Andrew Chen Archives

I’ve had two recent conversations in which people have mentioned the “grow first, monetize later” philosophy as one of the signs of the coming bubble apocalypse, and this post is to argue why it’s very smart and rational to focus on getting millions of users first. (This post is part of my 2011 blogging roadmap)

Regarding monetization, I’ll note that…

• in general, consumer products mostly suck at monetizing
• any business model built on 1% subscriptions of 0.1% ads need millions of users
• costs are ridiculously low for new startups, and N millions of users is not expensive
• ignore this for products like marketplaces where monetization is part of the value prop

An ad-based example
Here’s some quick math- let’s say that you are a typical seed-stage team of 4 trying to get your startup off the ground, and your burn is approximately $40k/month. If you monetize at $0.25 CPM, which is a pretty typical ad rate for every thousand ad impressions, then that means you need a whopping 160M ad impressions per month to break even[1]. Even if you get 2X or 10X that ad rate, you’re still in the millions of users to get there. Scary right?

A subscription-based example
Similarly, if your site has a freemium business model, you’ll find that something like 1% of users subscribe for a pretty nicely tuned freemium configuration. So if you have 1% of registered users paying you $5/month, that means your average user is worth $0.05. Given this, you’d need 800k registered users, and if only 10% of your users register, you’ll need millions of users to get there.

Ultimately, the key is new user growth
Given the difficulty of monetization for consumer products, ultimately the best way to get to breakeven isn’t to try to optimize the 1% subscription rate to 2%, but rather to pick a huge market, create a killer product, and try to acquire millions of users. Because this is the biggest risk, you want to focus on growth first and foremost.

Here’s a different analogy that Steve Blank uses to get at this- let’s say that you wanted to create a cancer-curing drug. You don’t need to crunch the business model for that- if you had it, it’s valuable. You don’t need to price test or do customer development. All the risk is in the science, so you just focus on the science.

Similarly, I’d argue that in consumer internet, the real risk is that you can’t get millions of users actively engaged in your product, and that risk is ultimately driven by growth and long-term user retention. Thus focus on that first, then figure out the monetization once you’re at scale.

Stuff is so cheap these days
Note also that running a site with millions of users is cheap. The cost of hiring developers/designers will vastly overshadow the cost of maintaining the infrastructure- all you need are a few dedicated servers or just use Amazon Web Services- unlike the 90s, you don’t need a huge datacenter to get started. Because these costs are pretty low, you can just focus on making sure your designers and developers are productive and you’re getting to product/market fit.

Ignore this advice for products where revenue is part of the value prop
Of course for products where you are helping people make money as the central value, you need to do this sooner rather than later, so that you can make the entire network happen. So if you’re building a marketplace, collect money early, even if you don’t take very much profit. Same with Groupon-esque startups.

[1] CPM to revenue calculation
$0.25 CPM = $0.25 for every 1000 impressions
$0.25 / 1000 = $0.00025 per ad
$40k burn / $0.00025 per ad = 160M ad impressions per month

This post is part of my recent 2011 blogging roadmap post, where I created an outline of going from zero to product/market fit. Getting to this endpoint is obviously a good goal in theory, but question is, what does it even mean to hit this goal?

The original definition
In Marc Andreessen’s original post on the topic, he writes:

Product/market fit means being in a good market with a product that can satisfy that market.

You can always feel when product/market fit isn’t happening. The customers aren’t quite getting value out of the product, word of mouth isn’t spreading, usage isn’t growing that fast, press reviews are kind of “blah”, the sales cycle takes too long, and lots of deals never close.

And you can always feel product/market fit when it’s happening. The customers are buying the product just as fast as you can make it — or usage is growing just as fast as you can add more servers. Money from customers is piling up in your company checking account. You’re hiring sales and customer support staff as fast as you can. Reporters are calling because they’ve heard about your hot new thing and they want to talk to you about it. You start getting entrepreneur of the year awards from Harvard Business School. Investment bankers are staking out your house. You could eat free for a year at Buck’s.

His partner, Ben Horowitz, follows it up with a bunch of other observations about the fact the event isn’t a “big bang” kind of event – instead, there’s lots of gray area as your product starts working for the market: I’d encourage everyone to read his subsequent post here.

