Andrew Chen Archives

A couple friends were in town recently, and I went with them to the Mechanical Museum at Fisherman’s Wharf, where they have lots of different old mechanical arcade machines. The oldest one was from the 1920s, and the average period looked to be 1950s or so.

It reminded me that while technology advances, human nature stays the same.

Love calculators
First up, we have a bunch of machines that are sort of like the "How good of a lover are you?" quizzes in  trashy women’s magazines. In general, you stick in a quarter, put your hand on the pad (or some other interactive action), and then it gives you a score.

Here are pictures of the machines I took on my iPhone – click to see a bigger version:

The main emotions that are being elicited in these cases are some combination of:

• narcissism
• curiosity
• competition

Extra points to the first (left-most) machine for the tagline:

"What do your friends call you behind your back?"

Is this really so different than the various quiz, comparison, and other applications on Facebook? And take a look at a site like this: Best Love Calculator. It even evokes the look and feel of these machines.

Telling the future
Also, we have machines like the ones below, which are focused on telling the future. The first machine is the most fun – you put your hand into the machines mouth! The second and third ones are both palm reading.

Here are the pictures:

Of course, horoscopes are still big these days, and people still inexplicably talk about their "signs" – I would ask "who knows why" but the answer to that is simply "because people are people." I’d boil the emotions down to:

• narcissism
• curiosity
• insecurity?

In the modern world, the entire clairvoyance thing is still around, via numerous websites that you find when you google for "psychic". Here’s a good example of what you get when you click on one of those ads. And don’t even get me started about John Edward’s Crossing Over and psychic shows like that.

Jackass and YouTube, oldschool style
These next machines are probably the most dated (and hilarious) – basically when you put in a quarter, you’re then able to watch some "kinky" (defined by the 50s). The first machine was the "French Execution," which played some music and you got to watch a guillotine chop off a miniature doll’s head. The second machine had you grabbing a hand-crank to operate a flip book with pretty boring material in it.

Here are the images:

This is obviously Jackass and the YouTube of the 1950s.

Emotionally, this is catering towards:

• novelty
• scarcity
• curiosity

Of course, the problem with these machines (unlike the other ones) is that after you’ve seen it, it doesn’t seem so special anymore, and you’re unlikely to watch it again. And of course, in a modern society where this type of stuff is available at a much more, ahem, liberal standard, it’s really boring. These were fun mainly because they show how dated the place is.

Simulation games
While Will Wright is often heralded for creating the Sim games for the PC, you can look further back than that to see simulation games. In these machines, we see one which is a helicopter machine and then a crane, both of which are operated using a set of simplistic controls.

Here are the images:

Interestingly enough, both simulation games were "directed" rather than undirected. They weren’t pure sandbox games (like SecondLife) but rather had a goal structure for the user. This is something that Erik Bethke (of GoPets) and I have talked about in the past.

Rather than just letting the user fly the helicopter, instead there were lights around the area where you were supposed to hover the chopper. The longer you hover, the more points you score. The lights rotate around the area, and you have to move the chopper there accordingly. The construction crane does the same thing, where you are supposed to grab as much dirt as you can in as little time as possible.

The emotions here are quite different than the other ones:

• aspiration
• fantasy
• competition

And of course, it’s obvious that the modern versions of this range from things like The Sims to MMOs to any other game that is about role-playing.

Differences with regular websites and Facebook apps
Interestingly enough, the machines above are actually quite different than what you would want to build for a modern consumer product.

The reason is that these machines are incented to:

• Have a great hook to draw a user in
• Make them give you a quarter
• Provide some value, but focused less on retention and more on pumping in more quarters

This is misaligned from websites, which share the attribute of drawing users in, but are focused around retention and constant usage, because that’s what drives advertising revenue.

Similarly, Facebook apps have a different incentive structure:

• Have a great hook to draw a user in
• ALSO, have a great hook to get the user to pass it along to their friends
• Provide some value, but mostly focused around virality
• Make the structure around frequent usage with continuing value

That’s why things like Magic 8 Balls and Fortune cookies and such are gimmicky products that might drive acquisition, but have problems with overall retention and active usage.

Conclusion
I often find that studying older historical products like this to be really fascinating. I think you can learn a lot about human psychology by looking at things like:

• card games
• physical architecture
• con artists
• old advertisements
• public speaking
• magic and psychics
• etc.

While many things are not directly applicable to the world, many of these have underlying themes and emotions that might be useful for modern entrepreneurs.

I just tinkered around with my blog a little bit to expose some of my "back catalog" of content that I’ve written over the last couple months. For those of you who have my blog on your feed reader, click here to check it out.

