and the like. It gives a unique view of what's going on. You could do a lot with this also – I'm too lazy to do the full analysis, but you could run the entire Crunchbase database through this and see what sites have broken into the mainstream and which ones have not. You could also run the entire Quantcast top 100k site list though this.

Another interesting thing would be to translate the color shades into index numbers, and then calculate the sites with the highest variance of scores (Twitter would probably rank highly in this), which would indicate polarizing products. Similarly, products that were successful outside the Silicon Valley area might be interesting investment candidates for venture capitalists to look at.

If anyone does further work on this, please shoot me an e-mail and I will link you!

BTW, if you enjoyed this article, check out the list of 50+ essays on this site, related to viral marketing, metrics, go-to-market strategies, online ads, etc.

UPDATE #3: I did some additional work timeslicing the data for YouTube and Webkinz to show how they grew over time. You can read the new post here.

UPDATE #2: Some more interesting analysis, which shows the spread of Twitter over time, month-by-month, using the same tools. General trend seems like it hit the coasts and then filled in the central areas.

UPDATE: Techcrunch does some great analysis here – they basically point out that instead of querying for "" and "" I should have searched for "twitter" and "techcrunch." Good points and worth reading more.

Written by Andrew Chen

August 12th, 2008 at 8:00 am

Posted in Uncategorized

Virtual goods: Who will be the of virtual item sales?

without comments

Web 2.0 + games
As I've previously written about, I'm hugely fascinated in the use of virtual goods to monetize social web properties. In particular, there's an increasing intersection of games and social websites as well as the people who make them. Obviously there's a lot of cross-pollination going on, particularly in the heated Facebook social gaming space. Another space where a lot of innovation is happening is the casual MMOG market, where more than a dozen venture-backed companies are coming to market in the next 12 months.

Virtual goods infrastructure
One of the opportunities that is emerging, as these companies release their product and consumers start to have a standardized set of expectations, is the virtual goods infrastructure. The interface for virtual item sales is starting to standardize. In many cases, you see common elements like:

  • Your avatar
  • A collection of items which can be navigated through
  • "Buy" buttons and other interaction with each item
  • Regardless of whether you can buy it, you can dress up your avatar with aspirational items

Yet while the interface standardizes, the rest of the process does not. In particular, how about issues like:

  • Payment policies what countries, what currencies, and how many methods?
  • Alternative payment cards (like Nexon cards in Target stores)
  • User-to-user sales
  • Number of currencies (dual? single? more?)
  • etc.

In fact, one realization is that all of this backend stuff is really just the same type of infrastructure that e-commerce sites have built for ages. And looking at the resulting analogy, one might ask: "Who's going to be the Amazon of virtual item sales?"

Let's examine what this quesiton means… and where their techniques apply

Ultimately, much of Amazon's success is due to their quantitative views on personalizing the user shopping experience. This includes issues like:

  • Product recommendations
  • Price testing
  • Product bundling
  • Search, browse, and navigational hierarchy
  • Reviews, ratings, lists, and metadata
  • Affiliate programs
  • Ad targeting
  • etc.

Right now, the virtual item stores I've seen are in their infancy – there's very little sophistication relative to what's possible as shown by Amazon, at least in e-commerce.

My question: Will one particular game (or game platform) do this all right? Or will it be a third-party vendor that provides this rich set of functionality to partner games?

I'd enjoy any suggestions for companies that are looking into this space – comments and suggestions appreciated!

Written by Andrew Chen

August 11th, 2008 at 8:30 am

Posted in Uncategorized

How to start a professional blog: 10 tips for new bloggers

without comments

Starting from scratch

I started my professional blog in late 2006 as I was packing my bags and moving from Seattle to San Francisco. In the first month, I was pleasantly surprised to see a couple ex-coworkers subscribe via e-mail, and didn’t think it would ever lead to more readership than that. Two years later, I have a nice community of a couple thousand subscribers, and I occasionally get the question of “how would you start a blog, if you were to do it over again?” I thought I’d share my thoughts on that.

Here’s the quick summary, for those who want a quick skim:

  1. Carpet bomb a key area and stake out mindshare
  2. Take time to find your voice
  3. Stay consistent on your blog format and topic
  4. Just show up
  5. Go deep on your topic of expertise
  6. Meatspace and the blogosphere are tightly connected
  7. Embrace the universal reader acquisition strategies for blogs
  8. Come up with new topics with brainstorms, news headlines, and notes-to-self
  9. Look at your analytics every day
  10. Don’t overdo it

Extended discussion below…

Carpet bomb a key area and stake out mindshare
Like all products and markets, the blogosphere has its own set of existing products and channels. For example, if someone asks me about a “VC blog” I might refer them to Fred Wilson, Mike Speiser, Venture Hacks, and others. If someone asks me for a “games blog,” I typically recommend folks like Raph Koster, Nabeel Hyatt, Daniel Cook at Lost Garden, and others. The point is, just like companies, blogs tend to achieve 30-second elevator pitch status, and it makes sense to figure out a theme for whatever you’re going to be writing. IMHO, as long as the space you’re writing about is growing, you can never be “too vertical” since you’ll easily attract a couple thousand supersmart people who care passionately about your particular sub-vertical.

In general, I find people describing this particular blog as “the viral marketing blog” more than anything else. I write about a bunch of other stuff other than that, but people seem very interested in the topic so I’ll take what I can get ;-)

Take time to find your voice
As I said before, it’s good to find a key area. That said, it takes some time to get there, and I spent the first couple months switching between a couple topics – be it personal stories, product design, and advertising. It wasn’t until almost a year in that I started writing about viral marketing, which this blog is probably most widely read for.

I found that as I wrote more consistently, and learned from other bloggers, I began to change the tone and voice of my articles. While some of the key elements were always there – essays rather than links, certain topical themes, etc. – I added much better formatting within the blog posts, photos, linking to other blogs, etc.

Stay consistent on your blog format and topic

Related to finding your voice, it turns out that blog format really matters. To completely oversimplify, there seem to be two very different kinds of blogs that are successful. Either you’re a “curator” of news, or you’re a primary content source.

The curator is someone who blogs often and throughout the day with links and snippets, and I would consider someone like Robert Scoble (or iJustine!) to be the Michael Jordan of this approach. The style is often more conversational and casual, and includes lots of little updates on what they are doing or reading or trying out. These guys can really “cover news” and are widely read because they can provide the first opinion on new stuff coming out.

The bad news is that the curator model requires you to stay on top of things ;-) For a guy like me, with a full-time job and blogging as a dirty habit, being a pure content creator is much more appealing. I will never get the traffic of the news curators, but I can go deep on a specific topic and get a laser-focused audience that just cares about the topics I write about.

