A processing rule documented in the Google Ads Help Center means offline conversions arriving more than seven days after the event never reach the data-driven attribution model, even though standard reporting columns record them in full.

The rule sits inside the Help Center page covering attribution reports, under a heading dealing with why those reports disagree with standard Google Ads reporting columns. It was spotted by Hana Kobzova and carried into wider circulation by Adriaan Dekker, a Google Ads consultant who writes about scaling companies on the platform, in a LinkedIn post carrying a 17-hour timestamp that places it on August 22, 2026. Dekker credited Kobzova with sharing the find. His summary of it was blunt: conversions uploaded after the window still appear in standard reports, but they are "ignored by the attribution engine".

That single sentence describes a gap between two sets of numbers that sit in the same account, drawn from the same events, and are frequently compared against one another during performance reviews.

An upload deadline only one report enforces

Google frames the discrepancies as expected rather than faulty. According to the documentation, differences in conversion totals between attribution reports, including the Model Comparison tool, and standard Google Ads reporting columns are normal, and are caused by data processing constraints and network configurations inside the attribution system.

The mechanism is a rolling window. The daily attribution modeling engine, identified in the text as Search funnels, processes conversions that occurred within the last seven days. An offline conversion uploaded more than seven days after the initial event took place is, according to Google, "bypassed by the attribution modeling calculations". Standard reports operate without that restriction and retroactively record uploaded offline conversions regardless of when they arrive.

The consequence is arithmetic. A lead that clicks a search ad on the first of the month, closes on the twelfth, and is uploaded on the fifteenth appears as a conversion in the Campaigns page columns. In the Model Comparison report, it does not exist. Neither figure is wrong. They are counting under different rules.

Google's stated remedy in the same passage is a latency target rather than a fix: keeping offline conversion upload latency below seven days, the documentation says, allows the modeling engine to incorporate those offline journeys into daily calculation suites, maintaining model accuracy and supporting automated bidding efficiency.

The wording carries an internal inconsistency

Two statements in the same bullet describe the gate differently, and they are not equivalent.

The first keys the rule to the age of the conversion: the attribution engine only processes conversions that occurred within the last seven days. The second keys it to upload latency: an offline conversion uploaded more than seven days after the initial event is bypassed.

Those tests diverge in ordinary cases. A conversion that occurred six days ago and is uploaded now satisfies both. A conversion that occurred fourteen days ago but was uploaded two days after the event satisfies the latency test while falling outside a strict seven-day event window. Whether the engine looks at the event date, the upload delay, or both, the documentation does not say. For advertisers running long sales cycles, the distinction determines whether historical backfills carry any modeling weight at all.

What the modeling engine leaves out

The upload gate is not the only exclusion. Under a heading covering network and campaign type exclusions, the documentation states that the Model Comparison engine is designed to explicitly exclude conversions originating from the Search Partner Network, Gmail, and App campaigns. Standard account reporting columns include conversions from those networks by default.

The comparison table published alongside that passage sets out the full divergence across six dimensions.

Time of event differs first. Attribution reports count by time of conversion. The Campaigns page counts by the time of the ad query preceding the click that led to the conversion. Adding by conversion time columns to the Campaigns page aligns the two.

Network coverage in attribution reports spans the Search Network, YouTube including Google video partners, the Google Display Network, and Discover. Performance Max campaigns are labelled as Cross-Network in that view and Demand Gen campaigns as Google owned channels. The Campaigns page adds search partners, Gmail, Google Maps and App.

Campaign coverage in attribution reports takes in Search, Shopping, Video, Display excluding pay for conversions, and Demand Gen. App campaigns are not supported. The Campaigns page adds App and Hotel.

Video format coverage is narrower on the attribution side, covering ad sequence, bumper, non-skippable in-stream and skippable in-stream. The Campaigns page additionally reports masthead, outstream and in-feed video.

Conversion coverage overlaps more closely. Attribution reports include Google Ads conversion tracking, Google Analytics goal and conversion import including Android app conversions, offline conversion import, call conversions covering click to call, call import and website call conversions, and app conversions. The Campaigns page carries all of those plus store visits. A footnote adds that engaged view conversions are supported in both Google Ads and Google Analytics 4 properties, while conversions attributed to impressions are not currently supported in GA4.

Four independent sources of divergence, then, before any question of model choice arises. An advertiser reconciling a Model Comparison figure against a Campaigns page figure is comparing outputs that differ by timing convention, network inclusion, campaign type inclusion and upload latency simultaneously.

Controls that bound the reports

The documentation also details the report controls governing what the attribution tab displays.

Date ranges reach back two years. Path report and attribution credit data older than two years is deleted and cannot be restored, a retention limit that sits well below the broader Google Ads reporting history.

The lookback window control adjusts to 30, 60 or 90 days, with most reports opening at 30. The Model Comparison report carries one additional option: a Default setting that matches the report's lookback window to the conversion window configured for each conversion action.

Google draws a firm line between that control and the conversion window, which are frequently conflated. A lookback window determines how far back from a conversion an ad interaction remains eligible for credit; the documentation's own example describes a 30-day lookback considering interactions from December 31 to January 30 for a conversion recorded on January 30. A conversion window is the period after an ad interaction during which a conversion is recorded at all. Any conversion without a preceding ad interaction inside its conversion window is not counted in Google Ads.

