Silent Disco Economics
I’ve been trying to get back to writing essays for a while now; the last piece I wrote for this blog was in January 2023. Frustratingly, my New Year’s Resolution was to publish 30,000 words. I don’t think I’m going to manage that, sadly. While I knew I’d be running our company, Waterfull, that involved much less content marketing than I naively hoped. I also (wrongly) worried that any effort I put into reading and writing for pleasure, and especially writing in public, was effort I really ought to be putting into making the company successful.
This piece is crystallised written version of the stories I found myself telling over and over again in the course of the year.
The mobile ad ecosystem had a good run, but things look pretty miserable in 2023; the industry expanded quickly, and is now dealing with a capacity problem. There are far too many studios, far too many game devs, and not enough revenue to sustain them all. The Swedish roll-ups are defaulting on debt, and the layoffs have been brutal.
I want to share three useful mental models that I talk about when I explain the mobile ads and mobile gaming business to people; they may not be 100% accurate, so let me know if you think I’m getting them wrong. Nonetheless, I think they’re a useful way to understand
1. how a free-to-play mobile studio, like my old employer Tripledot, functions;
2. how the mobile ad ecosystem worked in 2022-23); and
3. the key drivers of growth in mobile ad spend over the last 10 years.
Free-to-Play Economics
Anybody that really cares about the economics of mobile games should go read my essay on the subject here; you should also listen to Mitch Lasky and Blake Robbins’ podcast series, Gamecraft.
But it’s helpful to think about a mobile games company as having 3 components, which create a flywheel.
The product team is the largest team in the company - they build the game, and make it better. When they’ve built a good game, the user acquisition (UA) team can go and buy users for the game by spending money on various marketing channels. Once users are in the game, you can show them ads - maybe at the end of each level. At that point, it’s the job of the ad monetisation team (AdMon, or Monet if you’re French) to sell the right to show an ad to the players for as much money as possible. This team is small, and probably stressed.
Product teams optimise for retention - a cohorted metric, which refers to the number of users who installed the game on a particular calendar day, that logged in X days after the install. UA teams optimise for ROAS (return on ad spend) - another cohorted metric, the cumulative revenue which the users who installed the game on a given date generated by day X, divided by the amount it cost to acquire them. AdMon teams optimise for CPM - cost per mille, or thousand impressions.
Overall, the marketing team is oriented around a “payback period” - the time it takes for their cohorts to reach 100% ROAS. Everything after that is profit - although there are, of course, other operating costs. The payback period might not be the same for every game, but at a high level it’s set by the CEO and is a function of cost of capital, growth demands, and so on. At an extreme level, you can run a sustainable games company with a 365-day payback period, which might mean that on any given day, you’re spending 70% of revenue on marketing.
If all three of those teams do a good job, then you can pump obscene amounts of cash through the UA team, and expect to see it back the other side on the AdMon team. At Tripledot (one of the largest independent companies in the world), we often saw days with more than a million dollars in ad revenue.
The Auction House
Where does all this money come from?
Well, imagine you’ve got a valuable painting, and you want to sell it. You probably shouldn’t walk down the street hailing down passers-by - you should go to an auction house, like Sotheby’s. And when the auctioneer at Sotheby’s sells your painting, they don’t have random people in the room either - instead, the room holds a handful of art dealers, each of whom represent many clients - and possibly even the same clients. When a painting comes up for sale, each art dealer thinks about their clients’ tastes and chequebooks, and decides what, if anything, each of their clients would want to pay for that painting; this informs how they’ll bid in the auction to come, although they’ll keep a healthy margin for themselves.
That’s more or less how mobile ad auctions work. You’re the publisher, the owner of the painting, and when you decide to sell ad impressions in your game, you go to one of three auction houses - Applovin MAX, Google AdMob, or Unity LevelPlay, known as “mediation platforms”. The art dealers are “ad networks”, such as Google, Facebook, or Applovin - there are 10-15 of them. And they buy ads on behalf of advertisers - taking a 30% commission along the way.
Now, as the owner of the painting, you get to give Sotheby’s some instructions - a great one might be “don’t sell for less than a million dollars”. That’s called a “bid floor”. You might also say “don’t sell to that guy”. But in some mobile ad auctions, called waterfall ad auctions, it’s more complicated than that. These auctions work down from high to low, in discrete steps, and at each step only one participant is called to bid. For example, you might give Google the chance to buy at a $300 CPM - if they say no, move on to Unity at $280, and if they say no, then back to Google at $275. This process continues down to $0.01, often with around 100 price/network combinations, until someone buys at the price they’ve been offered, or the whole process runs out of time. As a mobile app publisher, it’s your responsibility to instruct the auction house/mediation platform on the running of this whole structure - you specify the price/network combinations. This is an important job - if you only offered one person the chance to buy at one price, then you’d be missing out on revenue, and the same is true for any given combination - just think about how you could tweak the highest price in the waterfall.
To get the best out of waterfalls, you need to optimise them with software and machine learning - you’ve got a different price structure for each app, user segment, and country - and that’s exactly what our first company, Waterfull, did.
If waterfalls sound complicated, then there’s another option - bidding - where everyone writes a bid down on a piece of paper, and the highest bid wins. But it’s tricky for ad networks to value impressions this way (I have this on good authority, but nobody’s ever explained the engineering to me), the mediation platforms take a 5% cut, and in general publishers tend to see lower revenue.