It’s obviously good to experiment and leave the door open for any and
all topics that interest you, but obviously once you begin to settle
and find your voice, it’s good to focus since then your readers will
know what to expect from you. There’s nothing worse than that guy that
writes one really good essay about the industry and then spends the
rest of the time writing about his dog!

Just show up
Hands down, the hardest thing about writing a blog is doing it regularly. I often just don’t have the energy to write, and have to consider it as a core part of my job in order to get it done. It’s especially true once you get past the first couple months and you’ve hit the top 90% of topics you wanted to get off your chest. Then coming up with new ideas is much harder.

In general, it seems like you have to maintain a tempo of at least 1 essay a week to be relevant. Any less than that, and people stop reading (or at least you’ll have all subscriber traffic and no one will just check your site). This blog is averaging about 1.5 posts a week, which I should probably work on, but it seems enough for at least some group of people to follow it. If I weren’t so lazy, I’d try to get at least 3-4 posts up per week, and possibly make them a little shorter. (Or one long one, and 2-3 news-related items)

Go deep on your topic of expertise
In general, I’ve found that you can never go “too deep” when covering a vertical. Some of the things I would have thought were the most esoteric – like viral loops and sharkfin graphs – have become the most widely read and widely linked posts. I originally hesitated to even post those since I thought they would be too obscure, but instead I found that people either appreciated it more. My guess is that it has to do with the fact that either they’re learning something completely brand new that they think is interesting, or at the very least you’re build “street cred” by not being the typical super-high-level analyst.

Meatspace and the blogosphere are tightly connected
Surprisingly (or unsurprisingly?) the world of San Francisco conferences, events, hanging-out, etc. are very much correlated with the blogosphere. A lot of readers of my blog either know me personally or know of me through a mutual friend – I get this sense since many of the inbound emails I get start with, “hi I read your blog and am friends with X” or I have lots of friends introducing me to FOAFs who want to talk about a topic from my blog. I think the point of this is just to say that you can grow your readership by being part of the conference circuit – either organizing or speaking or otherwise – and also your blog readership will lead to opportunities to get more visibility in meatspace. It’s all useful, so embrace it. And move to SF if you haven’t already ;-)

Embrace the universal reader acquisition strategies for blogs

When it comes to blogs, the user acquisition is pretty boring. You basically have the following sources of traffic, by importance:

  1. SEO, specifically Google
  2. Blog aggregators (like delicious, digg, etc.)
  3. Social platforms (like twitter, friendfeed, etc.)
  4. Individual blog links

Given that a lot of your blog traffic will come from SEO, it’s a good idea to try to own some keywords for a topic if you can. I get a ton of searches on viral loops and other viral marketing terms. It’s a good idea to add whatever your main keywords are to your blog title, blog topics, etc. I sometimes use blog titles like: “Facebook marketing: X” to get people to link back to me using those terms.

Similarly, whatever your expertise is, you might find vertical aggregators that drive a lot of traffic. For me, it’s Hacker News, Techmeme, and others. Identifying those key aggregators and submitting your articles is key.

Come up with new topics with brainstorms, news headlines, and notes-to-self
As mentioned above, it’s very very easy to run out of new topics to write about. Writer’s block seems to afflict me almost every week ;-) This is especially true once you hit the 2-3 month mark, since many of the topics that you might want to write about have already been covered in one angle or another.

My usual remedy to this is to employ a set of tactics that generate a healthy list of blog topics in my inbox, to be written one day or another. The first tactic is that when I’m in a good creative mood, I’ll often do a quick brainstorm of many potential topics and ideas. Some can be simple and explanatory, like “how to do X” or specific companies, or recollections of specific conversations that are worth blogging about. Similarly, I’ll also peruse sites like Techmeme and look for headlines that catch my attention. In particular, I often look for things that I think are either wrong, need clarification, or otherwise would compel me to rambling if someone told me about it in person. All of these ideas I will write up in very short outlines and e-mail to myself. Having a short outlin or starting a paragraph or two of the post helps me sketch out the idea in enough form to easily execute it at a later date. Otherwise, if you just have a fun headline but no body, going from 0 to 60 can be quite rough.

Look at your analytics every day
I look at the small amount of analytics on my blog on a frequent basis to understand what’s going on. It’s really nothing fancy, and certainly pales to the kind of instrumentation I’d do on an actual web project, but it’s good enough. More importantly, it helps you get some interaction with your passive users that aren’t leaving comments, and helps you figure out how to serve them better.

In general, I start by looking at my referrers every day, along with daily visitors/pageviews. I have sitemeter bookmarked on my phone so that I can glance at this all the time. I’ll look at what searches people are making via Lijit, and what searches are drawing people to this site. Another thing is to look at my Feedburner numbers and subscribers to see who is subscribing and how things are trending. I also get my top referrers and top content e-mailed to me on a weekly basis by Google Analytics, so I have an understanding of what people are looking at.

In general, in looking at this information I’m trying to assess a couple things:

  • Are there specific topic areas people are coming to the site for, that I should write more about?
  • Are there certain traffic sources that I should try to “develop” more? (Twitter is one good example of this)
  • What are the e-mail domains and companies that are visiting this blog, and how would I better serve those readers?
  • What are the searches people are making on the site, and are there any that aren’t returning any results?

The point of all of this is looking at your blog not as a “diary” as many people do it – instead of being fuly focused on yourself and what you want to write, you can think of your readers as your customer-base, and you’re trying to collect whatever knowledge you can to cater to their needs. Obviously, for part-time bloggers like myself, it’s important to balance your interests with the interests of your audience, but in general I think the philosophy holds.

Don’t overdo it
Finally, have fun ;-) After all, you can always quit! I often find myself not blogging for a week or two just because I don’t feel like it. I think that’s OK, since this isn’t my full-time job and I’m just doing it for fun. I think if I felt a lot more pressure to do this consistently, regardless of my enjoyment, I’d probably stop since it wouldn’t be fun anymore.

Written by Andrew Chen

August 11th, 2008 at 8:00 am

Posted in Uncategorized

Poll results for user acquisition: Viral, SEO, word-of-mouth, or other?

without comments

Recent poll
I recently had a poll on user acquisition and where people are putting their efforts. To summarize, most everyone put their focus on word-of-mouth and blogs, whereas less folks were looking at SEO/SEM/email/etc to drive their traffic.

Similarly, Facebook/OpenSocial was in a 3-way tie for 2nd place along with the other methods as well.

In general, I’m quite surprised! Here’s why:

Scalability of user acquisition methods
One my main considerations, when thinking about user acquisition, has to do with how well that method will scale over time. Ideally it’s a “set it and forget it” process (to quote informercial god Ron Popeil), where you can scale to millions of users after solving it once.

