Watch churn and runway while the numbers are still small

Small samples need different reporting

Most subscription analytics is designed for a business with enough customers that a monthly rate is stable. Below a few hundred customers it is not. A single cancellation swings a churn percentage by more than any intervention you could plausibly run, so a dashboard of headline rates produces a monthly ritual of over-reacting to noise.

What survives at this size is the detail underneath. Which specific customers left. What they said on the way out. Whether the ones who signed up after you changed the pricing page are behaving differently from the ones before it. Bastle is built so those are one click from the headline rather than an export away — every figure opens onto the records that produced it, and at your volumes that record set is short enough to read.

Cohorts before rates

Cohort retention tables group customers by signup month and follow each group forward, which is the only view that answers whether you are getting better. A blended churn rate mixes a cohort you acquired badly two years ago with the one you acquired carefully last month and reports their average, which describes neither.

Revenue retention cells above one hundred per cent mean a cohort expanded — usually the most useful cell on the page pre-PMF, because expansion is evidence that the people who stayed found more value, which is a stronger signal at low volume than an acquisition number. See how the metrics are calculated for the mechanics.

Ask at the moment, not a week later

Cancellation Insights intercepts the cancellation itself: reason captured in the flow, an offer matched to that reason presented immediately, follow-ups scheduled per reason afterwards. At pre-PMF the reason matters more than the save. Twelve people telling you the same thing in their own words is the most valuable dataset you have, and it only exists if you ask while they are still there.

Reasons are then weighted by the MRR they took rather than counted, which changes the ranking more often than people expect: the reason mentioned most is frequently not the reason costing most.

Involuntary churn is not a growth problem

Some share of the customers who stopped paying did not decide to. Cards expire, issuers decline, retries fail. Recover runs delinquency, card expiry and annual renewal as separate campaigns across email, SMS and an in-app banner, with a hosted card-update page that writes the new card back to your processor.

The part that matters for measurement: recovered revenue returns as a reactivation into the customer's original signup cohort, not as a new customer. Counting it as new would inflate acquisition and retention simultaneously, which is exactly the kind of quiet flattery that makes early metrics useless.

Runway, with the assumptions in the open

Forecasting is in beta and labelled that way everywhere. It pre-fills starting MRR, customer count, growth and churn from your history, lets you choose linear or compounding growth, models expenses and hiring, and reports runway in days per scenario, with accounting actuals syncing from QuickBooks Online or Xero. Treat it as a strong second opinion rather than the sole input to a hiring decision — over one to two quarters it is reasonable, and beyond that a small error in the churn assumption compounds into a large error in the answer.