Every SaaS forecast has a churn assumption at its heart, and in most of them it is a single number: 2 percent a month, say, applied to the whole base. For a young business with one plan and one kind of customer that is fine. For anyone else it produces a forecast that is confidently wrong, because churn in a real subscription business is not smooth. It is lumpy, concentrated and predictable once you look at it the right way.
Why blended churn misleads
Imagine a business with two segments. Enterprise customers on annual contracts churn at 8 percent a year, almost entirely at renewal. Small business customers on monthly plans churn at 4 percent a month, mostly in their first ninety days. The blended monthly churn might come out at 2 percent, but applying 2 percent evenly to next year's base gets both segments wrong. It understates the risk in the SMB cohort you just acquired and it ignores the cliff when a block of enterprise contracts comes up for renewal in March.
The same thing happens across tenure. Customers who have been with you for two years rarely leave. Customers in month two leave all the time. A blended rate treats them as the same.
Forecast from the base, by dimension
The fix is to forecast retention rather than churn, and to do it by the dimensions that actually explain behaviour. In most subscription businesses three dimensions do most of the work.
Segment. Plan tier, customer size or market. These usually have different contract lengths and different renewal dynamics.
Tenure or lifecycle stage. New, establishing, mature. Retention curves by tenure are remarkably stable within a segment, and they let you forecast the fate of a cohort from the day it signs.
Contract timing. For annual or multi-year contracts, the renewal month. Churn can only happen at renewal, so the forecast should concentrate it there.
With those in place the forecast works cohort by cohort. Each existing cohort of customers is rolled forward through its retention curve, with expansion and contraction rates applied at the same granularity. New business is layered on top from pipeline and conversion history, and each new cohort then follows the curve for its segment.
Where the data comes from
You cannot do this from the billing system's dashboard. You need invoice-level data: every invoice line, its customer, plan, service period and amount. From that you derive, for each customer in each month, a lifecycle state (new, expanding, stable, contracting, churned, reactivated) and a tenure. Those states, summed, give you the MRR bridge. Grouped by cohort, they give you the retention curves. The curves drive the forecast.
This is more work than a blended rate, but it is work you do once. The classification logic, once built, runs every month, and the forecast updates as the curves update.
Expansion matters as much as churn
Net revenue retention above 100 percent is what makes a SaaS business compound, and it comes from expansion within the base. Model expansion by the same dimensions. Customers typically expand at particular moments: after onboarding, at renewal, when they cross a usage threshold. A forecast that applies a flat expansion rate misses the timing and usually the size.
Testing the forecast
A retention-based forecast can be back-tested. Take the base as it stood twelve months ago, run it forward through the curves you have now, and compare with what actually happened. If the model would have forecast last year well, you can trust it for next year. If it would not, the error tells you which dimension you are missing. Blended-rate forecasts cannot be tested this way, which is one reason they survive so long: nobody can prove them wrong until the year is over.
What this looks like in practice
A monthly forecast cycle built this way takes about a day once the data pipeline exists. The output is an MRR forecast by month and segment, with the churn and expansion assumptions visible per cohort, a bridge from the opening base to the forecast closing base, and a list of the renewals and at-risk accounts that drive the near-term numbers. A board can interrogate any line of it, and the finance lead can explain any movement in terms of real customers.
Where to start
If your forecast is a growth rate on last month, the first step is to build the lifecycle classification from invoice data and look at the retention curves by segment and tenure. The curves alone will usually change how you think about the business. The forecast follows naturally from them.
This is the problem our SaaS analytics work addresses for clients, and it is the problem the Profitcura Revenue Intelligence platform is built to solve at scale. Either way, the principle is the same: forecast the customers you have, by how they actually behave, and the MRR takes care of itself.