The customer ID was shrinking lifetime value
Rebuilding the model around stable customer identity, complete payment paths, and mature cohorts changed which numbers were safe to use.
- Product record
- Wrong grain
- Stable account
- Customer grain
- Mature cohort value
- Decision-ready
The question behind the work.
A confidential subscription business needed a defensible view of customer value across plans, cohorts, and markets. The existing model mixed renewable product records, payment activity, and customer identity in ways that could distort lifetime and value.
“How should customer lifetime value be calculated when one customer can hold several renewable records and payments can follow more than one path?”
Evidence available
- Account, renewable-product, order, and payment histories
- Renewal paths and customer-identity relationships
- Plan economics, billing cadence, cohorts, and markets
- Retention curves and comparable observation windows
From raw evidence to a defensible answer.
Choose the customer entity
Trace how accounts, products, orders, and renewals relate so one person is not fragmented into several short-lived customers.
Reconcile every payment path
Separate completed, failed, and cancelled activity and confirm that alternative renewal flows enter the model consistently.
Align observation windows
Compare cohorts only when they have had the same opportunity to renew and distinguish mature value from incomplete recent history.
Separate commercial questions
Model initial conversion, cash flow, retention, and lifetime value independently before connecting them to acquisition decisions.
Set decision guardrails
Use ranges and sensitivity checks to define defensible acquisition and plan-level questions rather than one over-precise CLV figure.
The findings that survived validation.
A renewable-product identifier could fragment one customer across successive records, compressing apparent lifetime.
Incomplete or blended payment paths could change both retention and value conclusions.
Recent cohorts looked weaker when compared with customers who had more time to renew.
Plan conversion, collected cash, retention, and lifetime value needed separate definitions before they could guide investment.
What changed next.
One renewable record per lifecycle stage
The record was useful operationally but unsafe as the customer key.
Rebuild the model around a stable customer entity and map renewable records beneath it.
Payments in several flows
A single transaction path could omit legitimate value or include invalid activity.
Reconcile all payment routes and define status handling before calculating value.
Lower value in recent cohorts
Part of the apparent decline could be incomplete observation rather than worse customers.
Use equal exposure windows and clearly label observed versus projected value.
What the client could use next.
Delivered
- Validated customer-entity model
- Payment-path reconciliation
- Comparable cohort framework
- Plan-level value logic
- CLV assumptions and sensitivity note
- Acquisition decision guardrails
Where the evidence stops
- The public reconstruction withholds all company, market, date, plan, and monetary values.
- Projected value remains assumption-dependent and should be shown as a range, not a guaranteed outcome.
- This case documents model validation and decision design, not a measured post-implementation revenue result.
Have a question your dashboard cannot answer?
Start with the grain, evidence, and test before another metric becomes the strategy.