Churn Model

On these pages, I’ll write about projects I’ve undertaken or am working on at the moment.

Poker is an annual cyclical game (monthly, weekly, and daily too, but those are less interesting for now). There is a dip in the summer, caused by worse acquisition and worse churn rates.

The purpose of building a churn model is to run simulations of where the site will be if this calendar season is like the last. Let’s say that churn is random within a certain range, based on things like the weather in Stockholm, what’s on TV, if more weak players randomly played yesterday, and so on.

Given that, you can model what happens over the next few months, and present 95% confidence interval scenarios. You can then use that to ensure your acquisition and retention investment keeps you above certain minimum liquidity levels (as discussed in the liquidity project).

You can also see the impact of e.g. improving your software so that people play more days in a month, or e.g. the effect of running a reactivation campaign on your liquidity levels.

I’ve built models like this for two sites, using Excel. I’d like to do it with Python instead, as I can run many more simulations with much more ease, not to mention much faster. This is another product I’m interested in selling to other poker operators.