Compare conversion cohorts only after they have had time to mature

Compare age as well as calendar dates

Google documents conversion delay; its relevance depends on the action and business cycle. Build an observed maturity curve for your own reporting definition rather than borrowing a universal waiting period. Record successive snapshots of the same historical dates, including extraction time, outcome definition and whether dates represent interactions or completed conversions. Mixing those date bases changes what a cohort means.

Educational example, not a forecast: a fixed interaction-date cohort costs 600 units. Its snapshots show 12 outcomes after two days, 18 after seven days and 20 after fourteen. The displayed CPA moves from 50 to 33.33 to 30 without any new spend in that cohort. Comparing its two-day snapshot against another cohort's fourteen-day snapshot would manufacture an apparent deterioration. These figures illustrate the arithmetic; they do not establish a fourteen-day rule for your business.

One workable method is to exclude the immature right edge of both comparison windows. If recent cohorts usually remain materially incomplete for an observed interval, report a mature comparison ending before that edge and a separate provisional panel for the latest dates. Another method compares equally aged snapshots, such as each weekly cohort exactly seven days after its end. Use whichever approach your stored history can actually support.

Why the newest week can look worse

• A mature subtotal can support a budget review while a provisional subtotal remains useful for delivery and tracking checks. You do not need to ignore all recent information.
• An equal-age comparison avoids treating the benefit of additional observation time as campaign performance. Preserve snapshots if you want to make this comparison later.
• Separate sales-cycle delay from collection failure. A missing import and a late customer decision require different investigations and should not share one vague pending label.

Create a maturity-aware comparison

1. Select one outcome, its date basis and an account timezone. Collect a few historical cohorts with more than one snapshot; do not infer maturity from a single final-looking export.
2. Calculate the additional outcomes arriving between snapshots. Inspect variation by product, weekday and outcome type before deciding whether one common interval is reasonable.
3. Choose an operating rule for material incompleteness. Document why the selected lag fits this decision; a high-cost irreversible decision may require more maturity than routine monitoring.
4. Build older and newer windows of equal duration and weekday composition, each ending outside the excluded edge. Alternatively, recover equally aged snapshots and label their observation age.
5. Publish extraction timestamps and the excluded dates alongside the result. Use read-only analysis of available AdAce Ads records; if snapshot history is absent, say so rather than inventing an eventual conversion count.

Keep assumptions visible

Sources and further reading

• Google Ads: conversion delay