In a past professional life, I wrote algorithms on a commodity trading desk for an institutional investment fund. And over the past year, I’ve come to recognize the market for mobile users (which I call the secondary market for mobile users, or SMMU) as being fundamentally similar to that of other commodity spot markets, on which commodities are purchased for near-immediate delivery.
A number of differences between the SMMU and other commodity spot markets do, however, exist. The first difference lies in the quality of goods transacted over: regulated commodity markets standardize the quality of commodities traded, whereas the SMMU is completely unregulated. The quality of users delivered on the SMMU is random until the number of users bought reaches a very large number, at which point it converges around the population mean.
The second fundamental difference between the SMMU and other commodity spot markets is that the SMMU is completely opaque because the market’s technological infrastructure has intentionally been kept immature. Orders are submitted to many brokers (ad networks) via email; the real-time bid / ask spread is unknowable. This will change very soon as real-time bidding systems become more commonplace, but even then I doubt very much that a sophisticated market monitoring service will emerge in the next year. The brokers prefer this: given the demand / supply dynamic of the SMMU, brokers benefit from informational asymmetry and lack of transparency.
These structural differences, combined with the current state of mobile, lead to some interesting market dynamics:
- Informational asymmetry benefits sellers. Sellers know a user’s spending history and likelihood of converting (albeit only in their apps). Sellers know the ask price and the bid price. Buyers know nothing in this dynamic ex ante except their own bid price.
- Scale bestows a tremendous advantage. A massive developer buying and selling tens of thousands of users per day has a massive information advantage over a smaller developer; the large developer can track price trends on the SMMU (seasonality, etc.), it has more pricing information, and it has a better understanding of the true mean value of pools of acquired users. A developer buying one hundred users knows nothing about his purchase and nothing about the market; a developer buying one million users knows everything about his purchase and everything about the market.
- Neither buyers nor sellers can be assumed to act rationally. When I say, “act rationally”, I mean act in their own economic best interests based on careful analysis of the market and their own commodity portfolio (ie users). Size, in this case, doesn’t imply sophistication: many large developers lack even the most basic analytics system, and some small studios utilize cutting-edge analytics to the fullest and shrewdest extent.
So how should these factors influence user acquisition strategy? For large players (massive developers with deep pockets), the optimal strategy is simple:
- Build out a sophisticated analytics system. Track network prices and develop a precise LCV prediction mechanism.
- Buy enough users to ensure that the sample approaches the characteristics of the entire population.
- Sell users that don’t otherwise monetize in-app.
But this strategy is out of reach for most small developers that don’t have the technical expertise or resources to build and properly utilize an analytics system. How can these small players participate in the market without becoming victims of structural biases that favor much larger entities?
The fact of the matter is that the user acquisition strategy for a small player can’t rely solely on the SMMU; the dynamics of the SMMU create a situation where small players are playing suckers’ odds. For small developers, the SMMU must be only a component of a broader user acquisition strategy, revolving around these points:
- Build apps that are inherently viral and benefit from network effects. This is the foundation of a sustainable user acquisition strategy for a small developer, and it starts at the very most nascent concept stage. Small developers cannot compete on the SMMU; user acquisition must therefore be concentrated in non-market, free, virality-based mechanics.
- Purchase users on a “futures” basis at a locked-in price. CPA prices are one-directional in the long term; locking in a fixed price for a set amount of users from other developers alleviates the market risk of a massive launch from a large player (which causes CPA price spikes) or any other unpredicted price increase. A service like ChartBoost can facilitate traffic trades between parties.
- Don’t launch with a burst campaign. Seed the game with an initial set of users and track their behavior and spending patterns. Buying users without any knowledge of how previous users have monetized and engaged with an app is reckless; LCV can’t be predicted without an existing behavioral dataset.
- Hold the marketing department responsible for spending less on users than those users can be predicted to spend in-app. The CPA / LCV spread should be quantifiable, visible within the organization, and in favor of the developer. If this metric isn’t being reported, the user acquisition strategy isn’t being vetted adequately.
- Build out a portfolio of acquisition channels. App discovery is in the throes of disruption; user acquisition managers should keep abreast of emergent distribution / discovery services and utilize them on an early-adopter basis.
Utilization of the SMMU is a necessary evil for the smaller developer; over-utilization, however, can lead to costly acquisition campaigns that deliver very little in terms of return on investment. The “spot” market for users is vulnerable to massive swings, perennially upward-trending, and inscrutable without proprietary data. For small developers, the SMMU is a losing game.
Recent