Google's Seven-Day Offline Conversion Gate Hits Your Attribution Reports, Not Your Bidding. Google Rewrote the Page to Say So.
The short version
Google's attribution reports documentation now states that offline conversions uploaded more than seven days after the event are bypassed by the attribution reports modeling engine, including the Model Comparison tool. The same page states, in a heading Google added after the rule went viral in late August, that standard reporting columns and Smart Bidding "operate without this restriction" and optimize on late uploads across the full conversion window, up to 90 days for GCLID. So the practitioner alarm that late uploads are invisible to bidding is, on Google's current wording, wrong. What is true, and worth acting on, is narrower: any budget decision you make from an attribution report is made on a dataset that silently excludes late offline uploads, Search Partners, Gmail and App conversions. Low signal for bidding. Medium if Model Comparison is where you decide what to fund.
What Google announced
Nothing, in the usual way. The rule surfaced on August 22, 2026, when consultant Adriaan Dekker posted a passage from Google's About attribution reports help page, crediting Hana Kobzová with the find. The passage said the daily attribution modeling engine processes conversions that occurred within the last seven days and that an offline conversion uploaded more than seven days after the event "is bypassed by the attribution modeling calculations." Dekker's summary, that late uploads still appear in standard reports but are "ignored by the attribution engine," was accurate to the text and travelled fast, because the obvious next question is whether Smart Bidding, which runs on data-driven attribution, ignores them too. PPC Land's August 23 report raised exactly that question and noted the page did not answer it.
The page now does. As of this writing it carries a section headed "The 7-Day offline conversion upload gate" with two bullets. The first restates the limitation and scopes it: the engine "utilized in Attribution reports" only processes conversions within the last seven days, and a late offline upload "is bypassed by the attribution reports (such as the Model Comparison tool)." The second is new, and it is the sentence that matters: "Standard account reporting columns (such as 'Conversions' and 'All conv.') and Smart Bidding operate without this restriction and retroactively record and optimize on uploaded offline conversions according to your conversion action's configured conversion window (up to 90 days for GCLID)." Google's stated best practice is to keep upload latency under seven days so offline journeys are "fully visible across both attribution reports and standard performance views." The same section adds a second exclusion: the Model Comparison engine "is designed to explicitly exclude conversions originating from the Search Partner Network (SPN), Gmail, and App campaigns," which standard columns include by default.
What it actually means
Start with the correction, because it is the useful part. If someone forwarded you the LinkedIn post and you have been treating your weekly CRM export as a Smart Bidding problem, Google's current documentation says it is not one. Smart Bidding optimizes on late uploads inside the conversion window. A lead that clicked on the first, closed on the twelfth and uploaded on the fifteenth reaches bidding. The site's rule is to say when something does not matter, and for the bidding question, this does not.
Two things are worth noticing about how the correction arrived. First, Google clarified by editing a help page, with no changelog and no date on the section, which is how most 2026 clarifications arrive and why the original scare was possible. Second, the wording that caused the scare was ambiguous on Google's side, not the practitioners'. The August text keyed the rule to "the attribution modeling engine" without saying which one, and data-driven attribution is both a reporting model and the input to bidding. The rewrite resolves the ambiguity in the direction advertisers wanted. It does not say whether the ambiguity was ever a real gap or only a documentation one, and nobody outside Google can tell.
Now the part that survives. Attribution reports are a filtered view, and the filters were only recently written down. The Model Comparison tool is where a lot of teams go to argue about whether a campaign, network or keyword deserves budget, because it is the only place in the interface that shows the same conversions under different models side by side. Google's page now lists four ways that view differs from the Campaigns page: it counts by conversion time rather than click time, it excludes late offline uploads, it excludes Search Partners, Gmail and App entirely, and it labels Performance Max as a single "Cross-Network" row. For an offline-heavy lead-gen account with a weekly upload cadence, the Model Comparison report is looking at a minority of the conversions the account actually recorded. Any conclusion drawn from it about which campaign assists more, or which network to cut, is drawn from that minority.
The Search Partners exclusion deserves its own sentence, because it is where the money is. Practitioners have argued for years about whether partner inventory earns its spend. Google's documentation now says the report most people would use to settle that argument does not contain partner conversions at all. It contains partner spend, in the standard columns, next to a model view that has removed the conversions. That is not a bug, it is documented design, but it means the partner network can only ever look worse in a Model Comparison than it is.