One further caveat governs the Model Comparison report. Accounts that have only recently become eligible for cross-network attribution may see a notice that historical data is partially available, because the report filters out networks or campaigns without sufficient data.

Five reports, one tab

The attribution section groups five reports, reached through the Goals icon, the Measurements drop-down and the Attribution entry.

The Overview report reports how many days and ad interactions preceded conversion, how many conversions involved ad interactions on more than one device, and which campaigns most often assist.

The Conversion paths report shows common paths in two views. The default paths view consolidates successive ad interactions carrying the same dimension value into a single touchpoint annotated with a repeat count. Transition pathscollapse them into one interaction entirely, which suits analysis of movement between keywords, ad groups or campaigns. A Devices dimension breaks paths down by desktop, tablet and mobile, with the documentation noting that two adjacent path elements represent different devices but not necessarily different device types. A toggle adds YouTube impressions to the view; those impressions come from YouTube and Demand Gen campaigns.

The Path metrics report measures elapsed time and interaction counts. Average days to conversion is the default view, with a measure-from control switching between the first and the last ad interaction and a measure-in control switching between days and hours. An average interactions to conversion tab reports frequency. The documentation flags a scope limit that shortens every path shown: the report reflects only the keywords and ads inside the Google Ads account.

The Assisted conversions report distinguishes two credit types. An ad interaction that contributed to a conversion without immediately preceding it registers a click and view assist. An interaction directly before the conversion registers a last click conversion. The documentation works through a path reading YouTube, Search, YouTube, YouTube, in which YouTube collects two assists and one last click conversion. Because a single conversion can carry multiple assisting interactions, assist counts routinely exceed total conversions.

The Model Comparison report handles the models themselves. Advertisers managing multiple accounts through a manager account can run attribution reports at the manager level once cross-account conversion tracking is configured.

Why a seven-day gate matters for bidding

Data-driven attribution is not only a reporting view. Its output feeds Smart Bidding, and the documentation says as much when it links upload latency to model accuracy and automated bidding efficiency in the same sentence.

Offline conversion import exists because the conversion is separated in time from the click. PPC Land documented that mechanism in July 2026, when Google published a short explainer on the Google Click Identifier: the parameter is captured at click time, stored alongside the business record, and uploaded once the outcome is known, which may be days or weeks later. In business-to-business pipelines, a fortnight between click and closed deal is unremarkable. Under the documented gate, those conversions reach standard reporting and bypass the model.

The rule lands in the middle of a sustained restructuring of how offline data enters Google's advertising stack. Enhanced conversions for leads arrived in March 2022 as a way to supplement imported offline data with hashed customer identifiers. A mandatory conversion environment parameter for offline uploads followed in February 2025, justified at the time as an attribution accuracy measure. The Data Manager API became the unified ingestion point across Google Ads, DV360 and Analytics, and Google blocked new offline conversion import adopters from the older Ads API path on June 15, 2026. Enhanced conversions for web and leads collapsed into a single toggle over the same period.

Every one of those changes addressed how offline data gets in. None addressed what the attribution model does with it once inside.

The Search Partner Network exclusion carries its own history. Transparency on that inventory has expanded in stages, from full placement reporting in August 2025 through the removal of parked domain ads in February 2026 and an alpha letting selected Performance Max advertisers deselect the network in mid-2026. Conversions from that inventory nonetheless remain outside the Model Comparison engine while appearing in standard columns, which places any partner network return-on-spend assessment drawn from attribution reports on incomplete ground.

Documentation as the disclosure channel

The pattern by which this surfaced is familiar. Google publishes or amends a Help Center page without a release note, a practitioner notices, and a LinkedIn post carries it into wider circulation. Dekker has occupied that role repeatedly, from consent mode enforcement in July 2025 through the Performance Max partner network alpha in May 2026.

For measurement teams, the operational reading is narrow and specific. Attribution reports and standard columns were never intended to reconcile, and the documentation now names four structural reasons why. Anyone treating Model Comparison output as a complete account of conversions, or as a basis for shifting budget between networks, is working from a filtered view whose filters have only recently been written down.

Timeline

Summary

Who: Google, through its Google Ads Help Center documentation on attribution reports. The passage was surfaced publicly by Adriaan Dekker, a Google Ads consultant, who credited Hana Kobzova with sharing it. The audience affected consists of advertisers, agencies and measurement teams using offline conversion import alongside data-driven attribution.

What: Documentation stating that the daily attribution modeling engine, identified as Search funnels, only processes conversions occurring within the last seven days, and that an offline conversion uploaded more than seven days after the initial event is bypassed by attribution modeling calculations. Standard reports carry no such restriction. A second exclusion removes conversions originating from the Search Partner Network, Gmail and App campaigns from the Model Comparison engine, while standard reporting columns include them by default.

When: The LinkedIn post carrying the documentation excerpt was published on August 22, 2026. The Help Center page itself carries no publication date for the added section.

Where: The Google Ads Help Center article covering attribution reports, applying to Google Ads accounts globally, including manager accounts running cross-account conversion tracking.

Why: Data-driven attribution output feeds Smart Bidding, and the documentation ties upload latency directly to model accuracy and automated bidding efficiency. Offline conversion import exists to capture conversions separated in time from the ad click, a separation that routinely exceeds seven days in business-to-business and considered-purchase categories. Where it does, the conversion reaches standard reporting columns and never reaches the model, producing two divergent totals inside one account from a single set of events.