How to use Google Ads data-driven attribution, step by step

To use data-driven attribution in Google Ads, check each purchase action's model and eligibility on the Switch to DDA tab, then switch it there. Add the current model columns. If you bid with Target CPA or Target ROAS, move each target by the same percentage, and judge results after your average days to conversion.

Chances are you already use it. Google made data-driven the default for most conversion actions, and it moved actions on its retired models over to it. So this walkthrough starts with finding out, then spends its time on the part that costs money: bid targets.

Keep Google Ads open on the Goals menu. Everything below sits under Attribution or Conversions there, apart from the columns in steps 7 and 8.

Step by step

1. Open the Switch to DDA tab. It is where Google lets you move actions from a rules-based model, such as last click, to data-driven. It only includes actions that attribution in Google Ads supports. Menu path: Goals > Attribution > Switch to DDA tab.
2. Read the eligibility column. It is called Data-driven attribution (DDA) eligibility. Google's main help page says every action is eligible whatever its volume, while this tab's page sets volume bars for some action types. When they disagree, the column for your account wins. Menu path: Switch to DDA tab > Data-driven attribution (DDA) eligibility column.
3. Look for a scheduled switch. Google can pick actions for an automatic switch, and it emails admins 30 days ahead. You can let it happen, opt out, or switch sooner. Menu path: Goals > Attribution > Switch to DDA tab, then the action's row.
4. Note your average days to conversion. This number is your minimum wait after the switch. Google suggests letting that many days pass before you judge the results. Menu path: Goals > Attribution > Path metrics > Avg. days to conversion.
5. Preview the move by campaign. Model comparison sets last click and data-driven side by side, with Cost / conv. and Conv. value / cost for each. It leaves out conversions from Search partners, Gmail and App campaigns, and offline conversions uploaded more than 7 days late. Menu path: Goals > Attribution > Model comparison > Dimension: Campaigns > Compare and With drop-downs.
6. Make the switch. Click Switch to DDA in the action's row, or tick several rows and pick Switch attribution model to DDA from the Edit drop-down. Read the confirmation pop-up, then click Done. If a manager account runs your conversion tracking, change the model there. Menu path: Goals > Attribution > Switch to DDA. Settings route: Goals > Conversions > Summary > the action > Edit settings > Attribution model > Data-driven > Save > Done.
7. Add the current model columns. They show past conversions as if the new model had always applied. Add Conversions (current model), Cost / conv. (current model), Conv. value (current model) and Conv. value / cost (current model). Menu path: Campaigns page > Columns menu > Attribution section.
8. Reset Target CPA or Target ROAS. Work out how far Cost / conv. or Conv. value / cost moved under the new model, campaign by campaign. Then move each campaign's target by the same percentage, using a date range that skips the most recent few weeks. Menu path: Campaigns page, with the regular and current model columns side by side.
9. Wait before you judge. Expect decimals in Conversions, a dip in the most recent days, and credit sliding between campaigns. Google reports conversions on the date of the ad interaction, so recent days fill in late. Menu path: Goals > Attribution > Path metrics, for the number of days to wait.

A worked example

For illustration, take two Search campaigns that each spent €1,000 last month. Every number in this table is invented and round.

In this worked example, the switch moved €600 of credit from brand to generic. The bottom row is the point: the total stays at €8,000 in this worked example, so not one extra sale happened.

Now hold that against a margin. One store's Break-even sheet does the arithmetic for a 40% margin: break-even ROAS is 2.5x, which is 1 divided by 0.40.

At that margin, generic search looked like a loser under last click and a winner under data-driven. For illustration, its credited return went from 2.0 to 2.6, while the line it had to clear stayed at 2.5. The campaign crossed break-even on paper only.

Now the bid target. For illustration, say generic search ran on a Target ROAS of 250%. In this worked example, its Conv. value / cost rose by 30%, so Google's method raises the target by 30% as well, to 325%.

If you leave the target at 250%, the bidding sees a campaign beating its goal and pushes harder. It does so on credit that used to belong to brand. That is the overbidding Google's own guidance warns about.

Brand moves the other way. In this worked example its return fell by 10%, from 6.0 to 5.4, so its target comes down by 10% too.

None of this tells you whether generic search causes sales. The model moved credit between two campaigns that both sat on the same paths. Answering the causal question takes a holdout test, with a group that never sees the ads.

What to check when the report looks wrong

• Your purchase action is missing from the Switch to DDA tab. It may already be on data-driven, or attribution in Google Ads may not support it. Open its Edit settings to see which model it uses.
• The eligibility column and Google's main page disagree. They do, on paper. Go by the column in your own account.
• The last few days look weak after the switch. Conversions are reported on the date of the ad interaction, and credit now spreads over earlier clicks. Give it your average days to conversion before you call it a drop.
• Decimals appeared in Conversions. That is fractional credit: one sale shared between several ad interactions. Google says it shows up the first time you leave last click.
• Spend jumped on upper-funnel campaigns. Their credited CPA or ROAS improved after the switch, while their targets stayed put. Go back to step 8.
• Model comparison shows fewer conversions than the Campaigns page. It skips Search partners, Gmail, App campaigns and late offline uploads. Use the Campaigns page with the current model columns for the full count.
• A client account will not let you change the model. With cross-account conversion tracking, the model is chosen in the manager account.

What to do this week

1. Screenshot the Switch to DDA tab. Pass: for every purchase action, you can say which model it uses and whether it is eligible. Fail: an action you count as a sale is missing, or ineligible with no plan.
2. Add the four current model columns to the Campaigns page. Pass: you can see, per campaign, how far Conv. value / cost moves under data-driven. Fail: the columns match the regular ones everywhere, so the action is still on last click or has nothing to split.
3. Recalculate one target and book the review. Pick the campaign that moved most, apply the same-percentage method, and put a review in the calendar once your average days to conversion have passed. Pass: the target moved by the same percentage as its return. Fail: no target has changed since the last model change.

Check the homework. Your GA4 Attribution paths export already holds the evidence. Causality Engine reads that one file and shows what each channel caused next to what last-click gave it, in 1 to 2 minutes, for €99 once (excluding VAT), refundable within 30 days. Check the homework

Sources, 1 October 2026: About "Switch to DDA" (data-driven attribution) (Google Ads Help); About data-driven attribution (Google Ads Help); About attribution models (Google Ads Help); Best practices for managing attribution model changes (Google Ads Help); About attribution reports (Google Ads Help).

Related answers

• All Google Ads questions
• What is Google Ads data-driven attribution?
• How to check Google Ads attribution, step by step
• Data-driven attribution, defined
• Incrementality testing, explained