Snap Inc: How To Read the S-1

Whenever a hot company raises a round or IPOs, there’s a traditional Hacker News ritual where people dig through the S-1 looking for reasons to hate it. According to Hacker News, the typical well-funded tech company is a company that: 1. Has no real technology advantage 2. Is losing

Whenever a hot company raises a round or IPOs, there’s a traditional Hacker News ritual where people dig through the S-1 looking for reasons to hate it. According to Hacker News, the typical well-funded tech company is a company that:

  1. Has no real technology advantage
  2. Is losing buckets of money — who’d invest in that?
  3. 3a. Is a fad (if it’s consumer-facing)
  4. 3b. Is something nobody has ever heard of and potentially a fraud (if enterprise)

These are understandable reactions, and not because, statistically, everyone making these comments is one or two degrees of separation from someone less talented than them who is, thanks to that IPO, suddenly worth $20 million. And they’re directionally right: IPOs, in general, underperform equities in general. On average, you should not be buying into an IPO. But Hacker News cynicism tends to miss the mark. The actual situation is more complicated, and more interesting.

If you crowdsourced a hedge fund from Hacker News comments, you’d make 100% returns shorting a bunch of overhyped companies, and also post negative 5,000% returns from betting against Snap, Facebook, LinkedIn, and everybody else when they were hyped-but-underhyped.

What About Snap?

The first thing to understand about Snap is that it’s a historical contingency that they’re not worth more than Facebook right now. Facebook is an exception to the usual rule that messaging has higher engagement than pure social. And they’re an exception because they won social before this was true, and pivoted from there.

Facebook was originally a site for static identities, which evolved — painfully — into a site for photos and real-time status updates. And now it’s bootstrapping its user graph into messaging, while earning most of its revenue from the core profiles-plus-status-updates product. The Facebook of the future will monetize mostly through video, and keep users engaged through messaging.

Facebook is worth $385bn right now because they’ve been quick to pivot their product when they see a threat. Their first great pivot was the News Feed, when they switched from static profiles plus a list of recently-updated profiles to a list of recent profile updates. At first, this looked like a disaster: thanks mostly to the News Feed surfacing it, the Facebook group “Students Against the News Feed” quickly became the most popular group on the site. That’s harsh feedback. On the other hand, it’s a validation of the product, so they kept it. Now it’s hard to imagine Facebook as a directory rather than a feed.

In a few years, it will be hard to imagine Facebook as a feed, rather than a messaging app. In developing markets, messaging companies actually capture the competitive position Facebook occupies in the developed world. Tencent and Facebook-owned Whatsapp control the social graph in their home markets. The profiles-and-pics part is more of a sideline.

Messaging beats traditional social for two reasons.

  1. There’s a higher open rate. A message on Snapchat or Facebook Messenger is one-to-N, but a status update on Facebook or Twitter is one-to-approximately-N, where N is somewhere between two people (“I made a sandwich”) and five hundred (“We made a baby.”). Because you’re not choosing your audience size, it doesn’t make sense for you or the app to make your post interrupt someone else.
  2. Because of the high open rate, there’s a tighter viral loop. When your friend uses a messaging app, you get pinged, so you use it, and potentially ping someone else. Successful messaging apps tend to have a very high daily user to monthly user ratio, and get opened a dozen or more times a day. Opens per day are not a perfect metric, since new users are less engaged than older cohorts, so a growing app will have lower average usage than a mature one. But if you start to see older cohorts using the app less, that’s a sign that the viral effects are getting weaker, even if total active users and time spent keep rising. In fact, the first sign of a messaging app’s demise is probably that users have fewer, longer sessions; they clear out a queue of messages because the high-priority ones come from somewhere else.

This is why Facebook was willing to hire David Marcus from Paypal and give him carte blanche over the product. It’s also why Facebook ceded some of the world’s most valuable pixels — the upper left and upper right corners of the Facebook app — to messenger, a product that doesn’t really monetize.

In a fairer world, Team Snap, which figured out the importance of messaging, would be running the dominant identity-based app company in the developed world. But Facebook had a head start, which it secured through the bold strategic decision of launching when Evan Spiegel was in middle school.

The ultimate question for Snapchat, the only one that matters, is: can they beat Messenger and Instagram Stories in Facebook’s market, and then beat QQ, Line, and Whatsapp everywhere else? If you put together a basic toy model — here are their DAUs, here’s their average time per user, here’s the ad load (in terms of interruptions per hour) that people will tolerate, here’s the CPM, what you end up seeing is that everything is contingent on DAU growth. And when you think about how many teens and twentysomethings there are who aren’t already locked in to some other messaging platform, this DAU growth has to come from killing the competition.

The interruption-driven use case of messaging means that multiple messaging apps are not a stable state. There’s one you use for everything, and one you use for everything else. Over a long enough timeline, the only exceptions are niche platforms for small demographics (IRC for nerds, Yahoo Messenger for, allegedly, oil traders) or integrated products (Bloomberg Messenger, Slack — the default for intra-company messages, with a sideline serving communities whose members all use Slack at work, and Skype via its bundling with Windows 10). And as the sad state of SMS shows, even ubiquitous distribution can’t trump a sufficiently good user experience. (In fact, SMS’s status as a ubiquitous protocol owned by nobody in particular meant that every other messaging app could use it as a dumping ground for alerts.)

In theory, monetization is a problem Snap will need to solve. They’re bleeding money, they have negative gross margins, and no one is entirely sure if their ads work. On the other hand, Snap is in the business of providing real-time content, mostly video. And real-time video, according to PwC, gets roughly $73bn a year in the US alone. That number would be higher if people under 35 still watched live broadcast TV instead of spending their time on, to pick a random example, Snapchat.