The most substantial change was to expose a bunch of previous essays that might be new to some people. It’s on the left sidebar, underneath my new picture ;-)

Here’s the list of previous blog entries:

Wow, what an interesting announcement on the OpenSocial standard that’s coming out. As always, Marc Andreessen has some insightful commentary here.

I wanted to give some context on this entire thing from the app developers I’ve talked to:

RockYou and Slide in the lead at Facebook, developers looking for other channels
Overall, the app developers I’ve talked to feel like Facebook is somewhat of a lost cause, if you are writing a new app. Now that Slide and RockYou are fully entrenched as the top apps (just look at the leaderboard), they’ve been able to sustain their lead by cross-selling applications. For those who haven’t been following that, basically it means that with millions of installs, RockYou and Slide can easily cross-promote a new app through an existing one. This is done either at a feature-level, providing links to the new apps, or when you are installing the popular app, it takes you to an interstitial that asks if you want to install the new one.

Being able to launch quickly is important because your viral growth is dictated by:

new installs = existing installs * viral coefficient

They can build a bigger “seed” of existing installs, and thus, they can get more new installs. With the same product and the same viral-ness (determined by the coefficient), they will still pass you, no matter what.

Interestingly enough, of course, Slide and RockYou are listed as launch partners on OpenSocial, and it may be likely that they are gearing up to do it again.

Looking at other platforms
As a result of this, app developers have been looking for other social networking platforms to build on, some of which are included in the group announced, which include:

• Orkut
• Salesforce
• LinkedIn
• Ning
• Hi5 [Note: my employer MDV has an investment here]
• Plaxo
• Friendster
• Viadeo
• Oracle

This set of publishers is really fantastic, and encompass a huge number of users – from a unduplicated, registered users standpoint, you are talking about hundreds of millions of users overall.

That said, this group is also very different from each other – you have different demographics of users with different sets of goals, with different usage patterns located in different areas. In particular, it strikes me that LinkedIn, Oracle, Plaxo, and SalesForce are in one big lump of US-centric business audiences versus Orkut, Ning, Hi5, and Friendster, which is more consumer-oriented and substantially international.

I think this means that even if OpenSocial is able to allow widgets/gadgets/apps to function technically, it doesn’t guarantee that you can build in one place and not “localize” it to the target social network, and that you can get strong traction without understanding which platform you are getting yourself into.

Let’s look at how that might affect viral growth:

Maximizing growth and virality
One of the fun exercises that you can do to understand viral marketing is to install all the top Facebook apps and see how they drive a viral loop. In particular, which ones try to get you to invite right away, and how many try to get you to fill something out before you invite, etc.

A particular lesson you can glean from this is that virality is deeply ingrained inside consumer psychology. One example is Boozemail, which is run by my friend Adam Rifkin and his company Renkoo. A couple observations on Boozemail:

• The fundamental emotion it’s hooking is “reciprocity” (like gift giving)
• Thus, when someone buys you booze, you feel compelled to “buy the next round”
• Unlike gifts, however, booze can be bought for many people at once
• Booze is also something you buy regularly (not one-time, or for special events, like gifts)
• Also, booze is something that is uniquely college-oriented and 21+
• As a result, Boozemail is uniquely viral because when you receive a drink you want to send one back, but when you get to that screen, you can feel comfortable buying A LOT of people booze

Looking at this, you can learn that real world analogies matter a lot! And also that all the subtle hooks in the app – how it’s positioned, what emotion it hooks onto, how comfortable you feel sending it on, etc., all play into the viral loop.

From a user perspective, of course, Boozemail might work quite well in the US college-centric audiences, but bring it to Salesforce, and it might hold a different connotation.

Write-once-run-anywhere? Or Write-once-succeed-anywhere?
So returning to OpenSocial, if I were a developer on this platform, the problem that it’s solving is to ensure that I wouldn’t have to do a bunch of annoying things like use new markup languages or new API structures for every network, but ultimately, this is a “write-once-run-anywhere” proposition but not a “write-once-succeed-anywhere” proposition.

I think that for anyone who is developing applications, one very smart thing to do would be to understand the respective userbases of the companies participating, and ultimately choose one or two to focus on. In the cases where you are talking about audiences in the Philippines (Friendster) or Brazil (Orkut), it might be worthwhile to spend a lot of time trying to understand what internet paradigms work over there and why, and then focusing a series of apps to own that audience.

Either way, it’s a very interesting step, and it’ll be fun to see how things will develop.

Max is one of the best tech entrepreneurs in the Bay Area right now. Definitely worth reading this profile from the NYT: After First Succeeding, Young Tycoons Try, Try Again.

I guess this also goes hand-in-hand with the recent pictures from Valleywag on the PayPal mafia.

