Google Ads attribution models and conversion windows

What google ads attribution models actually credit#

Every conversion the account reports has already been through a small, invisible decision: which of the touchpoints a customer met on the way to converting gets the credit, and how much of it. That decision is what google ads attribution models exist to make, and it happens automatically, per conversion action, before you ever open a report. This entry on google ads attribution models and conversion windows covers both halves of that decision in turn.

Data-driven attribution looks across every recorded interaction, clicks and video engagements alike, on Search, Shopping, YouTube, Display and Demand Gen ads inside the account, and it credits whichever of them actually moved the customer (Google Ads Help, 2026). That is the entire trick behind cross network conversions: one model, one account, reading interactions across several networks as a single connected trail rather than several separate stories.

The setting itself is chosen for each conversion action, so a single account can quietly run several google ads attribution models at once. The label lives inside conversion tracking, and Google's own guidance is direct about the mechanics: "the 'Attribution model' setting in conversion tracking lets you decide how to attribute conversions for each conversion action" (Google Ads Help, 2026). A lead form and a purchase in the same account can, and often should, run different models entirely.

Last click has not disappeared. Google confirms you can still switch to it, since it "is still supported" alongside data-driven attribution (Google Ads Help, 2026). What has gone is the older rules-based family: first click, linear, time decay and position-based models are, in Google's own words, "no longer supported by Google" for new selection, and any conversion action still carrying one was automatically upgraded to data-driven attribution (Google Ads Help, 2026). Google also runs a standing "Switch to DDA" recommendation that moves eligible conversion actions off last click and onto data-driven attribution with 30 days' notice to the account (Google Ads Help, 2026).

Read that table as a den, not a menu you revisit weekly. Pick deliberately per conversion action, and google ads attribution models will keep doing quiet, correct work in the background. For the wider anatomy of the platform, start at the folkfox Google Ads hub and burrow outward from there.

Conversion window settings: click, view and engaged#

An attribution model decides how credit is split. Conversion window settings decide how far back in time that credit is allowed to reach, and the two are easy to conflate because they sit on the same settings screen.

The click-through window is the generous one. It can be set anywhere from 1 day up to 90 days on Search and Display campaigns, depending on the conversion source (Google Ads Help, 2026). View-through and engaged-view windows are separate settings entirely, each capped at 30 days rather than 90: the view-through window covers someone who saw the ad, never clicked, and converted later, while the engaged-view window tracks conversions after a qualifying video engagement (Google Ads Help, 2026).

Track these three settings separately and the account behaves predictably. Muddle them, and a campaign can look starved of conversions simply because the window closed before a slow-moving customer finished their journey. Conversion window settings are patient by design; treat them the same way, and read folkfox's guide to conversion lag before assuming a quiet week is a real decline.

Why the click date decides the credit#

Here is the part that catches even careful practitioners. Every primary conversion column is calculated against the date of the click, not the date the conversion actually happened (Google Ads Help, 2026). A click from three weeks ago that converts today is written back into that click's original row, which is why conversions can be reported up to 90 days after the click depending on the window chosen, and recent days always look thinner than they will once the data matures (Google Ads Help, 2026).

Switch the attribution model on a conversion action and the effect lands exactly there. Google's own guidance to advertisers is blunt: "if you change your attribution model, you should update your bids and targets for each conversion included in the 'Conversions' column", because the change in conversion attribution can otherwise cause over-bidding or under-bidding (Google Ads Help, 2026). Under a fractional-credit model, conversion credit is distributed between contributing ad interactions according to the model you have selected, which is exactly why decimals appear in the Conversions column the moment an account moves off last click (Google Ads Help, 2026).

Nothing about the past is secretly rewritten. Google ads attribution models simply re-divide credit that was always keyed to the click date, so treat a model change the way you would treat any material change to conversion tracking: expect the numbers to move, and know why before you react.

Changing the model is a material change#

Smart Bidding is only as sensible as the conversion signal it is fed, and an attribution model change alters that signal directly. Independent practitioner analysis of the Smart Bidding learning phase groups attribution model changes alongside the other conversion-related changes that can trigger a fresh learning status: "this also applies to changes in conversion value, attribution model, or count settings" (independent practitioner analysis, 2026). Treat a switch between google ads attribution models with the same caution you would give to swapping bid strategies, not as a cosmetic reporting tweak.

Google's own change guidance stops short of promising a learning reset, but it does confirm the reporting consequence that makes one likely: expect changes to campaign-level reporting with any attribution model change, and update bids and targets accordingly rather than leaving Smart Bidding chasing a target built on the old numbers (Google Ads Help, 2026). Calculate the percentage shift in cost per conversion between the old and new model, adjust the target by roughly that amount, and give the campaign a full cycle before judging it.

The practical routine: run the attribution reports before switching, not after, so you know roughly how much the numbers will move. Google ads attribution models reward patience and punish panic in almost exactly the same way a bid strategy does. Our PPC service builds this check into every account audit.

What google ads attribution models cannot see#

Every one of the google ads attribution models available inside the platform, however cleverly it distributes credit, is still reading the same limited evidence: clicks, views and engagements that Google itself recorded. That is a measurement of correlation, not causation, and independent incrementality research keeps finding the gap between the two is large.

Branded search is where this bites hardest. A customer who already searches your brand name was often coming anyway, and an in-platform model has no way to test that counterfactual, so it quietly over-credits the cheapest, most convenient click on the path. Only a genuine holdout or geo-experiment can prise correlation and causation apart.

Three further blind spots sit outside what any attribution report can show. A view-through conversion, where someone sees but never interacts with an ad before converting, is only counted at all inside its own short window and reported separately from the primary Conversions column, so unclicked exposure beyond that window is invisible (Google Ads Help, 2026).

Cross-device journeys without a Google sign-in cannot be directly observed either: Google states plainly that "cross-device conversions are modeled to account for users who start their journey on one device with an ad interaction" and complete it on another, which means privacy-safe modelling stands in for a signal nobody actually measured (Google Ads Help, 2026).

And an offline sale, a signed contract or a phone call that closes weeks later stays invisible until someone deliberately imports it back with its click ID, since offline conversion imports exist precisely because "an ad doesn't lead directly to an online sale, but instead starts a customer down a path that ultimately leads to a sale in the offline world" (Google Ads Help, 2026).

None of this makes google ads attribution models useless. It makes them one instrument among several, precise about what they can see and honestly blind about the rest. A cunning practitioner reads the reports and still books a holdout test before betting real budget on branded search.

How to choose and test google ads attribution models#

Choosing well does not need a statistics degree. It needs a short, repeatable routine, run before a switch rather than after the numbers have already moved.

Run that routine once per conversion action and google ads attribution models stop being a mystery dial and start being a setting you chose on purpose. For a second pair of eyes across the whole measurement stack, from conversion action to conversion window settings, talk to folkfox.

Frequently asked questions#