In a previous post called Why bloggers and press don’t matter for user acquisition, I point out how hard it is to scale the blogs/WOM gravy train, and I’ve also discussed why tech early adopters are often the wrong folks to target. I think in general I’m pretty surprised people are still going after blogs/WOM when it’s pretty obvious to me, at least, that it’s not a particularly great way to go.

I could be misinterpreting my own poll option, so let me know if I’m getting that wrong.

Written by Andrew Chen

August 4th, 2008 at 10:10 am

Posted in Uncategorized

7 ideas for billion dollar companies in the online ads

without comments

Given my previous Debbie Downer post on the advertising industry in a recession, I wanted to outline a couple major opportunities in the ad industry where huge companies could be built. This is a summary of many discussions that already out there, so it should be obvious to most folks in the ad industry but maybe new for everyone outside.

Here it is, not in any particular order:

  1. Online video
  2. User-generated content (blogs, social networks, etc.)
  3. Monetizing international traffic
  4. Vertical ad networks
  5. Cross-channel/cross-property targeting
  6. Continued securitization and efficient brokering of ad inventory
  7. New platforms and categories

Most of these may be obvious, but I’ll summarize all of them with a line or two of description. If I’m missing anything, please shoot me a note at voodoo [at] gmail and if I agree I’ll add it ;-)

Online video

Obviously video is a huge category since it’s the clearest metaphor to TV advertising, and no one really has cracked it. What’s the opportunity for folks outside of the big branded publishers? What are the key metrics? What’s the right ad unit? (hopefully not pre-roll)

User-generated content (blogs, social networks, etc.)
The buzz and incredible word-of-mouth for blogs and social networks is undeniable, but in order for this opportunity to scale, a saleable ad unit needs to be determined. This ad unit needs to be priced, measured, and otherwise standardized. What will it be? Is a standard of profile + news item + widget + etc emerging? Will blogs be treated similarly to social networks, or are they considered too different?

Monetizing international traffic

Many social networks and ad networks attract a tremendous amount of international traffic, much of which monetizes poorly because of a lack of direct relationships with advertisers in the source countries (and also ad spend internationally). Sometimes it’s akin to throwing away 30% of your pageviews because the 30% isn’t monetizable to a comparable degree. Who will solve this? Who will own the infrastructure to broker these transactions?

Vertical ad networks

As content becomes increasingly separated from distribution online using widgets, syndication, and aggregation, more and more of branded content exists outside the destination sites of branded publishers. Many of these guys are now organizing vertical networks to take advantage of larger, trusted networks. Is this model here to stay? How expansive will the influence of branded publishers be relative to the hustle and bustle of long tail publishers?

Cross-channel/cross-property targeting
As the amount of content and products increase online, user attention is getting fragmented – the same person that browses WSJ at a $60 CPM goes to their email service and is served an ad at $1 CPM. Similarly, the same person who is shopping for a car on Google is priced at $10 a click, but the same search elsewhere yields terrible monetization. Bridging the small areas of “high-context” online versus the larger swaths of high-engagement/low-context activities online will become an increasingly large opportunity. How will companies price, organize, and otherwise securitize this data? What will consumers think of their data getting aggregated? Will one company win, or will it be an open exchange? Or a co-op of similar sites?

Continued securitization and efficient brokering of ad inventory

As ad networks become more technologically advanced, allowing APIs, automated campaign management, a multitude of targeting options, the ad inventory of the world is becoming more efficient over time. (Though never efficient enough!) As exchanges, arbitrageurs, networks, and ad infrastructure become more mature, how much of the finance world can be directly applied? Will there be an organized futures, options, and derivatives market? Who will be the Bloomberg, the LTCM, the Instinet? What infrastructure hasn’t been built for ad industry that does already exists in the finance world?

New platforms and categories
And of course, the big wild card is that the available inventory online is constantly shifting as new consumer products are being created. If a new category of websites is created, like social networks or widgets in recent years, some of these opportunities have new and different advertising characteristics than others. Some may have extremely high CTRs, or capture the imagination of brand marketers.

For example, the ad market within games is still very much in flux, and who knows what standards will emerge there. Similarly, mobile continues to be a large opportunity that may open up as Android, Symbian, and the iPhone evolve.

Have a billion dollar ad play that’s not in here?
Email me cuz I’d be interested in hearing about it ;-)

Written by Andrew Chen

August 4th, 2008 at 8:00 am

Posted in Uncategorized

Quick link – NYT article on a “wow” product

without comments

This recent article on the NYT discusses the site called
If You Run a Red Light, Will Everyone Know?

After playing around with CriminalSearches for a couple minutes, I’m honestly wow’ed by how much data is in there. Basically it provides a free, searchable list of peoples’ names and all the stuff that’s on their rap sheet. I’d encourage you to try a couple searches to the comprehensiveness of this data.

This also reminds me of the power of Crime Maps to show you what’s going on in your neighborhood. In my neighborhood which is close to Bernal Heights / Mission, there are an incredibly large number of sex offenders, violent crimes committed, and other things that look horrific when plotted in Google maps next to my apartment!

Anyway, here’s a scary idea: If you cross-referenced your Facebook friends list with the CriminalSearches site, what would you get out??

I’d call that site ;-)

Written by Andrew Chen

August 3rd, 2008 at 2:06 pm

Posted in Uncategorized

50+ essays on viral marketing, social network monetization, product design and more

without comments

I finally decided to make a page listing all the reasonable essays I’ve written over the last 2 years on one page, which you can find here.

For the lazy folks, here’s a snapshot of the essays as of today (July 30):

Viral marketing and user acquisition

Engagement and product design

Online advertising and social network monetization


Media and games

Entrepreneurship and startup life in San Francisco

Written by Andrew Chen

July 30th, 2008 at 8:00 am

Posted in Uncategorized

Google Analytics report for top referrers to Futuristic Play

without comments

I have a Google Analytics report on my top referrers (and search keywords) mailed to me on a weekly basis, and I thought I’d share the results below. Note that this is visitor-only traffic, and doesn’t include RSS feed readership. Either way, blogging continues to be a very high effort-to-pageview ratio, is very inefficient, as I mentioned in a previous post Is blogging worth it? What’s the ROI?

As might be expected, the top traffic comes from a few sites:

  • Techmeme
  • Google
  • Twitter
  • Friendfeed
  • Hacker News

In particular, the top 5 make up the vast majority of the traffic, as one might expect from seeing the Power Law curve all over the place.

