Google Ads Attribution: Check the Columns Before You Change Bids

Data-driven attribution is the practical default for most conversion actions in Google Ads, and it gives fractional credit across every interaction that led to a sale, not just the last click. The moment you check your reports under this model, expect decimals in your Conversions column, a time lag on the most recent days, and credit shifts that will push you to update your bid or target settings.

Key Takeaways

Data-driven attribution is now the default in Google Ads because it better reflects which touchpoints genuinely influence conversions, shifting the focus from last-click models.

Table of Contents

• What attribution means and why it changes your reported numbers
• Which attribution models Google Ads supports now
• How the attribution reports actually work
• How to run the Model comparison report before switching
• Updating Target CPA and Target ROAS after a model change
• A decision checklist before you switch models
• How we operationalise attribution inside managed campaigns
• Where attribution fits in the bigger funnel picture
• What attribution insights actually change in a live account
• Where attribution data is genuinely limited
• How to troubleshoot attribution numbers that do not add up
• Using attribution to rebalance budget across channels
• The judgement call most accounts get wrong
• Let us run the attribution audit on your account
• FAQ
• Sources

What attribution means and why it changes your reported numbers

Attribution is the method Google Ads uses to decide which ad interactions get credit for a conversion. Pick a different model and the same conversion data gets redistributed differently across your campaigns, which means your Conversions and All conv. columns can tell two different stories from the same clicks.

This matters because data-driven attribution assigns fractional credit based on patterns Google’s models find in your account’s converting and non-converting paths, while last click gives the full conversion to whichever ad or keyword the customer clicked right before converting. A brand campaign that used to look like your best performer under last click can suddenly look average once data-driven attribution spreads credit back to the generic and display touchpoints that actually built the path.

The setting lives at the conversion action level, not the account level. That is the detail most people miss. If a bid strategy, like Target CPA or Target ROAS, uses that conversion action to optimise, changing its attribution model changes the signal the bidding algorithm is reading, and it will adjust its own behaviour in response, whether you planned for that or not.

Which attribution models Google Ads supports now

Google Ads has narrowed the field. Data-driven attribution is now the default model for most conversion actions, last click is still available as an option, and several older rules-based models (first click, linear, time decay, position-based) have been phased out.

Here is how the two remaining options actually behave:

• Data-driven attribution uses machine learning on your account’s own conversion and non-conversion paths to work out which touchpoints genuinely moved the needle, and it draws on data from Search, YouTube, Display and Demand Gen.
• Last click gives 100% of the credit to the final ad interaction before conversion, with no modelling involved.

Data-driven attribution performs best with volume behind it. Google’s own guidance points to roughly 200 conversions and 2,000 interactions within a rolling 30-day window as the level where the model works well. Below that, the model still runs, but you are asking it to find patterns in a thin data set.

Last click still earns its place for low-volume accounts, simple single-channel setups, or manual bidding where you want full transparency over exactly which click you are paying for, with no modelling layer in between.

How the attribution reports actually work

The attribution reports are where the multi-touch picture lives, and they tell a different story to your standard Campaigns page. Attribution reporting covers four views: Overview summarises top paths and assisted interactions, Conversion paths shows the exact sequence of touchpoints before each conversion, Path metrics breaks down average path length and days to conversion, and Assisted conversions shows which campaigns show up early in the journey but rarely get the final click.

The lookback window and the conversion window are not the same control, and mixing them up skews your reads. The conversion window sets how long after a click Google Ads will count a conversion at all. The lookback window sets how far back the attribution model is allowed to look for prior interactions when deciding how to split credit. Shorten the lookback window on a long-consideration product, say from 90 days to 30, and you will see fewer assisted conversions simply because the model is now blind to anything that happened more than a month before the sale.

Attribution reports do not pull from the same inventory as your Campaigns page. The Model Comparison engine excludes certain channels by design, including the Search Partner Network, Gmail and App campaigns, so a conversion count in attribution reporting can sit lower than the number you see in standard reporting for the same date range. That gap is not a tracking fault. It is a scope difference, and it catches out a lot of people checking both screens side by side for the first time.

How to run the Model comparison report before switching

Before you touch a live setting, use the Model comparison report to see what would change. It runs your historical conversions through an alternate model without altering anything in production.

1. Open the Model comparison report and choose the dimension you want to analyse, such as campaign or keyword.
2. Set Compare to your current model and With models to the alternative you are considering, then add the Conversions (current model) and Cost / conv. (current model) columns to the table.
3. Read the decimals. A row showing 14.6 conversions under data-driven attribution means fractional credit has been split across several touchpoints that contributed to those 14 or 15 actual sales, not a counting error.
4. Exclude the most recent three to seven days from your comparison. Conversions attributed to those days are still being modelled and will understate performance until the time lag settles.

Run this before any live switch and you get a preview of the swing, campaign by campaign, with zero risk to your actual bidding.

Updating Target CPA and Target ROAS after a model change

Switching attribution models redistributes credit, and redistributed credit changes your Cost / conv. and Conv. value / cost numbers even when nothing else about the campaign has moved. If your tCPA or tROAS targets stay untouched, you are now bidding against numbers that no longer reflect the account’s real performance.

