If you run Google Ads for anything that closes on a call, the most important number in your account is not in the conversions column. It is the number of days between a deal closing in your CRM and that row reaching Google. Past seven days, the conversion still shows up in your reports, but Google's attribution modelling steps over it, and modelling is what Smart Bidding learns from.
Google Ads excludes offline conversions uploaded more than seven days after the conversion event from its attribution modelling. They still appear in standard reporting columns, so nothing looks broken. Because data-driven attribution output feeds Smart Bidding, a slow CRM sync degrades bidding while your reports stay reassuring. The seven days run from the conversion, not the click.
What the rule actually says
Google's offline conversion import documentation ties upload latency directly to model accuracy: conversions imported more than seven days after they occurred are bypassed by attribution modelling, though they continue to appear in standard reporting columns. PPC Land reported the change in 2026, and the significance is in the second half of that sentence. Nothing errors. No diagnostic turns red. Your conversion count is unchanged. Only the data the bidding algorithm is allowed to learn from shrinks.
That matters because data-driven attribution is not just a reporting model. Its output is the credit assignment Smart Bidding optimises against. Conversions excluded from the model are, for bidding purposes, conversions that did not happen, even though you can see them on screen.
Two other 2026 changes are worth pinning to the same page, because accounts that got quietly broken this year usually got broken by one of them. From June 15, 2026, offline conversion imports and enhanced conversions for leads moved to the Data Manager API, and the Google Ads API's UploadClickConversions request was deprecated, with requests rejected from developer tokens that had not been uploading during the qualifying window. Google launched the Data Manager API on December 9, 2025 as a single first-party data ingestion layer across Google Ads, GA4 and DV360.
Three different clocks, and only one of them is seven days
Almost every confused conversation about this rule comes from collapsing three separate timers into one. They control different things and they fail in different ways.
| Clock | How long | What it controls | What breaks if you get it wrong |
|---|---|---|---|
| Click-through conversion window | 30 days by default, configurable up to 90 per conversion action | How long after an ad click a conversion can still be attributed to it | Deals that close after the window are never credited to the campaign that produced them |
| Conversion-to-upload latency | Seven days, measured from the conversion event to the moment you upload it | Whether attribution modelling, and therefore Smart Bidding, is allowed to learn from the row | Reports look complete, bidding quietly optimises on a partial, biased subset of your results |
| Conversion lag | However long your sales cycle actually is | How long you must wait before this week's performance data is trustworthy | Teams judge campaigns before the data has landed and cut the wrong ones |
The practical reading: a 60-day sales cycle is not a problem for the seven-day rule. A 60-day sales cycle with a monthly CSV upload is.
Why this hits lead generation accounts hardest
Ecommerce mostly escapes it, because the conversion fires in the browser at the moment of purchase. Lead generation does not, because the event worth optimising toward, a qualified opportunity or a closed deal, exists in a CRM and has to be carried back to Google by whatever process someone set up once and rarely revisits.
| How you sync your CRM | Typical conversion-to-upload latency | What Google models | What it means in practice |
|---|---|---|---|
| Native connector or Data Manager API, running daily | Under 24 hours | Everything | The intended setup. Bidding sees your real outcomes |
| Weekly batch on a fixed day | 1 to 7 days | Most rows, with the oldest at risk | Works until the batch slips a day, at which point the start of that week silently falls out of the model |
| Monthly upload | 8 to 38 days | Almost nothing | Your best data reaches reporting and never reaches bidding. This is the most common broken setup we see |
| Manual export when someone remembers | Weeks to months | Nothing | Smart Bidding is optimising entirely on form fills, whatever your dashboards say |
If your sync sits in either of the bottom two rows, the fix is not a better bidding strategy. It is a scheduled job. Everything else you change in the account before that is guesswork built on a model that never saw your outcomes.
Before touching bids, export closed deals for the last quarter with two dates: when the conversion happened and when it was uploaded. The median gap between those two columns tells you what share of your revenue data ever influenced bidding. Under seven days, you have a tuning problem. Over it, you have a plumbing problem.
The fix is not a faster CRM. It is a shorter conversion
Most teams respond to this by trying to compress their sales cycle, which is the wrong lever and takes a year. The seven-day rule constrains upload latency, not deal length, so the workable answer is to give the algorithm an earlier event to learn from and keep revenue as reporting. Here is a laddered setup for a business with a 45-day average cycle.
| Conversion action | When it fires | Role | Value |
|---|---|---|---|
| Form fill or call | Day 0, in the browser | Secondary, observation only | None. Counting it as primary is what trains the algorithm to buy junk leads |
| Qualified lead | Day 1 to 3, when sales confirms fit, uploaded within 24 hours | Primary bidding conversion | Expected value: average deal value multiplied by your qualified-to-won rate |
| Closed won | Day 45, uploaded within seven days of the close date | Secondary, observation only | Actual revenue, for reporting and for recalibrating the expected value above |
Worked through with real arithmetic: if your average deal is $12,000 and you close 22% of qualified leads, each qualified lead carries an expected value of $2,640. That is the number you upload on day 2, and it is a defensible one, because it is derived from your own historical close rate rather than invented. Google gets a modelled, value-bearing signal two days after the click instead of a signal it is not allowed to model 45 days later. Revenue reporting stays intact because closed-won still lands in the account, just as an observation rather than a bidding target.
