Yesterday’s CPA may still be unfinished

Monday spends $1,000 and shows ten purchases by Tuesday morning. The previous Monday shows twenty purchases on the same spend. It looks as though cost per acquisition doubled.

But the older group has had an extra week to convert. The comparison may be between an unfinished result and a more mature one.

Conversion lag is the time between an advertising interaction and the recorded outcome. Google’s conversion-lag reporting guidance describes how this delay affects metrics such as CPA and ROAS.

Compare cohorts at the same age

A cohort groups activity by a common starting point, such as the date of the ad click. To compare Monday with a previous Monday, inspect both after the same elapsed time where the reporting system supports it.

Consider this fictional history:

Time after the advertising day	Recorded purchases	CPA on $1,000 spend
One day	10	$100
Three days	16	$62.50
Seven days	20	$50

These are successive views of one cohort, not additional purchases to add together. A one-day result of ten purchases could be consistent with that history.

It could also be the beginning of a real deterioration. Lag explains why caution is needed; it does not prove that missing outcomes will arrive.

Separate customer delay from reporting delay

A customer might wait three days before buying. Separately, a completed purchase may take time to appear in a processed report. An offline import can introduce another delay.

Record the event time, import time where applicable, and report retrieval time. Check whether the report attributes the outcome to the interaction date or the conversion date.

The offline-conversion guide explains why an upload timestamp should not replace the time the business milestone occurred.

Avoid comparing an interaction-date report with a transaction-date order report without accounting for that difference.

Use an operating view and a decision view

The operating view monitors spend, availability, errors, and early signals. It can identify a broken checkout immediately, even while conversion performance remains immature.

The decision view uses an age-appropriate outcome window. Define the review delay using observed behavior for that conversion action, rather than selecting an arbitrary rule for every campaign.

Separate actions with different journeys. A newsletter signup and a completed enterprise sale are unlikely to mature on the same schedule.

Use the CPA calculator for arithmetic once the spend and outcome cohort are aligned.

Make the uncertainty visible

Label recent periods as provisional and show when they will be reviewed again. If the platform supplies a prediction, distinguish that estimate from recorded outcomes.

Keep snapshots so the team can see how results developed. A consistent delay pattern supports a more useful monitoring threshold; a sudden change in that pattern may itself indicate a problem.

The immediate task is to distinguish “not yet observed” from “not happening.” That requires time-aligned evidence, alongside checks that the campaign and customer journey are still functioning.