To read a campaign report correctly, you first need to know how each conversion was counted: by which attribution model, in which time window and against which source of truth. Only then is it worth breaking down the averages and asking what happened in the business itself during that period. When the report answers these three questions, it's much easier to make a decision from it.
In brief
- כל פלטפורמה סופרת אחרת. Adding up reports from different systems often counts the same sale more than once.
- מודל הייחוס משנה את התמונה. The exact same conversions will be split differently between channels depending on the chosen model.
- חלון ההמרה קובע מה נספר. A conversion that happened after the window closed simply won't appear in the report.
- דוח טוב נגמר בהחלטה. What worked, what didn't and what to do next month. Sometimes the decision is to move budget, and sometimes to keep going and measure.
What is a campaign report, and what is it supposed to answer?
A campaign report is a summary of what happened in advertising activity over a given period, and it is supposed to answer one question: continue, change or stop.
Many reports start with impressions and clicks and end with conversions and cost. That order is convenient for the systems, but not necessarily for the person who needs to decide. Anyone who receives reports from several platforms knows their numbers don't always reconcile with each other or with sales data.
So it's worth reversing the order. Start with the business goal and the business's own data, such as sales, qualified leads and new customers, and only then check what each channel contributed. When the report is built this way, it's easy to see whether the platforms' numbers match reality.
What is conversion attribution, and why do platforms report different numbers?
Conversion attribution is the method of deciding which of the customer's touchpoints with advertising gets credit for a conversion, and each platform applies its own method to the data it sees.
A customer might see a video on a social network, search for the brand on Google days later, click an ad and buy. The social network's advertising system sees an impression that preceded the purchase, and Google sees a click that preceded the same purchase. Each may record the conversion as its own, and when the reports are combined, you can get more sales than actually happened.
A common solution is a single source of truth, in which each conversion is counted once. It can be the analytics system, the CRM or the sales system, as long as the decision is made in advance and kept over time. A source of truth is an agreed basis for comparison, not absolute truth: GA4 and other systems are also affected by cookie consent, blocking and measurement settings. You keep reading the platforms' reports, but mainly to optimize within each channel, not to compare between them. It's also important to distinguish between attribution and incrementality. Attribution answers which touchpoint received credit for the conversion. Incrementality answers whether the conversion would have happened without the advertising, and is tested in an experiment that compares a group exposed to the advertising with a similar group that wasn't.
What is an attribution model, and which models are available in GA4 reports?
An attribution model is the rule by which credit for a conversion is divided among the touchpoints in the customer's path, and GA4 today has three main options.
According to the Google Analytics Help Center at the time of writing, in GA4 you can choose between data-driven attribution, Paid and organic last click and Google paid channels last click. The first click, linear, time decay and position-based models were removed in November 2023. It's also worth knowing that in GA4, direct traffic gets credit only in a path made up entirely of direct visits. As soon as another channel appears in the path, the credit moves to it.
Data-driven attribution is a model that uses machine learning to estimate how much each advertising touchpoint changed the likelihood of conversion. According to Google, the model compares paths that led to a conversion with paths that didn't, and takes into account, among other things, how much time passed until the conversion, which device the user came from, how many times they were exposed to ads and in what order. An important result for the report reader: conversions can be reattributed up to seven days after they happened, so data from the last few days may still shift.
Another detail that changes the reading: in GA4, changing the attribution model in reports also applies to historical data, while changing the lookback window applies only going forward. If someone changed settings in the middle of the period, the month-to-month comparison is no longer clean.
What is a conversion window, and how does it change the numbers?
A conversion window is the period after an ad interaction during which a conversion is still credited to it, and it directly determines how many conversions appear in the report.
In Google Ads, according to the Help Center at the time of writing, the default window for clicks is 30 days, for engaged video views three days, and for view-through one day. A window of one to 90 days can be set, depending on the conversion source, and Google recommends at least seven days. Changing the window is not retroactive, and automated bidding strategies optimize according to the selected window.
GA4 has an equivalent setting called the lookback window. According to Google, for new-user acquisition events, such as first_visit, the default is 30 days, with an option for 7, and for other key events the default is 90 days, with an option for 30 or 60. When comparing Google Ads with GA4, a difference in windows alone can explain part of the gap.
The practical implication is simple. For a product whose purchase decision takes weeks, a short window will make advertising look weaker than it is. For a cheap, readily available product, a very long window may attribute to advertising purchases that would have happened anyway.
Why does ROAS in the report need another metric alongside it?
ROAS measures the value of recorded conversions relative to advertising spend, meaning revenue and not profit. In a campaign report it's worth showing a guardrail metric alongside it, such as the share of new customers or gross profit after advertising costs, so the decision doesn't rest on revenue alone. We detailed the calculation of break-even ROAS and the difference between ROAS and POAS in the article on ROAS and profit.
Which metrics in a campaign report are misleading, and what should you ask about each?
Almost every metric in a campaign report answers one question and leaves others open, so it's worth knowing in advance what it doesn't tell you.