So the short answer is, there’s no easy test.

Now given that caveat, I’m going look at this through the lens of consumer internet to add some additional thoughts.

What is a market anyway? And how do you validate it’s real?
How do you even define a market for consumer internet? Ultimately, I concluded that the most useful definition of “market” is 100% consumer-centric. Here’s an attempt at a simple definition, focused on consumer internet:

A market consists of all the consumers who can search for and compare products for a use case they already have in mind.

This definition is very focused on the notion of pre-existing demand for products in your market, and is scoped narrowly to avoid confusion.

The most concrete test of pre-existing demand is using the Google Keyword Tool, which tells you how many people are searching on Google for a particular keyword. To try this out, you’d execute the following steps:

1. What keyword do people search to get to your site?
2. Put those keywords into Google Keyword Tool
3. How many people are searching for this keyword?

If the answer to #3 is large (millions or more), then you have a large market. This test is very concrete, and also very finicky. By design, terms like “vacation package” score high on this test, whereas “travel experiences” do not, even though an educated entrepreneur or investor might abstractly group them together. Similarly, by design, a person who’s building a “social network for musicians” might be inclined to list the # of musicians in the US as part of their market sizing, but under this test, you’d quickly see that there’s not too many people are specifically looking for that. Also interestingly enough, you’d never say there was a “Photoshop market” but a quick search will show that in fact almost 40 million searches per month on “photoshop,” and it might be a great strategy to position yourself relative to that keyword.

Validating that you are part of a pre-existing market comes with all sorts of benefits, which I’ll address in later posts. But for now, the most important benefit is that you know the # of potential customers is large.

(In general, I’ve been constantly confused about how to even define a market in consumer internet, given that there’s so much similar featureset between otherwise very different products. For example, early on, people talked about “social” as if it were a type of site, whereas now it’s seen as an aspect for all new products coming to the web. Similarly, people sometimes talk about “Facebook apps” as if it’s a market when, again, it’ll probably just end up an aspect of every new online service.)

What’s a great market?
What are other attributes that make a market attractive? For consumer internet, a great market is commonly defined by:

• a large number of potential users
• high growth in # of potential users
• ease of user acquisition

Not competition, in my opinion, because for consumer internet there is often literally billions of potential users, and you’re mostly competing against obscurity. So even if there’s a ton of competition, if it’s easy to acquire consumers to your product, that’s great! Then get a good enough product, and you’re ready to go.

Not monetization, in my opinion,  because making money is pretty straightforward. You can throw on some ads and get $0.1-$1 CPMs, or you can charge subscription rates and get 1% to convert, or you can do the virtual goods thing. The biggest risk in all of these monetization models is really about whether or not you can get millions of users or not.

Picking a great market leads to better products
Leading with a great market helps you execute your product design in a simpler and cleaner way. The reason is that once you’ve picked a big market, you can take the time to figure out some user-centric attributes upon which to compete. This leads to a strong intention for your product design, which drives a clean and cohesive UX. In a market of all black Model Ts, you can sell otherwise identical cars of different color and that’ll work. Picking the right attribute is it’s own topic though!

The important part here is that you can usually pick some key things in which your product is different, but then default the rest of the product decisions. This means that your product’s design can be more cohesive because you’re trying to do less, but better.

Once you’ve executed your product, then there are various ways to validate that it’s “good enough” and your product fits the market:

• When user testing, do people group your product in with the “right” competitive products?
• Do they understand the differentiation of your product versus your competitors?
• Will some segment of users in the overall market switch to your product?
• Are some users who’ve “rejected” the products in the market willing to try your product?
• How do your underlying metrics (DAU/MAU, +1 week retention, etc.) compare to your competitors?

All of the above are signals towards product/market fit. Thee above tests are interesting in that they fundamentally anchored on pre-existing competitive products in the category. In a new market, you don’t have the luxury of comparing yourself to other things.

In future posts, I’ll try to give some more concrete metrics based on my research for what are good numbers in each of these cases, but for now, the important idea is just that in a large existing market you have more datapoints to at least say, “my product is at least as good as the other guy’s.”