The fix is a straightforward percentage adjustment using the Model comparison report for geotargeted listing ads. Compare Cost / conv. (current model) against your old model’s Cost / conv. for the same date range and work out the percentage difference. If data-driven attribution shows a cost per conversion 15% higher than last click reported for the same period, as an illustrative structure, you would apply roughly that same percentage increase to your live tCPA target, because the bid strategy is now optimising against a higher baseline cost per action for the same volume of sales.

This is best practice specifically because Smart Bidding strategies read conversion data continuously. Leave tCPA where it was before the switch and the algorithm will either under-bid against a target that is now too aggressive, or over-spend chasing a target that no longer matches reality.

Time your changes around the time lag, not around a single day’s numbers. Wait the account’s average days-to-conversion before judging whether the new model has settled, pause any other major changes during that window so you are not reading two signals at once, and treat manual bidding accounts differently: without an algorithm reacting automatically, you have more room to watch and adjust gradually rather than recalibrating a live target the same day.

A decision checklist before you switch models

Not every account is ready for data-driven attribution to earn its keep, and the checklist below tells you quickly where you stand.

• Check conversion and interaction volume against the roughly 200 conversions and 2,000 interactions benchmark Google recommends over a rolling 30 days for the model to work well.
• Confirm offline conversion uploads land within seven days of the interaction, since the attribution engine processes conversions within that window and later uploads can be bypassed by the modelling calculations.
• Match your lookback window to your actual buying cycle rather than leaving the default in place.
• Note how much of your bidding already runs on automated strategies, since that is where a model change has the most downstream effect.

Pro Tip: Run the test on one conversion action first, not the whole account: pick your highest-volume action, compare models for two to three weeks, then decide before you touch anything else.

How we operationalise attribution inside managed campaigns

Reporting on attribution and acting on it are different jobs, and most accounts only do the first one. We run the Model comparison report alongside creative and conversion optimisation work, not as a separate audit, so a credit shift in the data shows up as a creative test or a landing page change, not just a line in a report.

Our first check on any account is always the same: pull Conversions (current model) by campaign, flag the ones with the biggest swing, then recalculate targets before touching spend. This article is written by Harry as part of our broader Google Ads management work.

Where attribution fits in the bigger funnel picture

Attribution inside Google Ads only tells you what happened between a click and a conversion that Google Ads itself tracked. It does not tell you what a customer did on your site before that click, or after it, on a different device or channel entirely.

That is the gap between attribution and funnel analysis. Funnel analysis looks at the whole customer journey, awareness through to purchase and repeat purchase, often stitched together from Google Analytics, your CRM and your ad platforms combined. Cross-channel data-driven attribution in GA4 extends the same modelling logic beyond Google Ads to cover organic search, email and other channels, which gives you a wider lens than the Google Ads interface can offer on its own.

Treat Google Ads attribution as one input into that bigger picture, not the whole picture. A campaign that shows weak assisted conversions inside Google Ads might still be doing real work earlier in the funnel, awareness and consideration, that only shows up when you line up Google Ads data against your broader analytics. If your Google Ads attribution report and your GA4 funnel report disagree about which channel deserves credit, that is not a bug. It usually means one tool is tracking a narrower slice of the journey than the other, and the fix is to decide which tool owns which part of the decision, not to force the two to match.

What attribution insights actually change in a live account

The useful test for any attribution insight is whether it changes a decision, not whether it looks interesting in a report. A campaign with strong assisted conversions but a weak last-click conversion rate is a candidate for a budget increase, not a pause, because last click alone would have recommended cutting it.

Here is the structure of that decision, without inventing a number that was never measured. Say a display campaign shows low direct conversions under last click but appears in 40% of converting paths in the Assisted conversions report, as an illustrative example. Under last click reporting, that campaign looks like a poor performer and a pause candidate. Under the attribution report, it looks like a path-starter that search campaigns are converting later. The correct move is to protect or grow that campaign’s budget and judge it on assisted conversions, not last-click conversions, because last click was never designed to credit the awareness stage of a funnel.

We will not quote a specific before-and-after percentage here that was not measured and dated, because an unverified number is worse than no number. The mechanism above is what matters: attribution insights change which campaigns you fund, not just which campaigns you report on, and the shift usually runs from paused display and YouTube campaigns toward funded ones once their assisted role becomes visible.

Where attribution data is genuinely limited

Attribution reporting is a model, not a measurement, and the most common misconception is treating it as a precise ledger of who caused what. Data-driven attribution is built from patterns in your own account’s converting and non-converting paths, so an account with thin volume gets a thinner, less reliable model, whatever the report shows you on screen.

There are structural limits too. The attribution engine excludes some inventory by design, including the Search Partner Network, Gmail and App campaigns, so your attribution reports will never reconcile perfectly with your standard Campaigns page, and that gap is not evidence of broken tracking. Cross-device journeys where a customer researches on mobile and buys on desktop without staying signed in are invisible to any attribution model that only sees logged interactions, Google’s or anyone else’s.

