Building a subscription model before the price was settled
A mobile math game was preparing a subscription offer across multiple markets while pricing, conversion and much of the cost base were still being tested.
Editable market clusters supported localized pricing without duplicating the entire model.
A trial stage separated product interest from the first paid subscription event.
Both the churn rate and its measurement period remained visible assumptions.
The architecture allowed longer prepaid plans to be introduced later.
WHY THIS WASN’T A TEMPLATE EXERCISE
The model had to respect
how the business actually moved.
The model architecture separates subscriber movement from price, geography and billing cadence, turning an incomplete launch brief into a system designed to learn with the product.
The most important input was still unknown
Subscription price was expected to emerge from live paywall experiments. Hard-coding one price would make the forecast obsolete as soon as a test won.
Multiple markets did not mean multiple models
Acquisition, pricing and conversion needed to vary by geography while subscriber logic stayed consistent and easy to audit.
The brief contained deliberate blanks
Headcount, operating expenses, assets and capital expenditure were not finalized. The structure had to surface those unknowns without pretending they were answered.
MODEL ARCHITECTURE
From operating activity
to a decision-ready view.
Each layer has one job. Together they keep the commercial story, unit economics and cash consequences on the same timeline.
Acquisition by market
Paid-media inputs create new prospects inside editable geographic groups rather than one blended global audience.
Trial conversion
A short trial separates product interest from the first paid subscription event.
Subscriber cohorts
New, retained and churned subscribers roll forward by plan and geography, with the churn period made explicit.
Price experiments
Paywall experiment outcomes can update a pricing table without changing the subscriber engine.
Revenue and cash
Monthly, quarterly and annual plans keep earned revenue distinct from the timing of customer collections.
WHAT THE ANALYSIS SURFACED
Useful answers,
without exposing client data.
The takeaways are intentionally qualitative. Exact assumptions, calculations and outputs remain inside the confidential client model.
The churn rate still needed a clock
A churn percentage cannot be interpreted responsibly until its period is confirmed. The model therefore treats both rate and cadence as visible assumptions.
Geographic pricing belongs in a mapping table
One subscriber engine can serve multiple market groups when local prices, currencies and acquisition assumptions are maintained as editable mappings.
A/B tests should update the model, not break it
Separating price from subscriber movement allows a winning test to flow through recurring revenue and cash without rebuilding historical cohorts.
Billing cadence changes cash before it changes demand
Quarterly and annual prepayments can improve near-term cash while the underlying subscription service is still recognized across its coverage period.
MODELING APPROACH
The working system
behind the answer.
- Geographic acquisition and pricing matrix
- Trial-to-paid conversion and subscriber cohort engine
- Paywall experiment input layer
- Monthly, quarterly and annual plan schedules
- Marketing, headcount and operating-cost assumptions
- Recurring-revenue, scenario and cash summary
CASE CONFIDENTIALITY
This anonymized case explains the subscription logic without exposing the product, markets, exact rates or source workbook. Proprietary assumptions have been removed or generalized, and the illustration is original concept art rather than a client interface.