Mobile game user segmentation through monetization behavior

User acquisition campaign targeting based on LCV — with the goal of spending less on a user than the revenue she generates — is a cost-reduction strategy: it doesn’t provide any upside to revenue generation. Once a user is in the ecosystem of an app, however, she still needs to be targeted, and her spending behavior in-game gives you ample information to categorize her via user segmentation. Each customer segment should be handled differently, because each customer segment responds to different stimuli and has a different potential for further monetization. The way a user engages with your app based on a very limited amount of time can provide extremely valuable insight into how to most profitably engage with her in the future.

I generally categorize users into three broad groups, defined below (with a rough estimate of what percentage of a game’s total playerbase they represent in parenthesis):

Those who pay enthusiastically (.5-1% of total)

Some users will purchase items in your game almost immediately — it’s part of their evaluation process. If these users retain, they’ll purchase items whenever they reach a pain point. Once you’ve identified these users, you should continue to provide them with a steady stream of opportunities to make purchases — cosmetic items, one-of-a-kind items, and big-ticket items all appeal to these users. The speed with which they make their first purchase, and what that purchase is, should help you identify the enthusiastic buyer. Keep these users happy with new content and they’ll continue to make purchases.

Those who pay reluctantly (2.5-4% of total)

The other sub-group of buyers are the value buyers — payers who will buy items only when presented with an extraordinary opportunity or discount. These users respond well to deep-discount sales and combination bundles. If you can convince these users that the value they’re getting (ie the pain point being resolved) with the purchase is worth more than the price of the item (usually via a reduction in future time commitment), they’ll make a purchase. These users generally make their first purchase mid-game, after they’ve already passed through a number of pain points. Deep-discount sales are the best means of getting these users to continually monetize in-game.

Those who will never pay (95-97% of total)

The vast majority of your players will never, ever monetize on their own. That doesn’t mean they don’t like the game — it could mean that they don’t have access to money (children), it could mean they perceive the existence of a social stigma around buying virtual goods, or it could simply mean they’re cheap. Whatever the case, unless you mechanize your game UI to a very granular level, you probably can’t identify these users very early on, although Bayesian algorithms can help you reduce the amount of time required before categorizing these users. The obvious method of monetizing the perpetual non-payer is with ads; if they like the game enough, they should be willing to endure the presence of ads in gameplay. And if this group represents a percentage of the total playerbase higher than 97%, it could be serving as a canary in the coal mine regarding your monetization mechanics.

Identifying the players in each of these groups requires some sort of classification system, be it sophisticated (such as a real-time classifier) or primitive. Likewise, a tool is needed to reach out to these players to engage with them based on the group they place themselves in. Usually this tool is a component of a broader CRM system; if this CRM system cooperates seamlessly with the BI stack, then optimizing monetization can be done automatically. If not, it can be a time-consuming, labor-intensive process.

I like to think of the user acquisition / LCV relationship as a perpetual system that achieves equilibrium over time (and, generally, a BI system should be built to do just this). User acquisition campaigns are budget-limited by LCV, but LCV can be augmented by optimizing monetization through behavioral segmentation — which thus influences further user acquisition campaigns. After acquiring enough data, with a decent machine learning infrastructure, your LCV calculations per channel should provide for efficient investment in acquisition spending.