I personally find that Twitter and Friendfeed being high on the list is kind of a funny thing. It’s one of the reasons why I’ve started to get more active on Twitter, since at least for my particular audience, people seem to use it quite a bit. (Follow me on Twitter if you already haven’t)

The other interesting point is that a bunch of my traffic comes from aggregators in the form of Hacker News, Techmeme, Google, and others. Excepting A-List bloggers like Robert Scoble, these guys tend to drive more traffic than individual blogs – they can also be more consistent in driving traffic from week to week.

Google Analytics for Futuristic PlayUpload a Document to Scribd
Read this document on Scribd: Google Analytics for Futuristic Play

Written by Andrew Chen

July 30th, 2008 at 8:00 am

Posted in Uncategorized

POLL: Where do you put most of your effort, for user acquisition?

without comments

Written by Andrew Chen

July 29th, 2008 at 8:00 am

Posted in Uncategorized

Google AdSense FAIL

without comments

That’s funny:

From the incredibly funny FAIL blog. (unsurprisingly, run by the same guys that run ICanHazCheezburger)

Written by Andrew Chen

July 28th, 2008 at 10:49 am

Posted in Uncategorized

AdAge article on agency perspectives of online versus traditional advertising

without comments

Around the bay area, I hear a lot of derogatory comments about brand advertisers and agencies about how they just “don’t get it,” how “TV is dead,” and other assertions. That may be true one day, but the traditional market is still quite big and it’s an uphill battle today.

There was recently a great AdAge article called $80 Billion? Online Display Market Is Being Overhyped that I want to pull some quotes from.

On the perception of where online ad spend fits into the overall marketing budget:

The inconvenient truth is that for all its new-media spin, display
advertising is “old” media — a commercial message to be placed next to
editorial or entertainment content. And we know by now that
measured-media growth has pretty much ground to a halt as marketers
continue to increase their dollars in unmeasured disciplines such as
web development, public relations and database marketing at the expense
of paid advertising. Ad spending among the top 100 U.S. advertisers
last year grew a paltry 1.7%, with measured media only up 0.3%.
Measured-media spending is in decline in Japan, and it’s not much
better in the U.K.

On how much money, percentage-wise, that brand advertisers spend online:

The question is: Should the fact Procter & Gamble spends only 1.5%
of its marketing budget on display ads be viewed as a warning signal by
online ad sellers, or as an opportunity? (Even Unilever, Ad Age’s
Digital Marketer of the Year, spends little more on display, allotting
it 2% of its budget.)

On their perspectives on other media channels to reach consumers:

Of course, part of why large companies such as P&G spend so little
on the web is because of the feedback they get from the marketing-mix
models they still use to determine media outlays: TV and other old
media still work. (P&G increased its magazine budget by 7% last

Ultimately, for startups that are trying to tackle the advertising space, the reasons above are a lot o the reasons why it’s hard to change agency behavior – either to make them use your self-service interface, or to use a different set of metrics to evaluate their spend, etc. The amount of dollars they’re spending online, while often strategic and very much growing, is simply not large enough to justify redoing their entire business model.

Read the whole thing here.

Written by Andrew Chen

July 28th, 2008 at 8:35 am

Posted in Uncategorized

omg I’m just a startup, I can’t do those fancy analytics!

without comments

Reader comments on user retention post
In my previous post on user retention strategies from the catalog marketing world, I received a number of comments commenting on the difficulty of implementing analytics systems when you don’t have the resources of a F500 company.

Another Andrew Chen writes:

I agree with the general sentiment of the post, but how would you suggest a startup deal with the problem? Without the marketing budgets of larger companies and without the time to do detailed analysis of its customers, how can a startup find the best messaging and segmentation?


Similarly, Rachanap writes:

RFM is a good idea in theory, however, even in traditional marketing world, like in CPG, companies don’t have the bandwidth to adopt it, let alone successfully implement it. Implementing such a truly effective and scalable marketing system needs a tool in place for segmenting the data, an then group the users (often needing expensive software) as well as additional infrastructure to capture that data, neither of which is an easy task. For startups this would be difficult, given their low bandwidth.

In essence, the question is, how do you take advantage of all of these different analytics systems given the constrained resources of a startup. Good question. Let’s tackle this below.

The role of analytics in startup decision-making
In general, a philosophy on the role of analytics within a startup is:

If you’re not going to do something about it, it may not be worth measuring.

(Similarly, if you want to act to improve something, you’ll want to measure it)

Don’t build metrics that aren’t going to be part of your day-to-day operations or don’t have potential to be incorporated as such. Building reports that no one looks at is just activity without accomplishment, and is a waste of time.

So instead, I recommend a “layers of onion” approach on figuring out what analytics you require. Early on, your goal may be to focus on creating a solid product that people like and stay engaged on. Obviously you could slap in Google analytics, but at the very least I’d recommend cohort-based analysis to get an actual feel for how well your products are retaining people. Similarly, you might reach a point where you want to be focused on traffic, in which case you’ll want to make sure you properly instrument to capture your viral loop. Later on, you’ll want to do the same for your ad inventory, to figure out what segments of audience and what sections of your site monetize the best.

The point is, develop the necessary metrics alongside whatever feature development that makes sense, don’t do any more than you need. Now when you are far enough along that segmenting your users based on behavior matters – likely this only because relevant once you hit a certain user acquisition threshold – then it may be important to implement RFM-based segmentation. Or not. It just depends on your product goals.

Metrics as a “product tax”
In fact, one way to view analytics is that they are a double-digit “tax” on your product development process because of a couple things:

  • It takes engineers lots of time and development effort
  • It produces numbers that people argue about
  • It requires machines, serious infrastructure, its own software, etc
  • Fundamentally, it slows down your feature development

As a rough estimate, I’ve found that it takes between 25-40% of your resources to do analytics REALLY well. So for every 3 engineers working on product features, you’d want to put 1 just on analytics. This may seem like a ton (and it is), but it throws off indispensible knowledge that you can’t get elsewhere, like:

  • Validating your assumptions
  • Pinpointing bottlenecks and key problems
  • Creating the ability to predict/model your business to make future decisions
  • It tells you which features actually are good and what features don’t matter

Question: Is it better to build 10 features where you don’t know what worked and what didn’t, or is it better to build LESS features but have a clear sense for what and why something worked? In my opinion, you want to learn as much as you can so you can “run up the score” on the features that work.

Prioritizing metrics development
The key to this philosophy is figuring out how to prioritze the metrics that you build relative to your features. Again, you want to only develop the analytics that you need to build out your product correctly – no more, no less.

In fact, it’s often wise to make a distinction between Operational reports versus Investigative reports,
each of which have their own goals and structure. The operational reports should reflect all the issues you care about in your business, and the investigative stuff generally satisfies curiosity (which is always a good thing).

