Your platform's first pilot has ended, and some participants want to continue. Before commissioning a large expansion, assemble a decision register: who completed useful work, who returned, what your team handled manually and which costs you can explain. Enthusiasm in a demonstration meeting cannot replace this evidence.
This evaluates a running pilot rather than listing first-version features. Evidence may justify adding a limited customer group, improving and remeasuring a particular workflow, or pausing. Explain the decision and its limits instead of converting every comment into another development feature.
Fix the unit of measurement first
In a fictional platform tracking equipment inspections for maintenance companies, several accounts may belong to one organization. Account count is not paying-customer count. Decide whether you are assessing a person, organization or completed task, retaining their relationships so employee invitations do not appear to be new commercial expansion.
Define useful activity: completing an inspection report that reaches its intended reviewer, rather than merely opening a page. Separate team tests from participant use, invitations from actual starts, and free trials from documented continuation agreements. These proposed definitions need product-owner approval.
Returning requires an observable window
Mixpanel documentation defines retention through starting and returning events and a measurement period, noting that some observation periods remain incomplete. Use the principle to select a sensible return interval for your task. A product used for periodic inspections need not be opened daily to provide value.
Do not compare someone with a complete observation period against someone who joined near its end and treat non-return as confirmed failure. Mark incomplete observation windows where appropriate and separate untrackable accounts. A percentage, especially from a small or selected sample, does not prove product-market fit across the whole market.
An evidence register before the next-stage decision
| Area | Pilot evidence | Question before expansion |
|---|---|---|
| Value and use. | Completed useful tasks, an identified sample and a consistent definition. | Does the result depend on customer work or team intervention? |
| Return and continuation. | Groups by start date, complete observation windows and explanations of interruption. | Which group is sufficiently understood and which needs more time? |
| Operational capacity. | Faults, interventions, support and manual steps. | What will accumulate with more customers, and who will handle it? |
| Cost. | Consumption, services and operational work with known sources and limits. | What is measured, approximately allocated or missing? |
This editorial register is not a scoring model automatically authorizing expansion. Add each finding's source, period and reviewer, including evidence that contradicts expectations. Commercial signals may be encouraging while operations reveal a blocker to resolve before broader access.
Expose manual work supporting the pilot
Your team may have created customer spaces, repaired files and completed tasks for participants. Do not hide that help inside a successful-use total. Record the intervention, frequency and reason: does it compensate for interface problems, training needs or an intentional service sold with the product? Each explanation affects the next stage differently.
GOV.UK's beta guidance discusses learning with a limited user group and support readiness. We draw on testing the ability to deliver service, without importing a government assessment procedure or schedule into commercial SaaS. Expansion needs review of unresolved problems and their owners, not merely another user count.
One divided bill is not confirmed cost per customer
AWS explains that SaaS customer-cost understanding requires connecting consumption to resource costs, with attribution precision depending on architecture. Ask what can be assigned and what remains shared, adding operational labor and services absent from the hosting bill. This article sets neither selling prices nor target profit margins.
If consumption has not been measured, make evidence collection an output of the next stage. Do not assume the pilot's average remains constant for every customer type or scale. Define a bounded expansion decision naming its audience, scope, assumptions, next review and conditions pausing additions when a material problem appears.
The first-pilot plan addresses what to build to begin learning. Here, send the results register to Nextwave's SaaS development team to turn remaining uncertainty into defined work, without assuming that pilot completion guarantees successful expansion.
References and decision limits
Documentation on Mixpanel retention, the beta phase and customer cost in AWS SaaS Lens was opened on 4 November 2026. The example and register are original illustrations, without customer results or revenue forecasts. AI assisted preparation, with information and translation review.
