Subscription Revenue Forecasting for a Global Software Company

The situation

A software company with subscription revenue across multiple global markets and $8M of revenue struggled to forecast recurring revenue and churn with any confidence.

Recurring revenue is easy to report and hard to forecast. The reported figure is a fact about last month; the forecast requires assumptions about retention, expansion and contraction that most companies hold only informally. The gap between the two widens as the customer base ages.

Multiple markets compound it. Churn behavior differs by geography for reasons that are usually commercial rather than financial, and a blended churn rate averages away exactly the differences that would explain what is happening.

What we did

Built a subscription forecasting model on historical data, cohort analysis and churn projections. Cohort analysis is what separates a subscription model from a spreadsheet with a growth rate in it. Customers acquired in different periods behave differently, and averaging them hides both the improvement and the deterioration.

Implemented tracking of new subscriptions, upgrades, downgrades and cancellations. Four movements, not two. A business that tracks only new and cancelled cannot see contraction, which is the quietest way recurring revenue erodes.

Delivered monthly analysis of revenue trends, churn by cohort and customer lifetime value. Monthly rather than quarterly, because cohort signals are only actionable while the cohort is still young enough to influence.

What changed

Subscription forecast accuracy improved 18%, with cohort-level churn and lifetime value visibility established.

Revenue projections became reliable enough to plan on. Retention patterns informed customer success strategy. And long-term planning ran on real recurring-revenue dynamics rather than on a growth assumption.

The second outcome is where the financial work turns operational. Cohort churn data is only interesting to finance; it is actionable to the team that owns retention, and that handoff is what converts analysis into a changed number.

What this means for subscription businesses

A blended churn rate is the most comfortable and least useful metric in a subscription business. It is stable, it looks reasonable, and it conceals the two things worth knowing: whether recent cohorts retain better than older ones, and which market is carrying the average.

Tracking all four subscription movements is the minimum viable instrumentation. Contraction in particular tends to be invisible in aggregate revenue during a growth phase, and it is the earliest signal that value delivery has slipped.

Forecast accuracy compounds into everything downstream — hiring, runway, and what a company can credibly tell investors. It is worth building before the number gets large enough that being wrong about it is expensive.