One more thing Google's page does not say. The rewrite scopes the gate to "attribution reports." It does not say what happens to the Conversion paths and Assisted conversions reports specifically, which run on the same engine, and it does not say whether a conversion that occurred nine days ago but was uploaded within two days of the event clears the window. PPC Land flagged that the original wording keyed the rule to event age in one sentence and upload latency in the next. The rewrite keeps both phrasings. For most accounts the difference is academic. For a long-cycle account backfilling history, it decides whether the backfill carries any modeling weight, and the page still does not answer it.
The gate is real. It just guards a report, not the bidder. Google took three weeks to say so, and said it by editing the page.
Signal or noise: low for bidding, medium if budget decisions come from attribution reports
If your interest in this was whether Smart Bidding sees your late CRM uploads, the answer on Google's current documentation is yes, and you can stop here. If your team uses the Model Comparison or Assisted conversions reports to decide what to fund, this is medium signal and permanent: those reports exclude late offline uploads, Search Partners, Gmail and App by design, and no setting changes that. The rating is deliberately lower than the August coverage implied, because the scary half of the story is the half Google has now walked back.
The Monday-morning playbook
- If the LinkedIn post triggered a fire drill, close it. Google's current wording is that Smart Bidding optimizes on late offline uploads inside the conversion window. Forward the sentence, not the scare.
- Measure your upload latency anyway. Pull the gap between conversion time and upload time for the last 90 days and find the median. Google's stated best practice is under seven days, and a weekly export runs a median around four with a tail past seven. If the tail is large, the fix is the export cadence, and it is an engineering ticket rather than a bid strategy change.
- Stop using Model Comparison to judge offline-heavy campaigns. For any campaign where offline import is the primary conversion action, the report is missing every conversion uploaded late. Judge those campaigns on the Campaigns page with "by conv. time" columns added, which Google's own table says aligns the timing convention.
- Stop using Model Comparison to judge Search Partners. It contains none of their conversions. If the partner-network decision is live in an account you manage, make it on standard columns segmented by network, or on an experiment, and say why in the notes.
- Add the four differences to whatever reconciliation doc your team keeps. Time of event, late-upload exclusion, network exclusion, and the single Performance Max row. Attribution reports and standard columns were never meant to match, and Google has now written down why.
- Watch the page. Google edited it once without notice and may again. The open question is whether the seven days count from the event or from the upload. If you backfill history, that answer matters, and it will arrive the same way this one did.
The bottom line
A real rule, a real scare, and a quiet correction in that order. The seven-day gate exists and it filters the reports many teams use to argue about budget, which is worth knowing permanently. It does not filter what Smart Bidding learns from, which is what everyone was afraid of and what Google has now said in writing. Read the page, not the post, and keep the upload latency under seven days for the reason Google gives rather than the one that went viral.
Frequently asked
Do late offline conversion uploads affect Smart Bidding in Google Ads?
Not according to Google's current documentation. The About attribution reports help page states that standard reporting columns and Smart Bidding operate without the seven-day restriction and retroactively record and optimize on uploaded offline conversions within the conversion action's configured conversion window, up to 90 days for GCLID uploads. The seven-day gate applies to the attribution reports modeling engine, including the Model Comparison tool. Google's best practice is still to keep upload latency under seven days so offline conversions appear in both views.
Why do Google Ads attribution reports show fewer conversions than the Campaigns page?
Google documents four structural differences. Attribution reports count by conversion time while the Campaigns page counts by the time of the ad click; attribution reports bypass offline conversions uploaded more than seven days after the event; the Model Comparison engine excludes conversions from the Search Partner Network, Gmail and App campaigns, which standard columns include; and Performance Max appears as a single Cross-Network row in attribution reports. Google describes these discrepancies as normal, not errors. Adding by conv. time columns to the Campaigns page aligns the timing convention but not the exclusions.
What is the 7-day offline conversion upload gate in Google Ads?
A processing rule documented on Google's About attribution reports page. The attribution reports modeling engine processes conversions that occurred within the last seven days, and an offline conversion uploaded more than seven days after the event is bypassed by attribution reports such as the Model Comparison tool. The rule surfaced publicly on August 22, 2026 via a LinkedIn post by Adriaan Dekker crediting Hana Kobzova. Google subsequently added wording clarifying that standard reporting columns and Smart Bidding are not subject to the gate. The page does not state whether the seven days are measured from the event date or the upload date.
Get the next brief in your inbox
When Google ships something that matters, you'll get the decode: what changed, whether it's signal or noise, and the move worth making this week. Free, no firehose, unsubscribe any time.