New markets are a danger to good product design
In fact, one of the scariest things to me about new markets is that doing great product design for them is extremely hard. It’s so unconstrained that it’s hard to do anything other than add features, see what sticks, and iterate. This is fun except that keeping a cohesive product experience is quite hard, and removing features is usually harder than adding them. So at the end, you incur tons of product design debt that never gets paid off. (It’s not a surprise to me that Apple has a history of simplifying already successful product categories, rather than inventing brand new ones from scratch)

Conclusion
To summarize my main points in this essay, I’ve come to some simplifying definitions on how to validate product/market fit in consumer internet. For market, if you constrain the definition to people who know how to search for products in your category, you can develop a pretty concrete test evaluating pre-existing demand. And by leading with a market, you can develop a central design intention that leads to better product design. This in turn can then be validated by comparing your product metrics to competitor numbers, as well as user tests that focus on grouping and differentiation.

This leaves lots of unanswered questions, but hopefully is a start to my new blogging roadmap! More to come soon.

One important question that comes up all the time is, what makes a product easy to market?  I had a fun chat about the topic with Eric Florenzano and Eric Maguire who worked together at Mochi Media with my sister Ada. They also recently did YCombinator.

After the chat, eflo wrote up a helpful summary of some of the ideas we covered. I wanted to quickly share them with some comments:

1. Come up with one resounding use case–one thesis for how people should use the product.  Preferably this fits in with something that users already do and already understand.

I’m going to try to start blogging again!
It’s been a long time since I was in a good blogging rhythm, and I’m going to try to start doing it again :-) In preparation for this, I put together an outline of an output-driven set of milestones around product, that takes you from zero to a P/M fit product thats ready to scale on marketing/tech/etc.

As far as I can tell, this is all standard fare for companies in Silicon Valley. My desire to write these posts is ultimately about documenting what’s working for people and spreading the knowledge beyond Palo Alto, CA :-) All of these topics are ultimately derived by both my own projects as well as my advisory roles at venture-backed startups. (Some of these are listed here)

If you like the outline and want to stay up to date, just subscribe and follow me on Twitter.

Without further ado, here’s the outline- I hope to write at least a post or two per week:

Blogging roadmap goals

• “output-driven” roadmap for going from zero to product/market fit
• for small hackerish teams building consumer internet products
• the intention is to create a scalable startup that is going after a huge market, and generate huge returns for venture capital investors
• goal is to get to P/M fit in shortest time possible, defer everything else
  • defers monetization
  • defers marketing
  • defers scaling
  • (this is all by design)
• P/M fit takes a non-deterministic amount of time to get there, insanely hard, you’ll probably fail anyway
• the problem is 90% contextual, make up your own rules as you go

Concept prototype

Picking a product and market

• build for yourself (start with intuition)
• have a long-term vision
• base it off something that’s already big and already working
  • big makes it easy to test and collect feedback
  • already working means you have a good sense for minimum product
  • also, there’s pre-existing distribution channels as well
• figure out the options for competitive differentiation – this is the core design intention
  • talk to a lot of users, do a lot of research, compare a lot of products in the space
• dimensions for competitive differentiation
  • competitive dimensions
  • vertical audience
  • design intention
  • cheaper/niche
  • targeting rejectors
• validating that there’s LOTS of pre-existing “pull” for the market
  • search keywords
  • app leaderboards
• ideal goal: simple product with fundamentally different core design intention for large pre-existing market
  • bonus points for baked-in distribution, monetization, etc. but don’t let this lead the idea!!!
  • usually one killer feature (not a bunch of features)
• prototype: Landing page
  • what’s a good landing page experiment?
  • headlines, copywriting, hero shot, etc.
  • unique URLs
• anti-patterns:
  • “someone’s already done this” (desire for originality)
  • monetization/strategy-driven product ideas
  • technology in search of a market
  • “Wall Street” markets
  • lumping yourself into an aspirational market
  • comprehensive featureset done poorly

Paper/Wireframe prototype

Designing the initial product

• go for the minimum desirable product
  • might work :-)
  • the central design intention drives the product design
  • supports only the core use case, as minimum as possible
  • core UX should be 2-3 pages
  • limited functionality, done well. “Less but better”
  • Should build bare bone prototype in less than 2 weeks (really!)
  • flow-based product design
  • user quotes, then fill in with UI
• low-fidelity prototyping tools
  • easier and cheaper to make changes
  • fix defects earlier (Toyota lean manufacturing model)
  • engineers always want to prototype in code, but then sunk-cost fallacy
  • get feedback from people and iterate
• prototype: Core user flows, mocked up and ready to build
• anti-patterns:
  • “database-up” design
  • feature creep and low product self-esteem (v1 should look like a feature!)
  • comprehensive featureset all of it done poorly
  • lots of pet features that don’t fit into the core design intention