So for folks with startups, I’d encourage you to ask: What are the outcomes that you care about? What are the assumptions you are making? If you’re a startup that’s spending copious amounts of time building features X, Y, and Z, what are you making the assumption of? That these features will make people want to extend their engagement times on the site? That these features will build word-of-mouth? What are the metrics that would prove you right or wrong in your assumptions?

Let’s discuss an example of a sample roadmap for a online photo-site.

Example product roadmap
I’ll close out this post with an example roadmap for a product that roughly had the same featureset as You may end up making these product decision out-of-order of what I have them, but the philosophy remains the same: Build analytics only when you need them, and align them to the key efforts behind your product development process.

Here’s an example sequencing of product development:

1) Core product and features

  • Obviously, you want people to be able to upload a photo and give a URL for it
  • These photos then can be displayed on other sites

1a) Analytics for core product

  • First add Google analytics since it’s free :-)
  • Build out a cohort matrix that tells you how many users join each week, how many come back, and what they do during their visits
  • Where are these images being uploaded to? A breakdown of image displays by the domain they were displayed on would be nice

2) Working on growing traffic

  • Once you have the core product done, perhaps you want to make some whizbang features to grow traffic
  • You might make a photo slideshow, or some awesome photo effects, etc.
  • Also, you might want to make one-click integration with Typepad/MySpace/etc via something like Gigya

2a) Analytics for traffic growth features

  • One key issue to measure is how much additional traffic is gained by these different methods. You may want to start annotating visits that come to your site based on what kind of widget or photo created it. Do sparkly photo galleries generate the most visits? Why? Make a report
  • Second feature might be an A/B testing framework. If you divide two groups, A and B, and show them different featuresets to see how the traffic grows for each



UPDATE: Thanks to Jeremy for the link and additional thoughts on spec’ing out metrics in the product process. Check them out.

Written by Andrew Chen

July 28th, 2008 at 6:00 am

Posted in Uncategorized

Hello world! (from Disqus)

without comments

Hello all from Seattle!

After much procrastinating and complaining about Typepad’s yearly subscription, I finally decided to pay for the Pro account and try out Disqus. We’ll see if it’s worth the money and upgrade.

I think for a lot of the products out there pitched at content creators – Google Knol being the latest one – there’s a misplaced emphasis on making money. As a blogger, I don’t really care about making a buck, since it’s not my full-time thing. I just want to write what I want to write and see what people think. Products like Disqus, Lijit, MyBlogLog, and other blog-related products I like all cater towards my interest in getting more interaction from my audience and knowing that they aren’t just anonymous web visitors. I think it’s important for folks working on these kinds of products to never forget that.

Anyway, I’ll have more essays later next week when I’m back in San Francisco. Talk to you guys then.

Written by Andrew Chen

July 25th, 2008 at 11:58 pm

Posted in Uncategorized

Experiment: Will adding @andrew_null to my blog title lead to more Twitter traffic?

without comments

I’ve gotten more active on Twitter recently – you can find my profile here.

I recently noticed on Summize aka Twitter Search that there’s a lot of people who blog the full name of my blog. You can see the results for a search on “Futuristic Play” here.

Thus it inspired me to try an experiment :-)

If I add @andrew_null to the name of my blog, will all the auto-linking that sites/apps do link more users back to my profile? We will find out!

If anyone else tries this out, let me know and I’d be interested in learning the results!

Written by Andrew Chen

July 23rd, 2008 at 11:33 am

Posted in Uncategorized

Virtual economy for notification “credits” on Facebook apps

without comments

Preferred program for Facebook apps?
There’s been some recent discussion around the tiered system that’s going to be put into place for Facebook apps to give them the right incentives to be a good citizen within the Facebook economy. I think this is a great start, and you’ll have plenty of apps and startups that are likely to want to participate.

That said, my thought would be that it doesn’t significantly change the dynamics on the platform since a human-based solution to the problem will inevitably cause significant gray area. In my estimation, the major issues around app behavior on Facebook all pretty much center around one thing: How much each app uses and abuses “viral channels” within the platform, which degrades the user experience.

Maximizing local happiness versus global
The problem is that each app, working in its own world, will NOT pursue the global maximum for user happiness. They simply don’t know what the other apps are doing. What this creates is a classic tragedy of the commons situation, where the commons is your Facebook newsfeed/notifications/e-mail inbox.

Proposal for a Facebook notifications point system
So here’s my proposal:

  • Each user session on an app is scored for engagement – meaning, pageviews/time-on-app/revisits and any other metrics that would be useful
  • The app creator then gains “points” for creating more engagement with their users
  • They can then use these points for different kinds of notifications. Some notifications that are more obtrusive cost more points (like emails), and others which are less annoying cost less

Then from a global context, Facebook can regulate the end-user experience by figuring out if their users are getting too many e-mails or too many notifications. If so, they can just raise prices so that notifications are more expensive to buy.

More analysis needed
I’ll be the first to admit I haven’t thought through all the implications of a system like this. In particular, here are a couple issues worth thinking about:

  • What are all the point-generating activities? Engagement, positive reviews, % retained install rate?
  • What are all the point-costing activities? Sending notifications? Uninstalls? Report as spam?
  • Will a secondary market emerge? Is it a good idea to let people pay for notification points?
  • What are the ways people will cheat the system? How do you counterbalance that?

Suggestions and comments welcome :)

Written by Andrew Chen

July 23rd, 2008 at 11:18 am

Posted in Uncategorized

Recency Frequency and Monetization (RFM): Optimizing your notifications strategy

without comments

Notifications strategy

Every startup has to think about their notifications strategy – it’s not enough to blast out e-mail whenever anyone on your site does anything, or to send out weekly newsletters. Notifications are an important part of how you manage your userbase’s lifecycle, and is a key part of any retention effort your business has.

I’ve often thought that online advertising folks are pure-plays for user acquisition. And that (traditional) games people are pure-plays for engagement. The question is, what’s the pure-play for retention? I’ve come to believe that businesses that focus on managing churn – be it telecom/wireless, catalog, financial services – are the guys who are the furthest ahead in understanding these concepts.

Learning from the catalog marketing world
I’ve recently been doing research in the retail and catalog marketing part of the world, where companies have thrown billions into understand when it’s a good idea to mail someone something and when it’s a bad idea. Unlike the online world, there’s generally a cost associated with sending out a “notification” and there’s usually a high LTV attached to each person, so the stakes are much higher.

Traditionally, companies in the catalog marketing world use a segmentation system called RFM:

  • Recency
  • Frequency
  • Monetization

Basically every user in your database is graded from 1 to 5 across each variable from a quintiles basis, where 555 is the best.

