- The report’s date range and each entrant’s conversion window are different settings.
- A recent entrant who has not finished is not automatically a final drop-off.
- Compare cohorts with the same eligibility, elapsed opportunity, and measurement definitions.
Separate entry date from time allowed to finish
The reporting date range tells you which entrants are included. A conversion window tells you how long an eligible entrant has to complete the defined steps. A person who arrived this morning has not had the same opportunity to finish a seven-day journey as someone who entered last week.
This matters when you evaluate a new landing page or onboarding change. A low current completion rate might reflect a cohort still in progress. It can also reflect genuine friction or missing tracking. The window check helps distinguish these possibilities; it does not explain the entire result.
Work through a mature and an immature cohort
Consider an illustrative seven-day window beginning at entry. At the end of day seven, an older cohort of 100 entrants has 20 completions. A newer cohort of 100 entrants has 8 completions after only one day. Both snapshots can be numerically correct, but they answer different questions.
You cannot infer a 60% deterioration by comparing 8% at day one with 20% at day seven. The new cohort may finish with fewer, the same, or more completions. Compare day-one rates with day-one rates for an early signal, or wait for each cohort’s seven-day opportunity to close.
| Cohort | Observed completions | Elapsed opportunity | Interpretation |
|---|---|---|---|
| Older 100 entrants | 20 / 100 = 20% | Full 7 days | Window complete |
| Newer 100 entrants | 8 / 100 = 8% | Only 1 day | Still in progress |
| Fair early comparison | Use both day-one snapshots | Same elapsed time | An early signal, not final conversion |
Choose the window from the decision
A same-visit form interaction and a considered purchase over several days are different processes. Choose a window that matches the question, document the start and end events, and keep the population consistent. Changing from a visit count to a connected-account count changes more than the time allowed.
In VisitLift, ordinary within-visit funnels and cross-visit purchase journeys have distinct definitions. Purchase journeys use a selected seven- or 30-day window and finish in verified paid Checkout under supported continuity or signed-account rules. Those settings are not a general promise of arbitrary cross-device or historical reconstruction.
Keep incomplete observations visible
Record pending entrants separately from completed and mature non-completing entrants when your analysis supports that distinction. Explain whether the report omits immature periods or displays their current partial rate. PostHog, for example, documents an option to hide incomplete periods in funnel trends.
Do not remove every recent low-converting entrant merely to make a chart look better. Apply the same preselected window rule to all comparison groups. Late-arriving records, timezone boundaries, environment filters, consent coverage, and event ordering also deserve checks before assigning a final status.
Make the next comparison reproducible
Write down entry dates, the elapsed-time cutoff, funnel steps, ordering rule, measurement unit, exclusions, and observed counts. When a change ships, compare a suitable earlier cohort with a later one using the same definitions and equivalent maturity. Note acquisition changes that could alter who enters.
Use the funnel calculator for arithmetic scenarios and the audit worksheet to record the definitions. Neither tool forecasts what the pending entrants will do. Treat an early rate as provisional and combine it with actual journey inspection and customer evidence before changing the product again.