A few definitions before the table. CTR is the click-through rate, the percentage of impressions that ended in a click. CPA is the average cost per conversion. A key event is an action defined in GA4 as important to the business, such as a purchase or a form submission.
| Metric | What it measures | What it doesn't tell you | The question to ask |
|---|---|---|---|
| Impressions | How many times the ad was shown | How many different people saw it, and whether they noticed it | What is the reach, and what is the average frequency? |
| CTR | What share of impressions ended in a click | Whether the people clicking are relevant, and whether they did anything after the click | What happened on the landing page after the click? |
| Cost per click | How much one click on the ad costs | Whether the click turned into a site visit, and whether that visit is worth anything to the business | What is the cost per action that really matters? |
| Conversions or key events | How many defined actions were recorded | Whether each conversion is a sale, and whether it was also counted on another platform | By which model and window was this counted? |
| CPA | Average cost per conversion | The quality of the conversions, and the gap between the campaigns that make up the average | What does it look like broken down by campaign, audience and device? |
| ROAS | Attributed revenue for every shekel of advertising | Profit, product costs, returns and incrementality | How much of this is new customers, and what remains after costs? |
| Video views | How many times the video was viewed according to the platform's definition | How much of the message got through, and whether it affected search or sales | What is the definition of a view, and what happened in branded searches? |
How do you read a campaign report step by step?
Read it from the business to the platform, not the other way around: first the business result, then the measurement settings, and only at the end the performance breakdown.
- מתחילים מנתוני העסק. Sales, qualified leads and new customers in the same period, from the agreed source of truth.
- בודקים איך נספרה ההמרה. Attribution model, conversion window, and whether someone changed settings in the middle of the period.
- משווים לתקופה דומה. A month with a holiday versus a regular month, or a promotion period versus a period without one, can look dramatic. It's better to compare with the equivalent period and mark unusual events.
- מפרקים את הממוצעים. An excellent average cost can hide one campaign that carries everything and three that spend budget without a sufficient result. Break it down by campaign, audience, device and placement.
- בודקים את השלב הבא. Clicks and engagement are easy to improve without the business gaining. The question is whether visitors stayed, left details, bought or came back.
- מתעלמים בזהירות מהימים האחרונים. Because of conversion delays and reattribution, the freshest data may still change.
- מסיימים בהחלטה. Three lines: what worked, what didn't, and what to do next month. Sometimes it's a budget change, and sometimes the decision is to keep things as they are and collect more data.
How we approach this at CONNECT
At CONNECT, part of the Bruckner Yaar Levi advertising group, we start every measurement project by agreeing on a single source of truth and its settings: what counts as a conversion, under which attribution model and in which time window. We document these settings in advance, so that later in the year the discussion is about results and not interpretations. The reports we deliver end with an explicit recommendation: move budget, change something in the campaign, or keep going and measure.
Between reports, AI agents for campaign monitoring agents track the accounts and flag anomalies in performance and spend pacing, so some problems can be identified before the end of the month. The agents don't replace the human reading of the report, but they shorten the time between a problem and the response.
Frequently asked questions
Why don't the numbers in Google Ads and GA4 match?
There are several common reasons. The two systems may use different attribution models and different time windows, so the same activity is counted differently. The definitions themselves also differ: according to Google, a key event in GA4 is not necessarily a conversion in Google Ads, and the historical gap between the two definitions was one of the reasons for separating the terms. It's worth choosing one source of truth and documenting each system's settings.
What is data-driven attribution in GA4?
Data-driven attribution is a model that uses machine learning to divide credit for a conversion among the touchpoints in the customer's path. Instead of a fixed rule, such as full credit to the last click, the model compares paths that ended in a conversion with paths that didn't, and estimates how much each touchpoint increased the likelihood of conversion. According to Google, conversions can be reattributed up to seven days after they happened.
How long should you wait before drawing conclusions from a report?
It's worth waiting at least until the period's conversion window has closed, and with data-driven attribution in GA4 a few more days, because credit for conversions can be updated afterwards. Budget decisions based only on the last few days may rest on partial data.
Which attribution model should you choose in GA4?
In many cases data-driven attribution is a reasonable starting point, because it doesn't give all the credit to one touchpoint. However, it is less transparent, so it's worth comparing it from time to time with the last-click model and seeing which channels receive more or less credit. Most important is to choose one model for reporting, document it and not switch it in the middle of a comparison period.
Why do yesterday's conversions change in the report?
There are two main reasons. The first is conversion delay: some customers buy days after the click, within the conversion window, so conversions keep accumulating even after the period has ended. The second is reattribution: according to Google, in the data-driven model in GA4, credit can be updated up to seven days after the conversion. So it's not advisable to make budget decisions based only on the data of the last few days.
ראו גם: Data & Measurement · Performance Campaigns · ROAS and profit · פרטיות ומדידה · תוכנית מדיה שנתית · Paid Campaign Management
Sources: Google Analytics – Attribution models · Google Analytics – Attribution settings · Google Analytics – Key events · Google Analytics – Advertising and attribution · Google Ads – Conversion window · Google Ads – Data-driven attribution · Google Ads – Target ROAS
רוצים לבנות דוח אחד שכל ההנהלה מסכימה עליו? השאירו פרטים.