Flight log / Transmission 057

Data-driven attribution in Google Ads: what it credits and what it can't see

Data-driven attribution is a very fair referee that only watches one team. Google recommends at least 200 conversions and 2,000 ad interactions in 30 days for it to work best, and plenty of B2B demo actions never get close.

Data-driven attribution, DDA for short, is how Google Ads decides which of your keywords, ads and campaigns get credit when someone converts. It has been the default for most conversion actions for a while, which means most accounts are using it whether or not anyone chose it. That's fine. What's not fine is reading its output as the full story of how a B2B buyer found you.

This guide covers how DDA assigns credit, the models still available, the data it needs, how lookback windows shape what it sees, the touchpoints it can't see at all, and what changes when you switch. For the bigger multi-channel picture, see our guide to cross-channel attribution. This post stays inside Google Ads.

What is data-driven attribution in Google Ads?

Google Ads Help describes data-driven attribution as a model that gives credit for conversions based on how people engage with your ads before converting, using data from your own account. It compares the paths of people who convert with the paths of people who don't, and gives more credit to interactions that are associated with a higher likelihood of converting. Each model is specific to the advertiser, so your DDA is not your competitor's DDA.

What it looks at, per Google: ad interactions on Search (including Shopping), YouTube, Display and Demand Gen, including clicks and video engagements. It applies to website conversions, store visit conversions and Google Analytics conversions. Smart Bidding strategies can use its results to bid.

One quiet line on the same page is worth knowing: last click and DDA can produce the same results in some situations, depending on how much data is available. If your account runs one campaign and buyers click one ad before converting, there isn't much path to distribute credit across.

Which attribution models are left in Google Ads?

Two. In April 2023 Google announced that first click, linear, time decay and position-based models were going away in Google Ads and Google Analytics 4. From June 2023 you could no longer select them for conversion actions that didn't already use them, and in September 2023 conversion actions still on them were switched to data-driven. Google said at the time that fewer than 3% of Google Ads web conversions used those models.

So the real choice is DDA or last click. You change it per conversion action under Goals, Summary, the conversion action, then Edit settings and Attribution model. Goals, Attribution, Model comparison shows the two side by side.

What are the data requirements for data-driven attribution?

Google's documentation is a little inconsistent here, so read it closely. The About data-driven attribution page says all conversion actions are eligible regardless of conversion or interaction volume, and recommends at least 200 conversions and 2,000 ad interactions on supported networks within 30 days for best performance. Below that, the model still works but may assign credit less precisely.

The help page for the "Switch to DDA" tab says something stricter for some conversion action types: 3,000 ad interactions and 300 conversions over 30 days to qualify, and 2,000 interactions and 200 conversions every 30 days to stay eligible. The tab includes a DDA eligibility column, so check that column rather than guessing which rule applies to your action. Google also notes that some actions may be selected for an automatic switch, with an email to account admins 30 days ahead.

For a B2B account spending $5K to $100K a month, the practical read is this. A demo request action with 40 conversions a month is eligible, but DDA has very little path data to learn from, so its credit split is closer to an educated guess. A higher-volume action like a content download gives it more to work with.

How does DDA change credit compared with last click?

Last click gives everything to the final ad clicked. In B2B, that's often a brand search, because buyers who first found you on a non-brand term come back later by typing your name. DDA usually moves some credit up the path. Here's that shift in a hypothetical account, using Google Ads' model comparison view on the same 100 conversions.

In this illustrative example, brand search drops from 62 credited conversions to 44 and non-brand rises from 28 to 36. Both columns add to 100, because DDA doesn't create conversions, it redistributes them. If your brand campaign looks less efficient after a switch, that's not a performance change, it's a bookkeeping change. Our brand bidding guide covers the separate question of whether brand clicks would have happened anyway.

DDA can also assign fractional credit, so a keyword can show 0.4 conversions. That's normal. It can also make low-volume keywords look like they convert at odd rates; judge those at the ad group or campaign level instead.

What lookback windows does data-driven attribution use?

DDA can only credit interactions inside the conversion window you set on each conversion action. Google's conversion windows page lists the defaults and ranges:

• Click-through window: default 30 days, and up to 90 days depending on the conversion source. Google recommends at least 7 days.
• Engaged-view window (video): default 3 days, 1 to 30 days.
• View-through window: default 1 day, 1 to 30 days.

Changes apply only to conversions going forward, so conversions that fell outside an old window aren't counted retroactively. For a lead form, 30 days is usually plenty. For a conversion action you import from your CRM, like an opportunity or a closed deal, it often isn't. Here's an illustrative curve for a hypothetical B2B account: the share of each conversion type, out of everything that happens within 90 days, that has happened by a given number of days after the first ad click.

In this hypothetical account the default 30-day window catches 92% of form fills but only 41% of imported opportunities. DDA would be building its model of what drives pipeline from fewer than half of the opportunities. If you import offline stages, set the window on those actions to cover your real sales cycle, up to the 90-day maximum, and pair them with enhanced conversions for leads so more of them match back to a click.

What can't data-driven attribution see?

DDA credits Google ad interactions. Everything else in a B2B deal is outside the frame. Here's an illustrative breakdown of touches in one hypothetical closed deal, by channel.

In this illustrative deal, the touches Google Ads DDA can credit are the non-brand and brand search ad clicks and any YouTube engagement, about 30% of the journey at most. LinkedIn ads, the webinar, sales calls, review sites and organic visits aren't Google ad interactions, so DDA spreads credit only across the Google slice. That slice will always add up to 100% of the conversion, which is exactly why it can look like the whole story.

The specific blind spots:

• Other ad platforms. LinkedIn, Microsoft Ads, Meta and the rest never enter Google's model.
• Offline and sales touches. Events, calls, emails, referrals.
• What happens after the conversion, unless you import it. A form fill from a student and one from a VP count the same.
• Interactions outside the window you set on the conversion action.
• Causation. DDA finds interactions associated with converting. It doesn't prove the conversion wouldn't have happened without them. That's what incrementality testing is for.

What happens when you switch to data-driven attribution?

Per Google Ads Help, changing a model affects only that conversion action and only going forward. It also changes what bid strategies using the Conversions column optimize toward, including Target CPA and Target ROAS. A few habits keep a switch from turning into a fire drill:

1. Compare first. Run Model comparison for the action over a recent date range. Google notes it filters out networks or campaigns without enough data.
2. Switch primary actions deliberately. Only primary conversions feed the Conversions column that bidding uses.
3. Annotate the date. Brand campaigns may look worse and upper-funnel campaigns better the next week. Nobody's performance changed.
4. Revisit CPA targets. If credit moves toward non-brand, its reported CPA falls and its tCPA target may now be too loose or too tight.
5. Hold off on other big changes for a couple of weeks so you can tell what the model change did.

Where data-driven attribution advice breaks down

• Low volume. Below Google's recommended 200 conversions and 2,000 interactions in 30 days, the model works with less data and may be less precise. Many B2B bottom-funnel actions live there.
• It's a black box. Google publishes a methodology document but not your model's weights. You can see outputs, not reasons.
• Last click isn't a safer truth. It's simpler, not more accurate. It overcredits brand and the final touch.
• Good inputs matter more than the model. If the conversion action counts junk leads, DDA will faithfully credit the keywords that produce junk.

FAQ