Media buying · Budget decisions

Google Ads conversion lag: when is ROAS ready to judge?

By Mirow · October 9, 2026 · 6 min read

Judge Google Ads ROAS after allowing for the purchase delay you actually observe, and compare traffic groups at the same age. A two-day-old result and a two-week-old result have had different amounts of time to collect purchases. Keep a firm spending limit while the newer result develops. Conversion lag is a reason to postpone a verdict, not permission to spend indefinitely.

The useful question is: how much of this result is still missing, and can the business afford to wait for it? You need a reporting checkpoint and a cash limit before the test starts.

Why recent ROAS can look worse

Conversion lag is the time between an ad interaction and the purchase it receives credit for. A shopper might click on Monday and buy on Thursday. Google Ads' primary conversion columns assign that purchase to the click date. Monday's spend can therefore stay fixed while Monday's reported conversion value changes. Google's separate “by conv. time” columns use the conversion date. Google Ads reporting dates.

Google explicitly warns that recent periods can look weaker because their spend has arrived but some purchases have not. Its estimates of remaining conversions are estimates, and conversion adjustments can move reported results down as well as up. Google Ads conversion delay.

This does not explain every disappointing day. A broken checkout, an unavailable product or a bad offer can also reduce sales. Check those problems immediately. If Google Ads and store revenue disagree, use the revenue reconciliation checklist first; waiting will not repair the wrong purchase value.

Choose the checkpoint from your history

Start with the purchase conversion action and the campaign type you intend to judge. Write down the account time zone, currency, attribution model, conversion window and value definition. Keep those settings consistent across the comparison.

To inspect historical delay, Google recommends a report ending at least 30 days ago, or further back for a longer conversion window. In a campaign, ad group or keyword report, segment by Conversions → Days to conversion. Google's delay-report instructions.

Read the distribution rather than borrowing a universal waiting period. If valuable purchases keep arriving after day seven, day seven is an interim review. If most of the value in several comparable historical groups is already visible much earlier, an earlier review may be useful. Check both purchase counts and value: a few late, high-value orders can matter more than many small ones.

The conversion window is the period in which an eligible purchase can receive credit. It is not proof that everyone buys at the end of that period. A shorter window also hides purchases outside it; changing the setting applies to conversions recorded going forward. Do not change the window to make a weak test look finished. Google Ads conversion windows.

Save dated exports for each fixed click-date group as it ages. Keep the selected click dates fixed while the export date advances. This preserves what you actually knew at each decision. A fresh export of an old date range shows today's updated values, not necessarily what the dashboard showed on the original review day.

Compare two-day results with two-day results

Here is a hypothetical example, not Mirow client data, using click-through purchases from a Search campaign. Each cohort is one day's ad clicks, with its full $3,000 of spend. “Age two” means two full days after that click day ends. The older cohort has saved snapshots at ages two, seven and fourteen. The newer cohort has reached only age two.

Assume one purchase action, with the store sending $100 of product sales value per credited purchase, after discounts and excluding tax. No refunds occur in this simplified example. Use whole purchase credits for readable arithmetic; the offer, product mix and reporting settings stay the same. Actual attribution can allocate fractional credit, so do not round account exports into whole orders. Google's conversion-column definitions.

Scroll horizontally to see all columns.

Comparing the newer cohort's 2.6x with the older cohort's 4.0x suggests a 35% decline. Comparing both at age two gives a 2.5% decline, calculated from the unrounded values: ($7,800 ÷ $3,000) ÷ ($8,000 ÷ $3,000) − 1. The displayed ratios are rounded to one decimal.

The newer cohort may still be worse. One comparison does not establish a trend or statistical significance. It does show that the apparent 35% decline includes a large difference in observation time.

The older cohort had recorded 80 of its 120 age-fourteen purchases at age two, or 66.7%. The projection 78 ÷ (80 ÷ 120) = 117 purchases assumes that timing repeats. Do not book those 117 purchases as earned. That would be a forecast assuming the old delay pattern repeats. Day fourteen is only this example's last snapshot, not a guarantee that all purchases or returns have settled.

Keep the budget decision in dollars

Now add a hypothetical $50 variable cost per credited purchase: product, payment fees, packing, fulfillment and shipping paid by the business after customer shipping charges. Refunds and return-related costs are zero in this simplified scenario. Include them in your own calculation, deducting each refund once and accounting separately for handling or shipping losses.

At age two, the newer cohort's modeled contribution after its media spend is 78 × ($100 − $50) − $3,000 = $900, or 11.5% of its $7,800 credited sales. The older cohort at the same age leaves $1,000, or 12.5% of $8,000. Both amounts are before fixed overhead.

These are estimates allocated to platform-credited sales. They are not reconciled store profit, cash received or proof that the ads caused the purchases. For the business decision, reconcile actual store sales and variable costs against the full media bill. If those costs are unavailable, show contribution as unavailable. The break-even ROAS guide explains the cost basis.

For example, a team might authorize at most another $1,000 of spend before its next review. That is a hypothetical risk limit set from available cash, not a recommendation to add $1,000 to this campaign. State who can stop the spend and what evidence would justify extending it. Historical lag cannot guarantee that future orders will fund today's bill.

Decide what can wait

Use a short decision table to separate an incomplete result from a problem that needs action now. “Wait” should always have a named review date and a spending limit.

Scroll horizontally to see all columns.

Before the next budget review, bring one saved age-matched comparison, the amount still authorized to spend, and the date you will examine the same click cohort again. Keep those fields in your weekly paid media report. That gives the team a specific decision today without pretending the newest ROAS is final.