Code prototype

Coding the initial product

• build the prototype as fast as possible
• fill in any blanks left out of the prototype
• use the product yourself, iterate on it while keeping with the core design intention
• focus on key flows and prioritize over ancillary ones
• don’t worry about corner cases
• get it ready to be used by other people
• prototype: Live product, usable by other people
• anti-patterns:
  • taking too long
  • losing focus of the central design intention
  • not adjusting based on intuition and usage
  • overarchitecting, trying to make it scalable or modular or future-proofing in general

Friends and family alpha testing

• private beta goals
  • clean up core experience
  • make product usable over multiple visits
  • validate the core design intention
  • not scalable
• recruiting friends and family
  • focus on retention
  • are users coming back?
• recruiting random people
  • Find people from the existing market, rejectors, and outside the market
  • Learn from extreme users
  • Craigslist
  • Usertesting
• user testing
  • do they get it?
  • how would you describe this to a friend?
  • usability – remove the friction
  • would they switch? (for existing market users)
  • Net promotor score
• interpreting user feedback and learning to say “no”
  • which users fall into the target market? Hear them out
  • which users don’t? It’s OK (and maybe even good!) to have them reject
  • try not to add new features unless absolutely necessary
  • what features can you remove that aren’t part of the core?
• prototype: Simple product, polished by real use
• anti-patterns
  • Delusion- it’s not working but you think it is
  • Melancholy from user testing
  • Adding features without interpreting
  • Adding features that violate core design intention
  • Listening to out-of-market users
• is it working?
  • people understand the product
  • some subset of your users like it and use it
  • you like it :-)

Random people beta testing

• traffic testing goals
  • start polishing your onboarding flow
  • develop options for distribution
  • build some basic stats infrastructure
  • not meant to be scalable
• User acquisition tactics
  • ads
  • PR + launch page + slow stream
  • partnerships
  • power through it
• Collecting feedback
  • surveys
  • help and problems
  • recruit users to talk to
• prototype: Spreadsheet for signup flow, more polished signup flow
• is it working?
  • signups are happening
  • people are going through the core flow
  • retention/recurring usage from target users
  • product still works for you, and your friends/family

User flow optimization

• model your usage and figure out your core drivers
  • this is completely product specific
  • two examples- daily deal versus a chat site
  • whats your “metric of love?”
• prototype your funnel – explore!
  • flow chart
  • excel
  • SQL
  • formalize/finalize with dashboards
• identify major bottlenecks for why the product’s not working
  • start at the beginning of the flow
  • fix bottlenecks with A/B tests
• is it working?
  • how do the metrics compare to the usage model?
  • 10% signup
  • +1 day retention and +1 week retention
  • DAU/MAU
• anti-patterns:
  • trying to fix problems in core UX when signup is the problem
  • over-architecting stats infrastructure
  • trying to use a generic analytics product to answer situational questions

Ready to scale?

• Hopefully the major checkboxes are checked – at this point you’d have:
  • Huge market
  • Differentiated product
  • Product makes sense to normal people
  • Product is working for IRL people
  • Product is working for non-IRL people
  • Well-understood and optimized user flows
  • Ready to scale up
• Non-scaleable marketing, tech, and otherwise- that’s fine
• Now scale everything else :-)

Crisis, terror, and melancholy

• Is it good enough?
• Nobody likes my product!
• My product is a mess!
• It’s taking too long!
• Investors hate my product!
• I’m iterating in circles!
• When to work on a completely new idea?
• Iterations are getting diminishing returns and people still don’t love the product

Final note: Thanks to my friends who helped review and add to this: Vinnie at Yipit, Alex at Penzu, Rob Fitzpatrick, Kevin at Hyperink, Jamie/Justin at Mocospace, Ada/Sachin at Connected, Noah at Appsumo, Jason at Kima, and the other folks who helped