Attribution Models and Conversion Windows — why your reported conversions shift without any customer behavior changing

Attribution models decide which ad interactions get credit for a conversion — they change *how conversions are reported per campaign*, never *how many conversions happened in total*. Misreading a post-switch reporting shift as a real performance change is one of the most common ways an account gets re-optimized against noise. This page covers what data-driven attribution (DDA) actually does, the conversion-window settings that gate when a conversion counts at all, and the counting caveats (cross-device, cross-account, historical comparisons) that distort what you see in reports.

Data-Driven Attribution Is the Default, and the Only Alternative to Last-Click

Google now supports only two attribution models: Last Click and Data-Driven Attribution (DDA). First-click, linear, time-decay, and position-based models were deprecated and automatically converted to DDA in October 2023 (support.google.com/google-ads/answer/6259715). DDA has been the default for all new conversion actions since October 2021 (support.google.com/google-ads/answer/6394265).

- Last-click: 100% of the credit goes to the final ad interaction before conversion. - DDA: uses machine learning on account-specific conversion data, comparing the paths of customers who converted against those who didn't, and redistributes credit across the contributing interactions (support.google.com/google-ads/answer/6259715, support.google.com/google-ads/answer/9203352 — "fractional credit"). - DDA covers Search (including Shopping), YouTube, Display, and Demand Gen ads, including website conversions, store visits, and Google Analytics data (support.google.com/google-ads/answer/6394265). - Data requirement: Google recommends at least 200 conversions and 2,000 ad interactions within a 30-day period in supported networks for DDA to perform optimally, though it can function below that threshold (support.google.com/google-ads/answer/6394265). - Paired with automated bidding, Google states DDA improves prediction accuracy of each ad's incremental impact, which can produce "additional conversions at the same cost-per-acquisition" (blog.google/products/ads-commerce/data-driven-attribution-new-default/).

Conversion Windows: Click-Through, View-Through, Engaged-View

A conversion window is the time after an ad interaction during which a conversion is still recorded (support.google.com/google-ads/answer/3123169). Three windows exist and each has its own default and range:

- Click-through window: counts conversions after a click. Default 30 days, customizable from 1 to 90 days depending on the conversion source. - View-through window: counts conversions after an impression with no click. Default 1 day, customizable from 1 to 30 days, or 1 to 4 weeks depending on the setting. - Engaged-view window (video ads only): counts conversions after a meaningful engagement with video content — not just a view. Default 3 days for web conversions, customizable up to 30 days (support.google.com/google-ads/answer/10048752). "Meaningful engagement" is format-specific: for YouTube skippable in-stream ads it means viewing at least 10 seconds (or the full ad if shorter); for in-feed and Shorts ads, 5 seconds of viewing qualifies. - Google recommends windows of at least 7 days to capture a richer set of conversion data (support.google.com/google-ads/answer/3123169).

Conversion windows are a prerequisite this page assumes correctly set — see Conversion Tracking Fundamentals — the layer every audit and every bid strategy depends on for how conversion sources and tag placement determine what gets counted at all before a window or attribution model can even apply.

The single most important operational fact for auditing reports

when you change the attribution model, the total number of conversions stays the same; DDA (or last-click) only redistributes credit for those same conversions differently across campaigns, ad groups, and keywords (support.google.com/google-ads/answer/10762625). Some campaigns will show more conversions, others fewer, purely because credit moved — not because actual customer conversions changed (support.google.com/google-ads/answer/9203352).

- The change is forward-looking only: it affects how conversions are reported in the "Conversions" and "All conversions" columns going forward. Historical conversions recorded before the switch continue to be reported under the previous model (support.google.com/google-ads/answer/9203352). - To see how historical data *would* look under a different model without actually switching, use the Model comparison report: add "current model" columns from the "Attribution" section of the Columns menu (support.google.com/google-ads/answer/9203352). - Bid targets must be adjusted after switching. Since per-campaign conversion counts shift, CPA/ROAS bid targets calibrated under the old model no longer match reality: calculate the percentage change in cost per conversion and adjust targets by that same percentage. Failing to do this "may result in over- or under-bidding" as perceived conversion value per campaign shifts (support.google.com/google-ads/answer/9203352). This is the direct link to Bidding Strategies: Manual, Smart Bidding, and When Each Applies — matching the strategy to the conversion data you actually have: an attribution switch without a corresponding bid-target recalibration silently breaks Target CPA/ROAS performance.

Cross-Device and Cross-Account Conversion Counting Caveats

- Cross-device conversions happen when a user clicks an ad on one device and converts on another. Google measures these with privacy-safe aggregated and anonymous data from users signed into Google services (support.google.com/google-ads/answer/6270625). They are reported in a dedicated "Cross-device conversions" column, separate from single-device totals. Google does not publish specific confidence or accuracy thresholds for when cross-device data gets reported. - Cross-account conversions require multiple Google Ads accounts linked under one manager account (MCC) and let you track multi-device user interactions across those accounts (support.google.com/google-ads/answer/3061730). - Mutual exclusivity: an account cannot use account-specific conversion actions and cross-account conversion actions simultaneously — only one or the other. If an account belongs to multiple managers, it can only draw conversion actions from one of them (support.google.com/google-ads/answer/3061730). - Migration trap: switching an account to cross-account tracking makes campaigns that targeted specific conversion actions fall back to the manager account's *default* conversion goals, which can silently change what a campaign is optimizing toward (support.google.com/google-ads/answer/3061730).

What Does NOT Work

Comparing pre-switch and post-switch conversion numbers directly to judge performance. The apparent change reflects how credit is now distributed, not a change in customer behavior (support.google.com/google-ads/answer/9203352). Two failure modes compound this:

- Time-lag illusion: right after switching models, recent days can show a temporary, slight dip in conversions purely from reporting delay — DDA credits interactions that occurred at different points in time, so the most recent days are undercounted until data catches up (support.google.com/google-ads/answer/9203352). Fix: exclude the most recent 2-3 weeks of data when evaluating the impact of an attribution model change. - Incomplete historical baselines: the Model comparison report may show only partial historical data if the account only recently became eligible for cross-network attribution (support.google.com/google-ads/answer/9203352) — a gap in the comparison table is missing eligibility, not zero conversions.

This is exactly the kind of misread that leads to overreacting during How to Audit an Existing Google Ads Account (Priority Order) — the sequence that stops you from trusting broken data: a campaign that "lost" conversions after a model switch may have lost nothing but reporting credit, and cutting its budget on that basis burns money on a phantom problem.

Related

- Conversion Tracking Fundamentals — the layer every audit and every bid strategy depends on — attribution only redistributes credit for conversions that were captured correctly in the first place; broken or duplicated tracking upstream makes any attribution model's output meaningless. - Bidding Strategies: Manual, Smart Bidding, and When Each Applies — matching the strategy to the conversion data you actually have — Target CPA/ROAS bid targets must be recalibrated after an attribution model switch, or the algorithm bids against a conversion value that no longer matches reported reality. - How to Audit an Existing Google Ads Account (Priority Order) — the sequence that stops you from trusting broken data — when auditing, a per-campaign conversion swing right after an attribution model change is a reporting artifact to identify and exclude, not a performance signal to act on.