Here’s the list (I’m such a hater):

• Podcasting
• Blog search engines
• Semantic web
• Vertical search
• Google Maps mashups
• Social shopping

… and now, 2-D e-mail, which has been replaced by 3-D email:

Winner-take-all categories: YouTube as King of the Hill
I was recently chatting with a friend and the topic of YouTube-like exits came up. In that version, it’s winner-take-all and the #1 leader gets bought for $1.6B, and everyone else is worth much less. Similarly, Skype might be an example of this since they were bought for so much, yet other folks in the category didn’t cash out as well.

Compare that to something like CRM, where back in the day, you had dozens of vendors hawking their own versions of products, and you had many different winners worth lots of money. In addition to Siebel, SAP, Oracle, Salesforce, Microsoft, etc., you also have tons of consulting firms who generate lots of money as well.

Ad networks in 2003
Now in 2003, ad networks were around, and a couple had already exited for a fair amount of cash, yet at the same time they were considered pretty undifferentiated and low-margin. Between lead generation companies, comparison shopping engines, affiliate networks, and ad networks, the entire online advertising space was just a morass of noise without a lot of technology.

It was also not clear that a player like Google wouldn’t end up with a winner-take-all outcome. The thought was, if they were able to aggregate more advertisers (200k+) than everyone else, then they should be able to pay more dollars for inventory, which would lead to them picking up more advertisers, and so on. It’s a virtuous cycle of lock-in that has driven a lot of revenue.

The big surprise (for me)
So of course, in 2003, it would have been considered fairly dumb to start a new ad network, right? You might say the same thing for social networks right now :)

But truth is, a couple interesting things happened in the remnant advertising world:

• There was a huge jump in remnant inventory led by UGC sites
• Leadgen, SEMs, and other big advertisers were running out of high-quality search inventory
• CPC didn’t turn into a silver bullet to address inefficient ad buying

So interesting enough, all three of these factors came into play, forcing advertisers to continue exploring for good inventory out in the long tail of the internet.

The first one is obvious – sites like MySpace, Facebook, YouTube, etc are all still very hard to monetize. The second one is less obvious: Because search was such a great thing for ROI-driven advertisers, they would buy it until they were bidding so much that all their margins disappeared. As a result, they’d have to start buying cheaper (and more experimental) types of inventory to find positive ROI businesses. Thus, they’d be more likely to experiment with new publishers and new types of advertising, just to see what happens. And finally, CPC is great, but there’s a lot of uncertainty in the equation. Publishers want guaranteed revenue (via a CPM), advertisers want guaranteed revenue (via a CPA), and CPC is just a mid-point solution.

Of course, all this inefficiency made it so that many different ad networks could co-exist with each other, by constantly grabbing onto new publisher ad inventory, then losing it, then grabbing it later on. The instability of all of these relationships, coupled with the willingness to experiment by all parties, made it easy for a lot of new ad networks to break into the market and get to a critical mass.

One statement that started a $10B+ buying spree
And finally, it only took one signal by Google to set off the big buying spree: Display advertising is important.

After that, Blue Lithium, Right Media, Doubleclick, 24/7, aQuantive, Tacoda, etc. were all several hundred million to several billion dollars acquisitions based on the premise that all the vendors with relationships and who have gotten traction in the market are worth a ton of money. Even though the ad network business looked like a low-margin, commodity business, along with ad serving and other types of infrastructure, it was irrelevant because it grew out of a winner-take-all business where revenue was speculative. Pretty interesting how that worked itself out…

My overall lesson from all of this is that a lot of times, people view things as "winner take all" and sometimes it is that way – but in this case:

mature industry + real revenue + adjacent space heating up
= huge outcomes for everyone

Worth looking at what else is out there with that kind of model brewing.

Good summary in the NYT today about how hard it is to get good numbers in online advertising: How Many Site Hits? Depends Who’s Counting – New York Times.

How many people visited Style.com, the online home of Vogue and W magazines, last month? Was it 421,000, or, more optimistically, 497,000? Or was the real number more than three times higher, perhaps 1.8 million?

The main discussion ends up being around comparing the panel-based numbers (Nielsen, ComScore, and others) versus the server logs that people have.

In truth, the panel-based folks should probably offer a pixel to publishers to get even deeper counts – panels can be very off, which works in TV because there’s no other options, but not in online advertising.

On another note, I’m getting my wisdom teeth out this afternoon so I won’t be blogging for the next week or so. Wish me luck!

InsideFacebook has a great article on Facebook’s recent effort to standardize application invites:

Since the beginning of the Platform, Facebook has had to tweak the
invitation rules a few times to limit the impact spammers could have at
large. At first, Facebook had no limits on invitations – apps could
spam all your friends every day, and many apps had little behind their
full page of checkboxes with no skip button, yet were able to grow to
millions of users in a matter of a couple of weeks. After that,
Facebook limited the number of invitations a user of an application
could send to 10 per day – and later to 20 per day, though only 10 at
once.