I’ll probably write something soon about why in the end Waterfull wasn’t a great business - but we had a great time building it, we learned a lot, and I’ll always be grateful for the kindness of the people I met in the mobile ad industry.
The Silent Disco Effect
What gave birth to the mobile ad industry in the first place? I’ll set the scene with a piece of context, and then talk about a few key trends that drove growth over the last few years - and ultimately stymied it.
Investors like a tailwind, and a wonderful way to think about tailwinds is what I’m going to christen the Silent Disco Effect. I don’t think Americans know this, but in the UK it’s common at Oxbridge balls and posh 18th birthday parties to have a “silent disco” at the end of the night - everyone wears headphones that can be set to either the blue, red, or green channel; each channel has it’s own “radio station”, playing the same music, and when you switch between them the LEDs on your headphones change colour.
Now, we gon' break this thang down in just a few seconds.
If Hey Ya comes on the red channel, and the blue and green are a bit meh, then you’ll see the whole crowd start to shift from blue and green over to red. People will excited tap their friends on the shoulder, pointing at their red headphones, and expect their friend to shift over - and if they don’t, well, y’all don’t hear me, you just want to dance.
Anyway, the whole crowd shifts to red, until another certified banger like Holiday by Dizzee Rascal comes on the green channel, and the crowd starts to shift back the other way. Maybe some people don’t like Dizzee, and they want to listen to Katy Perry on the blue channel, so you get a bit of fragmentation.
As an investor though, you’re looking for those waves where everyone starts to switch from the blue channel to the red channel. Two great examples here are cloud and cards. You can make an argument that the basis for 1. cloud and 2. fintech is the gradual shift from on-prem software to cloud-based services and from cash to cards respectively. Before the Bakersfield Amex drop, nobody was listening to Hey Ya - and these days, if you want to buy something, you know what to dooOOOoooOOOooo.
If you’re still reading, please forgive me.
Four trends have driven the stratospheric growth of the mobile ad ecosystem over the last decade, from nothing to about $36bn/year in mobile ad revenue globally: device growth, the shift to programmatic, dollars following attention, and the invention of performance marketing.
The first main trend is device growth; at one point (2006?) nobody had a smartphone, and now everybody in the world does. That creates a bunch of people who you can show ads to. This is an Outkast-style trend - everyone’s on it. This provides a growing supply of users who will watch ads.
Second, global ad spend - from top agencies like WPP and Omnicom down to in-house teams - has been shifting from direct deals, the martini-sipping Mad Men thing where people buy naming rights for a sports team, or exclusive rights to advertise on a magazine cover, to programmatic deals, where each impression is bid for individually. This is Jeff Green’s world, and we’re all just living in it.
Third, there’s the famous Mary Meeker graph - as users moved to spend more time on their phones and laptops, advertisers followed them, under the basic assumption that time spent should be proportional to ad spend. And in 2020 and 2021, people spent more time on their phones than ever before.
These latter two trends combined to create demand for mobile ads from advertisers - they would spend money to get their brand in front of users. That being said, maybe this is more of a Katy Perry vs. Dizzee Rascal moment - the expectation is not that we’ll get 100% penetration, but that mobile programmatic ads would rise from 0% of global ad spend (before 2010) to … more than that.
Once the ad networks figured out how to make mobile advertising work (not a straightforward process!), mobile game companies could get their nascent UA and AdMon teams on the case. But underlying this whole thing was a fourth key shift - from brand to performance marketing.
Back in the day, marketing was hard to measure - you could put ads in front of people, but you wouldn’t be able to tell if they worked. As Sam Wanamaker once said “Half the money I spend on advertising is wasted; the trouble is, I don't know which half.”
But in mobile, you could track clicks and follow user IDs, laying a trail of cookies and links to figure out exactly which ads were working on which users. That leads to performance marketing - using last-click attribution, you can do a decent job of figuring out exactly where each user comes from, and the ROAS earned on each dollar you spend. Every business that advertises sits somewhere on this spectrum from brand to performance - and the mobile game companies were right out on one end.
With these tools, UA teams can precisely target ads and measure channel performance. Never mind Ogilvy and Saatchi - this job is about capital allocation, not creativity. However, with that comes public outrage about “scandals” like Cambridge Analytica; and well-meaning, if shakily-grounded, concerns about privacy. If you’re curious, you should read this essay.
The combination of device growth, precise targeting, and programmatic demand following users to mobile created immense growth between 2014, when Applovin was founded, and 2021. These trends put Ferraris in parking lots across Israel, and helped my boss afford season tickets at both Arsenal and Spurs at the same time. But in the last few years, things have started to turn.
Apple’s 2020 WWDC conference announced the end of IDFA - ID For Advertisers. IDFA was a persistent identifier that allowed iPhone users to be tracked across apps, plausibly a breach of privacy but invaluable for ad tracking. Without this identifier, ad networks know less about users, which causes them to bid lower, and UA teams can’t measure their spend. Imprecise targeting creates inefficiency, sucking cash out of the system.
At the same time, smartphones have saturated the global market, and we’ve emphatically come down from pandemic screen-time highs. Advertisers have moved more spend over to mobile, but their budgets are being cut in tougher economic times, and higher interest rates ratchet up the bar for a 365-day payback model.
I’ve got no idea what the future holds for the mobile app ecosystem - it’s probably got something to do with AI, but I couldn’t find a way to build a business there. I’ve immensely enjoyed working in the space, and it’s been a wonderful education.