Don’t mean to disappoint you, but this idea isn’t nearly as devious or dark as the name implies. It’s just a straightforward, really smart, thing to do.
The term derives from sociology: groups of people who share certain characteristics are “cohorts”. Instead of focusing on average behaviors of a large population, cohort analysis looks to find and understand patterns among subgroups, and how those change over time. Especially for new businesses, that analysis can reveal some surprising and important data… including what the company might be worth.
Probably the most common application is in computing lifetime customer value for new companies. Let’s take one with a subscription model. Obviously, a key question in a valuation will be: what’s our renewal rate look like? But just taking an overall average of all users might be very misleading when many are new. Instead, you really want to look at “cohorts” of users to see how they behave. At the most basic level, this means grouping those who’ve been around, say, 1, 3, 6, 12, months, whatever, to see each group’s renewal rates. A more sophisticated approach might add layer in other factors to further define the cohorts, like the source or type of user, geographic or demographic factors, etc. A little basic stats work should then allow us to project what renewal patterns will look like for our new users… and so what kind of lifetime customer value we can expect from them. And that lets us figure out which targets are the most promising, and what we can pay for customer acquisition. Of course, knowing your LCV, acquisition costs, and margins also provides a rough guide to future profitability.
Naturally, the analysis can get much more complicated, and useful. For example, many social gaming companies perform frequent, deep dive cohort analysis to find, say, the hidden patterns among users buying virtual goods… and then tweak things accordingly.
Probably the best thing about cohort analysis, though, is how often you’ll find surprising patterns that let you improve profitability quickly: who knew our 24- 32 year old North Dakota users buy five times more lipstick than average? And, it also keeps you from making mistakes based on “averages” that totally mask the truth: companies whose customer gender distribution mimics the general population should probably avoid spending advertising dollars targeting their average customer, the one with one testicle and one ovary.