Nevertheless, that didn’t stop developers from still using
borderline tactics. Because invitations are such a vital part of the
user application experience, Facebook ultimately decided to bring more
of the UI in house to prevent poor experiences from becoming common
amongst Facebook applications.

This standardization brings to an end one of the most interesting things I was seeing on Facebook, which was "invite targeting." That’d mean targeting invites based on attributes like:

• Demographics
• Other apps installed
• Interests
• Friend relationship strength
• etc

If you think of every invite as an ad for the application, and you only have 10 per day to send out, you’re going to want to target the invites to the people mostly likely to click through and install it. Thus, if you’re making a movies app, you want to make sure you are sending to people who don’t already have the application, list a lot of movies in their interests, etc. Then you can start to A/B test which attributes are the most important, track stats and conversions, and other sophisticated things.

Net/net, it may actually be a good thing for Facebook to track conversions for invites, and make reward applications that are good at inviting the right people – this targeting process is likely a useful thing that enhances the experience, not detracts from it.

The leader in online advertising tries to monetize social network inventory
Great article from the WSJ on the issues Google has faced monetizing user-generated content. Here’s a couple excerpts:

Then, when Google tried putting ads on the site, it ran into trouble.
Critics in Brazil released a report showing advertisements on Orkut
alongside pictures of naked children and abused animals. Google
immediately suspended the ads, but the Mountain View, Calif., company
is still grappling with the fallout from critics’ Orkut campaign.

And of course, for brand advertisers this is even more embarrassing because just one ad can cause a lot damage, versus direct response guys who just lose out on reduced ROI:

Liquor maker Diageo PLC of London says it stopped advertising on all of
Google’s properties after learning that its ads ran alongside
pornographic images on the site. Spokesman Stuart Kirby says Diageo
didn’t realize that ads for its Johnnie Walker brand had appeared on
Orkut, where many users are below legal drinking age.

… and more:

On Aug. 17, Mr. Tavares sent an 18-page complaint to Brazil’s
advertising watchdog, known as CONAR, documenting cases of embarrassing
juxtapositions: advertisements for Diageo’s Johnnie Walker whiskey next
to pornographic images; a pet store pitch on a community dedicated to
stabbing animals with knives. In the report, Mr. Tavares alleged that
Google’s "flagrant illegalities" had resulted in ads appearing next to
"barbaric" content.

Key obstacles for UGC + brand advertising
The intersection of issues that cause this come down to three key points:

1. UGC content means some percentage leakage of inappropriate content
2. Blind ad networks mean that advertisers don’t know where they are buying
3. Brand damage can happen via one single false negative – it’s not about % of leakage

This, in addition to users not being in a buying mindset is what causes social networks to be so hard to monetize via brand advertising. The only way to solve this is to use a direct ad sales team that is selling a very transparent ad unit on a page with no UGC content (editorial pages, section homepages, etc.). Of course, this means that all the profile views end up being managed via remnant ad networks, which is no good either.

Of course, this also means that anyone who can figure out how to make this inventory safe is sitting on a billion dollar+ business.

It’s been a couple hours since the CommunityNext conference got started – and things have been going smoothly so far.

The first day is focused on developers – you can see the schedule here. Akash, CTO of Hi5 kicked off with a great discussion of scaling up to 70 million registered users, and how to architect a system that scales up. Since then, there’s been a bunch of demos from successful app/widget creators.

The room is completely full, with a standing room only audience with overflow out the door:

Here’s the scene right outside of the main presentation room, with lots of people with laptops out, PuTTY terminals and IDEs open.

… More updates soon.

I got a couple quotes in articles recently on the Facebook apps frenzy:

• Wired: Selling ads on Facebook is a risky but tempting business
• Fortune: New ad outfit targets social-networking sites
• New York Times: Swarm of software developers creating features for Facebook
• Business 2.0: The Facebook economy

I end up being one of the skeptical voices in the articles, but my intention is matter-of-factly describe the hurdles the sites have to overcome on the advertising side in order to succeed. I think it’s possible, yet will take some time. These are lessons I learned from Revenue Science, trying to monetize traffic on several of the top social networks in the world including MySpace.

The specific issues that pop up from the monetization side shouldn’t detract from the fact that adding deeper hooks is something that every network should do, and Facebook has done it with incredible execution. They are a visionary in the category.

However, that still doesn’t mean it’ll be easy for the entrepreneurs to make money :-)

Other blogs on this topic:

• How much is a Facebook user worth, anyway?
• Inside Facebook interview
• 5 things that make your social network monetize like crap
• Inventory glut in social media