Video is the king of advertising media because it can support such a wide spectrum of ad types. A low-quality video can be monetized with a simple text ad, which is roughly as easy to ignore as text ads anywhere else. But since there’s an infinite supply of crummy video content, there’s infinite inventory for these cheap ads. But video also supports high production-value, highly-interruptive ads. And the analytics are better; when users skip an ad, you know, down to the second, when you started boring them. Text can’t beat that without expensive eye-tracking studies.

I Could Build That In A Weekend

Comments sections are full of people who are pretty sure they could build Snapchat, Reddit, or Facebook in a weekend. (AlphaGo, maybe over Christmas break.) And that’s a fair criticism, but you will notice that the right weekend to build Snapchat was in 2011, and the right thing to build this weekend and relentlessly scale for the next half-decade is something else.

Basically no company outside of the life sciences sector relies primarily on its technological competitive advantage by the time it goes public. It’s still an important ingredient, but by the time a tech company is earning enough to IPO, its network effects and pricing power mean more than its ability to launch features. Product improvements still drive growth, especially in the more distant future — but the specific products the company will develop are hard to predict. When Facebook IPOed, we knew they planned on getting material revenue from mobile, but nobody was predicting messenger, live video, Instagram’s revenue contribution, or VR.

You can build an app in a weekend (it will suck, but you can build it). Building an open-this-app-every-day habit for 158m people takes longer.

Where To Start Reading The S-1

You know how there are lots of novelists who were frustrated lawyers before they quit to pursue their passion? Well, documents written by lawyers are, necessarily, documents written by people who have not stopped practicing law to indulge in their love for the written word. The ostensible goal of a prospectus is to sell a security — to explain to you, the investor, exactly why you should be excited to be a shareholder of Snap, Inc. The actual goal of a prospectus is to avoid lawsuits, and the way to avoid lawsuits is to pedantically explain every possible negative contingency.

For a lot of investments, the negative contingencies are the most interesting part! If you’re buying a bond issued by, say, a steel company, then the negatives matter a lot more than the positives. How old is their plant? How does their cost structure compare to steel companies in other parts of the world? How soon will they need to replace equipment, and how big a check will be involved?

Snap is not a risk-first company. Snap is a risk-last company. Only if things go very, very right will it matter that they could have gone wrong. So once you’ve read the summary, I highly recommend skipping the 32 pages(!) of risk factors, the disclosure about data sources cited in the prospectus, the use of proceeds section (Snap will be using their IPO funds to build Snap), and the financials (we’ll get back to them). Head to page 60, where they explain the business.

Snap’s value is based on a pretty simple equation: DAUs X ad load X CPM, minus costs, discounted back to the present. Annoyingly, GAAP accounting doesn’t perfectly match up with reality when it comes to expenses. A big chunk of Snap’s cost of revenue is better thought of as a fixed cost that won’t scale with revenue. While their gross margin last year was -112%, the incremental cost of another dollar of Snap revenue is not 112 cents. It might not even be 12 cents: sales commissions, customer service, hosting, and bandwidth are the only costs that will scale linearly with revenue (ignoring Spectacles for the moment).

The right order of operations is: Understand the business qualitatively (“Management’s Discussion and Analysis of Financial Condition and Results of Operations”) -> Approximately quantify it (DAUs, revenue per DAU) -> Extrapolate, think about what would have to go right for Snapchat to be a profitable business earning $4bn/year or more in revenue -> start guesstimating long-term costs based on their current financials, keeping in mind that operating expenses are mostly driven by what their competitors do, and are thus pretty fundamentally unknowable.

As a rough Fermi estimate, right now there are probably about a hundred people (about 95 of whom are within a ten-minute walk of the Metro North, BART, or London Underground) who have created a new Excel file called “SNAP Model v1,” plugged in SNAP’s financials, and started extrapolating forward to see when they could imagine the company turning a profit. They’re probably replacing net DAU adds with some estimate of gross adds and churn, then figuring out roughly what it costs Snap to get a new user, what that user’s lifetime value will be, and how plausible it is for Snap’s user retention to go way up. What they’re all seeing is that the only variable that really swings things is DAUs. Snap can’t get much better at selling ads without making some kind of monetization record, and their users are already pretty engaged given that almost all the content on the app is being created by other users.

$25 Billion?

You have to make a lot of assumptions to get to a $25bn valuation. In my toy model, for example, they have ~300m daily users by 2020, which makes Snap a phenomenon that either crosses global boundaries or somehow convinces a lot of middle-aged and older people to use something they think is synonymous with sexting.

There are two reasons to think Snap can achieve a $25 billion valuation.

  1. They got a $10 bilion valuation when they raised their Series D, despite having 74m users and essentially no monetization. Clearly Evan Spiegel can lay out a vision that prompts sophisticated investors to part with large sums under significant uncertainty.
  2. A discounted cash flow valuation treats investors as homogeneous. But they’re not! If Snapchat has a meaningful chance of being worth $25bn to a financial owner, it’s necessarily worth a lot more than that to Facebook, Line, Tencent, maybe even Google. For as long as Facebook’s revenue per DAU is higher than Snapchat’s, preventing the loss of that DAU is worth more to Facebook than winning that DAU is to Snapchat.

So, Snapchat is in a fascinating game theoretic conundrum. They can raise the money they need to be worth a lot as an independent company, but they’ll get attractive terms if and only if investors think they’ll nearly, but not quite, succeed. Every time Snap reports earnings, the stock is going to immediately lurch up or down based on DAUs, then gyrate a little more after hours as management comments on engagement trends.

If you’re investing in Snapchat, you’re really investing in a protracted M&A negotiation with a company that’s always working its way towards being able to walk away from the deal.