Nobody has measured a universal “accuracy rate” for data-driven attribution across all accounts, and any number circulating that claims one is not traceable to anything real. The honest position is that attribution models are consistently useful for relative comparisons between your own campaigns and much less reliable as an absolute measure of true causation, because no model, including Google’s, can see every channel and device a real customer actually used.

How to troubleshoot attribution numbers that do not add up

When the numbers in your attribution report do not match what you expected, check the mechanical causes before assuming the data is wrong. Three issues cover most discrepancies.

First, check the time lag. A sudden conversion dip on the last three to seven days is almost always the attribution engine still modelling recent interactions, not a real drop in performance, and it corrects itself once those days settle.

Second, check your offline conversion upload timing. Uploads landing more than seven days after the interaction can be bypassed by the attribution modelling calculations entirely, which quietly understates assisted conversions for any campaign that relies on delayed offline data, like a sales team logging a deal weeks after the lead came in.

Third, check scope before you check maths. Attribution reports exclude certain networks and campaign types by design, so a number that looks low in Model comparison compared to your Campaigns page is often a coverage difference, not a tracking error. Reconcile by filtering your standard reports to the same included networks before comparing totals.

If none of those explains the gap, isolate the conversion action in question and run the Model comparison report on it alone over a fixed date range, excluding the most recent week, before concluding anything is actually broken.

Using attribution to rebalance budget across channels

Attribution reporting earns its budget seat once you stop reading it campaign by campaign and start reading it by funnel role. Group your campaigns into top-of-funnel (display, YouTube, broad search), mid-funnel (generic search, shopping) and bottom-funnel (brand search, retargeting), then pull Assisted conversions for each group rather than for each campaign individually.

A top-of-funnel group that shows heavy assisted-conversion activity but weak last-click numbers is underfunded if you are still allocating budget on last-click performance alone, because last click will always undervalue the channels that start journeys rather than finish them. Shift budget toward that group gradually, in line with the lookback window you have set, and re-check the Model comparison report after each shift rather than assuming the first move was the right size.

Treat this as an ongoing rebalancing exercise, not a one-off correction. Buying cycles shift with seasonality, and a lookback window set correctly in one quarter can understate assisted activity in another if your customers’ research period lengthens, so revisit the Path metrics report every few months to confirm your window still matches reality.

The judgement call most accounts get wrong

The conventional advice on attribution stops at “switch to data-driven and let Google handle it”, and that is where most accounts leave the job half-done. The model switch is the easy half. The half that actually protects performance is going back into tCPA and tROAS targets and adjusting them to match the new Cost / conv. baseline, and most accounts skip that step entirely.

We think the bigger mistake is treating attribution as a reporting exercise rather than a budget decision. A campaign with strong assisted conversions and weak last-click numbers is not a reporting curiosity, it is a funding decision sitting unmade. The accounts getting real value from attribution data are the ones that moved budget toward path-starting campaigns, not the ones that just admired a nicer-looking report.

Prioritise in this order: run the Model comparison report first, adjust bidding targets second, and only then start reallocating budget between campaign groups. Skipping straight to budget moves without adjusting targets first means you are optimising against stale numbers while you reorganise spend, which compounds the error rather than fixing it.

— Harry

Let us run the attribution audit on your account

Checking attribution properly takes time most in-house teams do not have spare between campaign builds and creative requests, and that is exactly the gap our Google Ads management work is built to close. We run the Model comparison report, recalculate bidding targets against the current model, and fold the result straight into creative and conversion optimisation decisions, so insight becomes action rather than another tab left open.

Our attribution audit covers:

• A Model comparison report run against your highest-volume conversion actions.
• A tCPA/tROAS recalibration based on Cost / conv. (current model) versus your existing targets.
• A lookback and conversion window check against your actual buying cycle.
• A budget reallocation recommendation across funnel stages, backed by Assisted conversions data.

The audit carries no obligation to sign on, and we turn around initial recommendations fast. Book a consultation and we will walk you through what your own attribution data is actually telling you.

FAQ

What is attribution in Google Ads?

Attribution is the method Google Ads uses to decide which ad clicks or interactions get credit when a customer converts. Data-driven attribution is the default model for most conversion actions and splits credit fractionally across the path, while last click gives all the credit to the final interaction.

What does 7 day click 1 day view attribution mean?

This refers to conversion windows, not attribution models: a 7-day click window counts a conversion if it happens within seven days of a click, and a 1-day view window counts a conversion within one day of an ad impression the customer saw but did not click. These windows control eligibility for counting a conversion at all, separately from how credit gets split once a conversion is counted.

What is 50/50 attribution?

An even split, sometimes called linear or position-based attribution in older systems, divided conversion credit evenly or by fixed percentage between touchpoints in a path. Google Ads has phased out these fixed rules-based models in favour of data-driven attribution, which calculates fractional credit from actual account patterns rather than a preset percentage.

How much does Google Ads pay per 1000 views?

Google Ads is a platform advertisers pay to run campaigns on, not a publisher network that pays per view, so there is no payout figure to quote here. Advertisers set their own cost per click or cost per thousand impressions bids, and what you pay depends entirely on your bidding strategy and auction competition.