Recalculate the multiplier quarterly. A close rate that drifts from 22% to 14% and is never updated means every bid in the account is priced off a number that stopped being true two quarters ago. If you are also trying to raise that close rate rather than just track it, the on-site half of the problem is a cost per qualified lead question, not a bidding one.
A six-point audit you can run this week
- Check the upload timestamps. Confirm your integration sends the true conversion date and time, not the moment the file was processed. A connector that stamps upload time makes every row look fresh while the underlying attribution is wrong.
- Measure your own latency. Median days between conversion event and upload, by conversion action. Anything at or above seven is invisible to the model.
- Check the integration path. Anything still pointed at the Google Ads API's UploadClickConversions has been on borrowed time since June 15, 2026. Confirm it moved to the Data Manager API.
- Compare your conversion window to your sales cycle. If the median deal closes on day 52 and your click-through window is the 30-day default, the campaign that produced the deal never gets credit for it, regardless of upload speed.
- Check which action is primary. If closed-won is a primary bidding conversion and arrives late, bidding is learning from whichever deals happened to close fast, which is a biased sample, not a smaller one.
- Reconcile counts. Deals in the CRM for a period, versus offline conversions Google recorded for the same period. Gaps here are usually a matching failure, not a modelling one.
Where enhanced conversions for leads fits
Enhanced conversions for leads matches conversions using hashed first-party customer data, such as an email address collected on the form, rather than relying only on a stored click identifier passed through your CRM. That makes it the more resilient route for lead journeys that pass through phone calls, sales tools and long gaps where a click ID gets lost, and it moved to the same Data Manager API path in June 2026. It does not exempt you from the seven-day guidance, and it does not fix a monthly upload schedule. It fixes matching, not latency, and most accounts need both working before offline data is worth trusting. If you are still rebuilding your measurement layer more broadly, the wider signal loss problem is the context this sits inside.
Frequently asked questions
Does Google Ads ignore offline conversions uploaded after seven days?
Not entirely. Conversions uploaded more than seven days after the conversion event still appear in standard reporting columns, but Google's attribution modelling bypasses them. Since data-driven attribution output feeds Smart Bidding, late uploads inform your reports without informing your bidding.
Is the seven days counted from the ad click or from the conversion?
From the conversion. The seven-day guidance measures the gap between when the conversion occurred and when you uploaded it. How long after the click the conversion happened is governed by a separate setting, the click-through conversion window, which defaults to 30 days and can be extended to 90.
What if my sales cycle is three months?
A long cycle is fine. Upload each conversion within seven days of the event itself, and use a mid-funnel event, typically a sales-qualified lead, as the primary bidding conversion with a value derived from your close rate. Keep closed-won as a secondary conversion for revenue reporting.
What changed with the Data Manager API on June 15, 2026?
Offline conversion imports and enhanced conversions for leads uploads migrated to the Data Manager API, and the Google Ads API's UploadClickConversions request was deprecated, with requests rejected from developer tokens that had not been uploading during the qualifying window. Google launched the Data Manager API on December 9, 2025 as a unified first-party data layer across Google Ads, GA4 and DV360.
How do I tell whether late uploads are hurting my account?
Export closed deals for a full quarter with both the conversion date and the upload date, then take the median gap. If it exceeds seven days, the majority of your outcome data never reached the attribution model, which usually shows up as Smart Bidding chasing cheap form fills while pipeline quality drifts.
Should closed-won revenue be my primary conversion in Google Ads?
Usually not, if deals take longer than a week or two to close. Late-arriving conversions are excluded from modelling, so a closed-won primary trains bidding on the subset of deals that happened to close quickly. Use a qualified-lead event as primary with an expected value, and keep closed-won as an observation conversion.
The takeaway
This is a measurement rule with a bidding consequence, and it fails quietly by design: the reports keep looking right while the model that spends your money is fed a shrinking, skewed sample. Measure the gap between your conversion events and your uploads before you change anything else in the account. If it is over seven days, fix the schedule first, then move your primary conversion earlier in the funnel and give it a value derived from your own close rate. Everything else in the account is downstream of that.
Sources & further reading
- PPC Land, "Google Ads attribution ignores offline conversions uploaded after 7 days" (2026): ppc.land/google-ads-attribution-ignores-offline-conversions-uploaded-after-7-days
- Google Ads Help, "About offline conversion imports": support.google.com/google-ads/answer/2998031
- Google Ads Help, "Fix discrepancies and errors in offline conversion imports": support.google.com/google-ads/answer/13321563
- Search Engine Land, "Google is moving offline conversion imports out of the Google Ads API": searchengineland.com/google-is-moving-offline-conversion-imports-out-of-the-google-ads-api-477669
- Two Octobers, "Digital Marketing Updates: September 2026": twooctobers.com/blog/digital-marketing-updates-september-2026