The idea here is that the higher the numbers, the more likely the recipient is to respond to an offer. Thus, 555 customers are more likely to mail back something than 455, and Recency also plays a higher factor in response than the Frequency of that customer. (Which is why it’s RFM and not FRM)

But how do you know when someone would have responded anyway?
An interesting problem then occurs because what if all of your 555 customers would have responded anyway? Then you’re sending them catalogs but it’s pointless. I want to point at a great paper on “uplift marketing” that answers exactly this question. I attached it as a Scribd embed below, but if you’re reading this in an RSS reader, you’ll have to click through to this blog.

Anyway, the authors from a company called Stochasic Solutions divide the world into 4 groups:

  1. Persuadables
  2. Sure Things
  3. Lost Causes
  4. Sleeping Dogs

The idea is that Persuadables are the ones who are affected by your marketing (and notifications), whereas Sure Things would have embraced your products anyway.

Then the Lost Causes are never going to respond, no matter how much marketing you send their way.

And finally, an interesting group are the Sleeping Dogs, who are likely to use your product but as soon as you’re reminding them (especially too often), then they are likely to quit your service. Ouch. (Gym memberships anyone??)

The point is, the Persuadables are the ONLY group who can improve your business. Everyone either costs you dollars by expending marketing budget that didn’t need to get spent, or would quit your service upon remembering they ought to cancel.

Don’t assume that more notifications is better
I’ll let you guys read the document yourself, but I think the overall point is that first off, segmenting your audience based on their responsiveness is a smart idea. You’re able to figure out key variables that drive their participation, which leads to some nice strategies on the retention side. On the other hand, don’t forget that the Persuadables are the true audience that can be retained via notification. Everyone else is negative, and if you’re overly aggressive, it’s easy to turn those Persuadables into Sleeping Dogs.

The full document below:

Generating Incremental SalesUpload a Document to Scribd
Read this document on Scribd: Generating Incremental Sales

Written by Andrew Chen

July 23rd, 2008 at 10:52 am

Posted in Uncategorized

Are Web 2.0 startups wasting their time with Web 2.0 early adopters?

without comments

Cookie cutter go-to-market strategies

I run into a lot of startups that have identical strategies for getting to market. When you talk to them, a lot of them will talk about the same questions:

  1. When are you launching?
  2. What do you do to get on Techcrunch/Venturebeat/DEMO/etc.?
  3. How do you think Twitter/Friendfeed/etc got their successful launches?

I think there are some big assumptions that go into the questions above, which we’ll discuss in this post. But let me first show you a very famous diagram.

The traditional “Crossing the Chasm” curve

The “technology adoption” curve above was popularized by Geoffrey Moore’s Crossing the Chasm book, and mostly deals with enterprise software. (I had the pleasure of spending some time with him while I was at Mohr Davidow, where he is a venture partner)

People often have the same image in their minds when thinking about go-to-market strategies for consumer internet as well. The general idea is to go through the typical flow:

  1. Get a bunch of early adopters (aka Techcrunch readers) excited about your product
  2. They blog, twitter, and promote your product to their friends
  3. Eventually this process will reach the mainstream and you’ll get the wider market

I also want to bring up the definition used by Moore to describe a market, which is a bunch of folks that reference each other when making purchasing decisions. In this case, it’s the bloggers and alpha nerds that are the market.

But what if this is NOT the starting point for your market? Who are the other early adopters and visionaries?

Redefining enthusiasts and visionaries
The major point I will make here is that Techcrunch and related blogs reach an audience of early adopters, but these may not be the earlier adopters that you want. After all, what’s the point of launching a music startup on Techcrunch, for example, if your startup is primarily for the teen mass market?

I want to point out a great article posted by Malcolm Gladwell (of Tipping Point) years ago, called The Coolhunt, which you can read here. The article discusses the emergence of bottoms-up “cool:”

Once, when fashion trends were set by the big couture houses-when cool
was trickle- down-that wasn’t important. But sometime in the past few
decades things got turned over, and fashion became trickle-up. It’s now
about chase and flight-designers and retailers and the mass consumer
giving chase to the elusive prey of street cool-and the rise of
coolhunting as a profession shows how serious the chase has become. The
sneakers of Nike and Reebok used to come out yearly. Now a new style
comes out every season. Apparel designers used to have an
eighteen-month lead time between concept and sale. Now they’re reducing
that to a year, or even six months, in order to react faster to new
ideas from the street. The paradox, of course, is that the better
coolhunters become at bringing the mainstream close to the cutting
edge, the more elusive the cutting edge becomes. This is the first rule
of the cool: The quicker the chase, the quicker the flight. The act of
discovering what’s cool is what causes cool to move on, which explains
the triumphant circularity of coolhunting

Where Gladwell uses the term “cool,” the people in the technology industry use the phrase “early adopter.” Within every target market, no matter how mainstream, there are early adopters or “cool” people who are more likely to uptake these new products and have friends reference them for decisions.

Where are your “real” early adopters?
And thus as a corollary, if your market is moms, there are cool moms that are likely to try out the new technology. And if your market is Asian immigrants, there are cool members of that group who are trying out new technology.

My point is simply this:

In 99% of all cases, the Techcrunch early adopter crowd is probably NOT the ideal early adopter crowd to go after – your target market lives somewhere else

The exceptions I’ll make to this are B2B tech startups like Gnip, or companies primarily trying to target VCs in their announcements.

So where are your early adopters you want to be going after? I don’t know, but that’s the kind of research that you should be doing to create a compelling, differentiated go-to-market strategy that anything but cookie cutter.

One last thing to think about… where did LOLcats and the FAIL meme start? Start by reading about 4chan. Also check out SomethingAwful, Genmay, Encyclopedia Dramatica, and all the other sites that expose some of the darkest underbellies of the internet ;-) Between these sites, Yahoo Groups, IRC, and other old-school communications platforms, I’m guessing a large number of internet memes are generated there by early adopters of some kind.

Written by Andrew Chen

July 21st, 2008 at 7:32 am

Posted in Uncategorized

More gloomy news on the advertising recession

without comments

Silicon Alley Insider sums it up:

Ouch. Better start turning those widget companies into virtual goods startups…

Written by Andrew Chen

July 19th, 2008 at 9:39 pm

Posted in Uncategorized

I summarize Foo Camp 2008 activities so you don’t have to

without comments

Attending the O’Reilly Foo Camp
This last weekend I attended Foo Camp, which is a yearly O’Reilly event. It was a lot of fun because of the crowd – an eclectic mix of artists, philosophers, geeks, writers, and business folks. Special thanks for Dave McClure for letting me hitch along in his car and connecting me to the O’Reilly folks. Also, Bryce from OATV was kind enough to let me crash some of the Startup Camp sessions.

