Before you diagnose anything, check whether the window you are reading is old enough to be complete. If any day in it is less than 4 days old, the number is wrong and there is nothing to diagnose yet.

On this page

• How long does Google Ads conversion data take to settle?
• What is the difference between conversion lag and report settlement?
• Why did testing settlement on a 7-day window give the wrong answer?
• Why does my ROAS look great at month end and collapse on the 1st?
• Is the Google Ads conversions column counting what you think it is?
• What should I check first when Google Ads conversions drop?

How long does Google Ads conversion data take to settle?

A by-conversion-date figure takes about 4 days to finish being written. Measured against a saved snapshot on an apparel account, the same windows re-read 6 days later:

Four days old or more, it is finished. Younger than that, it is severely understated. At 1 day old the figure nearly doubled on re-read and cost per acquisition moved from $80.11 to $43.60 with nothing changing in the account.

What is the difference between conversion lag and report settlement?

Click-to-conversion lag is how long a person takes to convert after clicking. It describes buyer behaviour.

Report settlement is how long Google takes to finish writing a day's row. A conversion is filed on the date it happened, so there is no long tail, but the row is not complete the moment the day ends.

Never use the first to correct the second. Applying click-lag data as a settlement multiplier manufactures a recovery that is not there.

Why did testing settlement on a 7-day window give the wrong answer?

Settlement was first validated on a 7-day window whose last day or two were fresh. Dilution held the measured backfill to +3.5%, which read as "settles fast." That figure was then applied to a 3-day window sitting entirely inside the fresh zone.

Both windows "ended 1 day ago." Completely different exposure. The 3-day window read as a 55% collapse and nearly triggered a strategy reversal. It was an artifact. Nothing had happened.

Test at day grain, or on a matched-length window. Never validate settlement on a long window and apply the result to a short one. A 3-day window that is 1 day old is 100% unsettled.

Why does my ROAS look great at month end and collapse on the 1st?

Google enforces a monthly spend ceiling of daily budget multiplied by 30.4. Near it, delivery throttles for the last days of the month and resumes at full pace on the 1st.

That creates a mirror-image pair of artifacts. The pre-reset window is flattered: spend is throttled low while conversions keep arriving from earlier high-spend days. The post-reset window is punished: spend jumps back to full pace against conversions seeded by the throttled days.

Measured on an apparel account in July: $7,935 spent against an $8,056 ceiling, 98.5% consumed. Delivery cut to $212 on the 30th and $132 on the 31st, then resumed at $303 a day. The post-reset window read as a 63% collapse in new-customer efficiency. Combining the two stretches cancelled most of it.

Before reading any window that spans a month boundary: sum the month's spend against daily budget times 30.4. If it is near 100%, the last two or three days are throttled, not signal.

Is the Google Ads conversions column counting what you think it is?

The all-conversions column is a sum over every conversion action attached to the account, not a count of people. Duplicate, legacy and test actions are normal on a mature account, so the aggregate routinely over-counts by two to three times.

Measured on a telehealth account: a campaign reported 3 conversions on one day. Segmented by conversion action name it was one person registering, counted three times — once by the live action, once by a Test action nobody had removed, and once by a duplicate.

A live Test conversion action in production is a defect. Check for one on any account whose numbers look implausibly healthy.

What should I check first when Google Ads conversions drop?

Most published answers list twenty causes in no order, so the reader checks everything at once and learns nothing. Go in this sequence and stop when a gate closes.

1. Is the window complete? If not, stop. Re-read in 4 days.
2. Does the window span a month boundary? Check the monthly ceiling.
3. Is the metric what you think it is? Segment by conversion action name.
4. Did tracking change? Tag deployment, consent banner, checkout change, theme update. Check the site before the account.
5. Did somebody change the account? Change history, grouped by user. It is more often yes than people expect, and it is not always a person.
6. Only now, auction and demand. Impression share lost to rank, cost-per-click decomposition, seasonality, competitor entry.

Related questions

How many days should I wait before reading Google Ads conversion data?

At least 4. Measured against a saved snapshot, windows ending 4 or more days ago backfilled 0.0% on re-read, while a window ending 1 day ago backfilled 83.7%. Never read a window containing days younger than 4 days old.

Does conversion lag bucket tell me how much a conversion-date figure will backfill?

No. Conversion lag bucket measures click-to-conversion time, which is buyer behaviour. Report settlement is how long Google takes to finish writing a day's row. Applying one to the other manufactures a recovery that is not there.

What is the Google Ads monthly budget cap?

Google enforces a monthly spend ceiling of your daily budget multiplied by 30.4. Near the ceiling, delivery is throttled for the final days of the month and resumes at full pace on the 1st.

Why does Google Ads show more conversions than my order system?

The aggregate conversions column sums every conversion action attached to the account, including duplicates, legacy actions and test actions. Segment by conversion action name before believing any count.

Can I read the most recent few days using another data source?

Yes. Any source that keys on the order's own creation timestamp has no settlement lag, so it is the only valid lens on the most recent few days. Google's conversion-date figure literally cannot read them yet.

Written by Usama Khan, who runs Google Ads for e-commerce brands and home-services companies at AdventurAds. Figures are from accounts under active management and are anonymised. Last reviewed 2026-08-26.