How to read your Google Ads attribution report — and why it's probably lying to you

Pull up the attribution report in Google Ads and you get a number: conversions attributed to your campaigns across the selected date range. It looks precise. It updates daily. It has a model selector at the top that most people have never changed.

Then you open GA4 and the number is different. Then you check your CRM and the number is different again. At this point most people either pick the highest number and move on, or spend an afternoon trying to reconcile three systems that were never designed to agree.

Neither approach is right. The attribution report is not broken — it is doing exactly what it was built to do. The problem is that what it was built to do is narrower and more specific than most people realise. Understanding that gap is what turns the report from a source of confusion into a genuinely useful decision-making tool.

What the attribution report is actually measuring

The Google Ads attribution report answers one specific question: across the touchpoints Google observed, how should conversion credit be distributed?

It does not ask whether Google caused the conversion. It does not consider what happened on other channels. It does not know about the Meta ad that ran three days before the search, or the email newsletter that reactivated a lapsed customer, or the trade show where someone first heard about the product. It only sees what passes through Google's ad ecosystem: impressions, clicks, and the conversion events you have told it to track.

This is not a flaw. It is a design constraint. The report is built to help you make decisions about Google Ads — which campaigns, which keywords, which audiences deserve more budget. Used for that purpose, it is a reasonable tool. The problem comes when it is used to answer a broader question it was never designed for: is our advertising working?

Data-driven attribution: the black box problem

If you have looked at the attribution model selector in Google Ads recently, you will have noticed that data-driven attribution (DDA) is now the default and is described as the most accurate model available. That claim is worth examining.

DDA uses machine learning to assign fractional credit to touchpoints based on observed conversion paths in your account. Instead of giving 100% credit to the last click, it distributes credit across the interactions that appear to have influenced the outcome. In principle, this is better than last-click. In practice, it introduces a problem: you cannot see what it credited, or why.

Last-click attribution is blunt, but it is transparent. You know exactly which campaign, which keyword, and which ad received credit. DDA is a black box. Google does not surface the weighting assigned to each touchpoint. You can see the aggregate output — this campaign got X conversions — but you cannot audit the calculation behind it.

There is also a systemic bias worth noting. DDA is trained on observed conversion paths within Google's ecosystem. It can only credit what it can see. Upper-funnel interactions that happened to coincide with eventual conversions get weighted into the model, but the model has no way to distinguish between correlation and causation. A YouTube ad that ran while a customer was already in-market gets credit it may not have earned.

"If you add up what Google attributes to itself and what Meta attributes to itself across a single customer journey, the total often exceeds your actual sales. Both platforms are technically correct within their own reporting window. Neither is telling you the whole story."

This is not unique to Google. Meta's attribution report has the same architecture: it distributes credit across Meta-visible touchpoints and cannot see anything outside its ecosystem. Run both platforms simultaneously and both will claim credit for the same conversions. A customer who saw a Meta ad on Tuesday, searched on Google on Thursday, and converted on Friday will appear in both reports as a conversion attributed to that platform. Neither report is wrong. They are each telling you a partial truth as though it is the complete picture.

This is the structural reason why above-platform measurement matters. If you are running multiple channels and relying solely on platform-reported attribution to judge what is working, you are using a mirror that can only reflect what it can see. We covered the practical alternative — Media Efficiency Ratio — in detail in the ROAS article and the MER/POAS implementation guide. Attribution reports are useful inputs into that picture. They are not the picture itself.

Attribution windows: the settings nobody changes

Every Google Ads account has attribution window settings that determine how far back Google looks when connecting a conversion to an ad interaction. The defaults are 30 days for clicks and 1 day for view-throughs. Most accounts have never changed them.

These defaults were not set based on your business. They are Google's baseline, applied universally regardless of whether you sell impulse-purchase products online or high-consideration B2B services with a three-month sales cycle. The result is that for most accounts, the attribution window is either too long or too short — and in both cases it is distorting the conversion count in ways that affect real budget decisions.

The tool that tells you what your window should actually be is the Conversion Delay report in Google Ads. It shows the distribution of time between first click and conversion across your account. If 85% of your conversions happen within seven days of first click, a 30-day window is counting conversions that have increasingly tenuous connections to the ad interaction. If a significant portion convert beyond 30 days, your current window is cutting those off entirely.

Most accounts have never opened this report. It takes five minutes to run and it will tell you, with your own data, whether your current attribution window is set up to reflect how your customers actually buy.

The GA4 discrepancy: why the numbers will never match

The question comes up in almost every client review: why does Google Ads show more conversions than GA4? Sometimes it is the reverse: GA4 shows more. The numbers rarely agree, and the gap can be 20%, 30%, or more.

This is not a tracking error. It is expected behaviour — the result of three fundamental differences between how the two systems work.

1. Different attribution models by default. Google Ads uses data-driven attribution to assign conversion credit. GA4 can be configured to use various models depending on which report you are reading. Even when both are set to the same model in theory, they implement it differently at the level of session and event counting.

2. Different counting methodology. Google Ads counts conversions — an event that fires when a conversion action is completed. By default, it can count multiple conversions from a single user. GA4 can count sessions, users, or events depending on how your configuration is set up. A customer who completed a purchase twice in the same period might appear as two conversions in Google Ads and one in GA4, or the reverse, depending on your settings.

3. Cross-device and cross-browser gaps. Google Ads benefits from signed-in Google account data, which allows it to connect a search on a work laptop with a conversion on a mobile phone. GA4 relies primarily on cookies and its own client ID, which does not cross device or browser boundaries. A journey that Google Ads tracks as a single user across three sessions might appear as three separate users in GA4.

The cross-channel blind spot

Google's attribution report is, by definition, a Google-only view of the world. It can tell you a great deal about what happened within Google's ecosystem. It cannot tell you anything about what happened before a customer entered that ecosystem, or what other channels contributed to the outcome.

In a single-channel business this is a minor limitation. For any business running Google alongside Meta, email, organic search, or any other channel, it is significant. Consider a straightforward scenario: a prospect first sees a Meta retargeting ad on Monday. On Wednesday they do a Google search, click a Shopping ad, and browse without purchasing. On Friday they do another Google search, click a Search ad, and convert. Google's attribution report distributes credit between the two Google touchpoints. Meta's report claims the retargeting impression as an attributed conversion. The actual journey involved three channels. Two of them are claiming full or partial credit, and the combined attribution adds up to more than one sale.

Neither platform is being deliberately misleading. They are each reporting what they can see. But the combined effect of reading both reports as truth is a systematically inflated picture of advertising efficiency. Each channel looks better in isolation than it actually is in context.

This is the core argument for measuring at the business level rather than the channel level — using total revenue divided by total media spend (MER) as the primary scorecard, and treating platform-reported attribution as a directional signal about what to do within each channel, not as evidence that the channel is working. We have covered the mechanics of that framework in detail across the ROAS and MER/POAS articles.

Three things worth doing before you read the report next time

The attribution report is not the enemy. It is a specific tool with a specific scope, and within that scope it does useful work. The problems come from using it to answer questions it was not built to answer: whether your advertising is working, whether a channel deserves its budget, whether you are generating more value than you are spending. Those are business-level questions. The attribution report is a campaign-level tool.

Read it for what it is good at. Use it to make decisions inside Google Ads. Then take a step back and measure the outcome at the level where it actually matters.