I wanted to link to a couple of the blog posts and photo galleries out there about the event:

Summaries of Foo Camp activities
I went to a large number of sessions – there were something like a dozen rooms with 10 sessions throughout the day, so over 3 days there were a ton of speakers and activities.

Anyway, here are some summaries of what went on:

Camping at the conference
Grab a spot to sleep in the O’Reilly offices – it’s cold and wet outside

Session on distributed social networks
Everyone at Foo wants data portability, companies resist it, and consumers don’t care

Session on economics not being a branch of physics
Lots of borrowing + forcing banks to mark-to-market the value of their assets + borrowing restrictions that are depending on assets means that banks catastrophically dump their assets in a spiral of death if anyone in the ecosystem has any problems

Showering facilities
With a half dozen showers at the conference and 350+ attendees, the only time to shower is at weird hours, like 2 or 3am

Session on granular measurement of carbon footprint
We’re all screwed unless we reduce our carbon footprint to 2500 somethingaruther

Session on Playboy’s existential crisis
 Hugh Hefner got brought back as part of the playboy brand because of Viagra, but when he dies the brand will get hurt unless they appoint George Clooney as the next Hugh Hefner, kinda like James Bond

Day 3 of the conference
With a half dozen showers and 350+ attendees, a lot of people seem to choose to not shower

If someone has a grin on their face when they are telling everyone else they’re a villager, it means they are actually a werewolf since no one could be that happy to be a villager

Session on the future of news
Journalism is in danger because consumers want celebrity gossip and they won’t pay more for investigative reporting – Rupert Murdoch is to blame and Seymour Hirch is the best writer ever!1!!1!!

Session on hiring great people
No one in san francisco can hire anyone locally, it’s all about importing smart people from outside the city

Session on corporate sing-a-longs
No matter how many people you invite to attend a corporate sing-a-long session, none of them will sing

OK, please tell me if I missed anything ;-)

Written by Andrew Chen

July 16th, 2008 at 6:11 pm

Posted in Uncategorized

In Seattle from the 25th to 30th – who wants to meet up?

without comments

I’m going to a wedding in Seattle next weekend, and will stick around a couple days to catch up with people. If anyone would like to meet up while I’m back there, shoot me an e-mail.

In particular, I’d enjoy meeting:

  • smart engineers and future entrepreneurs
  • advertising people (agency, ad network, etc)
  • database marketing and stats folks
  • startup entrepreneurs, especially the ones who are looking to go BIG! ;-)

Shoot me a note at voodoo [at] gmail!

Written by Andrew Chen

July 15th, 2008 at 4:15 pm

Posted in Uncategorized

Online advertising during a recession: 5 key trends for ad-based startups

without comments

One of my top 25 search queries is “recession advertising” so I thought I’d expand on this topic a little bit, since important parts of the economy continue to implode and the folks who are thinking about business models should be worried.

Ultimately, the dynamics here are complex and uncertain, but here some of the key trends worth watching if you’re an advertising-based startup:

  1. Accelerating movement of offline to online ad spend
  2. Brand areas weak, direct response will be less affected
  3. Weak areas to watch: Video, social networks, communication, etc.
  4. Rise of direct-to-consumer revenues?
  5. Timing is everything

Let’s dive into these topics more below…

1) Acceleration movement of offline to online ad spend
The first key issue revolves around the fact that advertising spend is already shifting online from other types of media. In the direct response sector, classifieds are obviously moving from newspapers to services like Monster and Craigslist. In brand advertising, dollars are moving from TV onto high-quality publishers on the internet. An article from AdAge last year articulates this theory:

Many analysts now agree that when marketing budgets come under pressure
in a stressed economy, those sectors that can best document their
connection to ROI, such as search-engine advertising, are far more
attractive to corporate chiefs than other kinds of less-trackable
traditional advertising.

The point is, when your marketing wallet shrinks yet the market gets even more competitive, then companies in crisis will start incorporating methods other than the tried-and-true. Even though a lot of brand-based advertising has horribly opaque measurements – clickthroughs, surveys, and other gross metrics don’t provide much – it’s still better than the TV ad sales guy who asks for huge upfronts without providing much in transparency.

So consider this movement of dollars from offline to online a big plus for any ad-based startup.

2) Brand areas weak, direct response will be less affected
Of course, as the quote above alluded to, the strength of a company’s online advertising revenue has a lot to do with the kind of advertising that the company enables. For companies that are focused purely on brand advertising, there will still be hits in budget as the typical reactions – a flight to quality, a flight to metrics – affect brand-oriented startups.

So if your world is focused on engagement, eyeballs, and branding opportunities, things may still get worse before they get better.

On the other hand, the more transactional and close-to-the-money your company is, the more you can expect your revenues to grow and maybe even thrive during this time. These include companies in the following lines of business:

  • ecommerce
  • search
  • classifieds
  • shopping comparison
  • remnant ad networks
  • lead generation
  • product reviews
  • etc.

No matter how bad things get, from a relative standpoint the above businesses are still much better than direct mail, yellow pages, newspapers, and the like. The competitive pressure may make these industries shine – and Google will only get stronger!

3) Weak areas to watch: Video, social networks, communication, etc.
Unfortunately, some of the weakest areas for online spend during a recession are also some of the hottest spaces for startups right now. In general, startups based in video, social networks, and communication applications are some of the most brand-dependent companies out there. The problem is that generally, they have a hard time monetizing pageviews because users aren’t in a buying mindset when using the products.

Because of this, you need to be at a critical mass point to be relevant to agencies – and of course, this bar can be expected to rise over time in the case the economy is sputtering. Why spend a dollar with a no-name publisher when you can buy premium inventory for relatively cheap CPMs?

This is not to say that there won’t be significant opportunities in this space. For example, I remain quite bullish on web properties like MySpace and Bebo, even as they’re brand-focused, because they are attached to organizations that know how to sell brand-based advertising. Similarly, the trends in vertical ad networks provide an interesting opportunity for startups to partner with more established media companies to drive higher revenues as well.

4) Rise of direct-to-consumer revenues?
In the case of a long period of recession, another key opportunity will be for brand-oriented properties to transition their businesses into direct-to-consumer opportunities. Does it surprise you that YouTube is looking into affiliate-based revenue ideas? Or that Slide is thinking about direct-to-consumer opportunities as well? These are smart folks, and they understand that unlike brand advertising, if you can get direct monetization to work, it’s stable, scalable, but just very very hard.

And of course, virtual goods fits into this as well, but you all knew that.

