Subscription forecasts: cohorts, churn, revenue and cash

Forecast subscription customers by acquisition cohort, applying retention to the customers exposed to churn and adding new activations separately. Then calculate the ending monthly recurring run rate under a stated price definition. Keep that run rate separate from revenue earned during the month and cash collected from annual plans.

Define what counts as an active paying subscription

Decide how trials, free accounts, pauses, delinquent subscriptions, discounts and annual plans enter the forecast. Use the same definition in the opening count and new-customer additions. An activation that never becomes a paying account should not automatically create a full-price recurring balance.

For the example, a customer means one active paid subscription. Price is $50 per month, with no upgrades, downgrades or discounts. New customers are all active at month-end, and churn is not applied to them in their activation month. The percentages below are explicit planning assumptions, not industry retention benchmarks.

Roll the customer cohorts into the next month

Start month one with 1,000 customers. At 4% churn, 960 remain. Add 80 new customers to finish with 1,040 and an ending recurring run rate of 1,040 × $50 = $52,000.

For month two, assume the older 960 retain at 96%, but the recent 80 retain at 90%. Add another 80 new customers. Keep the cohorts separate:

Ending recurring run rate is $53,680. Applying 96% to all 1,040 opening customers and adding 80 gives 1,078.4 customers and $53,920, overstating this case by $240. Expected forecast counts can be fractional; actual subscriber records count whole customers.

In a spreadsheet, give each acquisition month its own row and each forecast month a column. Multiply a cohort's previous ending count by the retention appropriate to its age, then add new activations only in that cohort's starting month. This makes a weak first renewal visible instead of hiding it inside one blended churn rate.

Distinguish customer churn from revenue churn

Suppose a different customer base has 80 small accounts paying $50 and 20 large accounts paying $200. Its opening recurring amount is $8,000. Losing four small accounts is 4% customer churn but $200 ÷ $8,000 = 2.5% lost recurring revenue. Losing four large accounts is still 4% customer churn but removes $800, or 10% of recurring revenue.

Use account counts to forecast support demand and customer retention; use price-weighted movements to forecast recurring revenue. Show expansion, contraction and reactivation separately if they matter. Otherwise, upgrades can offset cancellations in the revenue total while the customer count continues to fall.

Keep annual cash out of the monthly run rate

Assume ten customers each pay $2,400 upfront for twelve months of an evenly delivered service beginning on the first day of the month. Cash receipts are $24,000 at the start. Under a monthly-normalized recurring definition, those plans add $2,000 to the ending monthly run rate, not $24,000.

For this simple example, assume revenue is earned evenly at $2,000 a month with no other obligations or adjustments. Actual recognition depends on the contract and applicable policy. IFRS 15 ties revenue to satisfaction of performance obligations, which is distinct from collecting the customer's cash. A new account activated late in the month can contribute its full normalized ending run rate while generating only part of a month's earned revenue.

Forecast annual renewal receipts in the renewal month, not as one-twelfth cash every month. Refunds, nonpayment, service credits and plan changes need their own assumptions. The annual cash collected may need to support service delivery long after the payment arrives.

Connect the customer forecast to acquisition and delivery

At an assumed $100 acquisition cost, 80 new activations require $8,000 of acquisition spending if the spending and activation assumptions use the same cohort. Include its cash date and the support or onboarding capacity those customers need. A growing run rate does not prove that acquisition pays back.

Use the first-order economics guide for the acquisition-cost boundary, the staffed-hours method where onboarding depends on people, and the cash forecast to separate billing and collection. Re-estimate retention by cohort age from mature observed cohorts before treating a chosen percentage as a durable forecast.