“Churn” is a surprisingly imprecise term.
Some ways to “manipulate” churn, which may be 100% legitimate:
- Don’t count churn until after 90 days. This is quite common in self-service businesses. If a customer churns in a week, why count them?
- Don’t count churn if they come back within 365 days. This is less common, but not super uncommon, in apps with more episodic use. If we only use your app a few times a year, we may cancel but come back in 90 days again as a paying customer.
- Don’t count churn in pilots. This is the enterprise version of the first point. Should paid pilots count in ARR? Companies differ here. But if you include pilots in ARR, and even 30% of them don’t convert — that will hurt your churn a lot.
- Segment out your low-end customers (or some other segment). This is pretty common. The smaller the customer, the higher the churn, usually. Asterisk out your small guys and your churn goes way down.
- Don’t also measure NPS, and pair NPS/CSAT and churn together. If you just measure churn but not NPS and CSAT, you won’t really know what’s going on. Low churn with low NPS is still not a recipe for long-term success. Hiding poor NPS is a problem.
I don’t think any of these are wrong or even manipulative in fact. I think you in fact should segment churn almost always, unless all your customers really are the “same”.
Most important is measuring churn carefully and constantly — and setting goals to driving it down. Less important is exactly how you calculate “base” churn. Just improve it. Every quarter.