The difficult part about these approaches is that unlike ad-based models which allow you to monetize 100% of your audience in one fashion or another, transactional revenues can usually only squeeze cash out of 1-5% of your audience – so what do you do with the rest of them? Are they just loss leaders?

Similarly, it seems that the best transactional revenue models have to be “productized” into actual features within the web property. Building a virtual goods infrastructure is not an easy task, nor is it simple to convince users that all those digital bits actually have value. It’s not something that’s as easy as copying-and-pasting some Javascript code onto a page to display ads.

5) Timing is everything
And finally, I’d like to close this post with the observation that in the advertising world, particularly in new media channels like online advertising, timing is everything. The brand-oriented web properties that exist today were built in the 2003-2005 era, when brand advertising wasn’t so healthy. Similarly, Google was created during a period where online ads was out of vogue, and they had to figure out a model that works.

For the new startups that are building their business plans from scratch today, I think there remains tremendous opportunities in the advertising-supported model. It pays, as many investors can attest, to be counter-cyclical. Perhaps the startups being incorporated this year who reach scale 3-4 years from now will be the ones that really kill the TV ad market by doing things we can’t even imagine today.

Good luck! I’m going to YPulse tomorrow – Foo was great – and will be back blogging in no time.

Written by Andrew Chen

July 14th, 2008 at 7:30 am

Posted in Uncategorized

Off to conferences for the next few days…

without comments

I’ll be at Foo and YPulse over the next week or so. If anyone else is going, please give me a shout and we’ll meet up! In the meantime, expect little to zero blogging.

If I’m not too lazy, I may log into twitter or blog a bit about the events.

You have Twitter too? If you’re feeling bored, you can follow my tweets here. Or the main Foo conference chatter is here. (Note that with the 140 character restriction, it’s hard for me to type out my usual essays!)

Written by Andrew Chen

July 10th, 2008 at 12:49 pm

Posted in Uncategorized

Yahoo’s BOSS program doesn’t go far enough: Why not open up Yahoo search traffic?

without comments

Yahoo BOSS is a neat toy for mashup fans, but doesn’t help where you need it to
There’s recently been a spate of articles about Yahoo’s cutely named BOSS program (Build your Own Search Service) at Cnet, Techcrunch, GigaOm, and others. Techcrunch’s headline is that “Yahoo Radically Opens Web Search With BOSS.”

Yet when I read more about the program, it seems pretty bland IMHO:

BOSS allows developers to submit queries (and their associated
parameters) via an API to retrieve up to 50 web, image, news, or
spelling results in XML or JSON format at a time. Per Yahoo’s policy,
developers will be required to display its ads next to, or within,
their results (although this requirement won’t be imposed until later,
Yahoo plans to offer CPM fees as an alternative, and academics will be
exempt from any such attempts at monetization completely).

To me, this seems like a neat little service, but it won’t change the world- this will just let companies do neat search mashups. Given all the hype around APIs and mashups in recent years, can you think of one mashup company that was commercially successful? I challenge the readers of this blog to name some examples ;-)

The problem is, how will the companies that implement this cool technology end up with any traffic? Seems like this program is a recipe for a bunch of neat PR-generating techie projects without real traction, which is arguably not what Yahoo needs right now.

Search engine result pages as Platforms

I mentioned in a previous blog post on Google’s second click that ultimately, the search engine results page is the “platform” on which these portals build. After capturing the first click, via search, they can start building out products to capture subsequent traffic, for example driving address queries into Google Maps. The power in this approach is that huge amounts of traffic go through the SERPs, which then drive traffic to related properties.

Now compare this approach to what Yahoo is offering through BOSS, which lets you harness their infrastructure but doesn’t provide any traffic to back you up.

Opening up Yahoo Search traffic, not just the APIs

The extreme approach – well not even that extreme these days, given Facebook – would be to let developers build extensions to the search engine that actually run on top of the * domain. They can provide an API, do app approvals, and direct only small bits of traffic to each app to test them out – then ramp up the ones that perform better than anything else. There are difficult pieces necessary to make this work, but if done well, it has the potential to change the search game by letting developers target small groups of queries the way that advertisers have been able to.

Maybe in the scheme of things, this is too risky of a move for Yahoo at a time when they are focused on smoothing out revenue growth so everyone can keep their jobs ;-) So perhaps a better player to try something like this out would be Ask, Looksmart, or even Microsoft!

Thoughts and suggestions welcome…

UPDATE: I was pleasantly surprised to hear that the ideas above are implemented in early form by Yahoo SearchMonkey, which I heard about but clearly need to dig into. It looks like these search add-ons go into a clunky gallery that users have to add to their results, which means they still need to work on helping developers get better distribution among the Yahoo search users. You’d think that one of the lessons from the Facebook platform would be that platforms need to build distribution channels into their product above and beyond the APIs that are provided. Thanks to Oren for illuminating me on the SearchMonkey point.

Written by Andrew Chen

July 9th, 2008 at 11:11 pm

Posted in Uncategorized

Poll results from “How would you launch a new product?”

without comments

I had previously posted a poll for the folks reading this blog to vote on a poll for how they’d launch a new product, and here were the results (as of this afternoon), ranked in order:

How do you launch a new product or service?

  1. Start with a site, then add app(s) later 56% (102 votes)
  2. Start with app(s), then add a site later 21% (38 votes)
  3. Destination site only 10% (19 votes)
  4. Social network app(s) only 8% (14 votes)

There were also a couple votes for “Other” which, from the comments and emails, seemed to be people who launched with both at the same time, or people looking at iPhone development, etc.

Frankly, I’m surprised by the results!

I really expected that there would be more folks to be emphasizing app and widget development rather than destination sites. Another way to look at this is that 66% of the votes had you starting with a destination site first, with apps being an afterthought.

Perhaps with all the Facebook and OpenSocial excitement that I’m exposed to in the Bay Area, it’s easy to perceive that *everyone* is working on apps when in fact there are many more traditional approaches out there.

Written by Andrew Chen

July 8th, 2008 at 6:06 pm

Posted in Uncategorized

Most ungoogley product ever from Google?

without comments

Google Launches Virtual World Called Lively.

I’m actually impressed by the fact that it’s so…. not Google. It seems like something a media company would be more likely to make, not a product made by a company of PhDs. Maybe it’s all those Yahoo refugees that are working there now?

UPDATE: Some more details from this informative article from Venturebeat. The product was created by one of the early founders of IMVU, and is much more directly positioned against that company than SecondLife. It’s built mostly as super-rich chat, has a focus on conversations rather than walking around a world, etc.

Written by Andrew Chen

July 8th, 2008 at 4:31 pm

Posted in Uncategorized