An ad report is read from the outside in: what was spent and what came back first, then the attribution window those figures are counted under, and only after that the intermediate metrics. Read the other way round, starting at CTR or at ROAS, any campaign can be made to look good or bad depending on which row you pick.
The reading order, and why the order matters
Almost nobody reads a report, they glance at it. The eye jumps to the biggest number, to the one shaded green, or to whatever the client asked about last month, and everything else gets interpreted to fit that first impression. A fixed reading order is the only practical defence, because it forces you to see the money before the explanation.
| Order | What you look at | What it does not tell you yet |
|---|---|---|
| 1 | Spend for the period | Whether it was well distributed |
| 2 | Result for the period | Whether the result is yours |
| 3 | Attribution window in use | Nothing alone, but it conditions everything above |
| 4 | Cost per result against target | Why it moved |
| 5 | Delivery: impressions, frequency, reach | Whether the ad or the audience is the problem |
| 6 | Clicks and conversion on the destination | Whether the page holds what the ad promised |
The first two rows are facts about a bank account and nobody argues with them. The third is a reading condition rather than a figure: without it the other five mean nothing. The last three are diagnosis, which is where you decide what to change, not whether something is working. Row five is the one most often skipped, and it separates the two problems that look identical on a results slide: an audience that has seen the same ad too many times and an ad that nobody wanted to see in the first place. The first shows up as frequency climbing while cost per result climbs with it, the second as impressions arriving normally and nothing happening after them.
When the order gets inverted you get the report everyone has sat through: six slides of excellent intermediate metrics with no mention of money, or the opposite, a headline of superb results that deflates the moment somebody asks since when each conversion is being counted.
The window changes everything, and it is almost never printed
The attribution window is the period during which a platform claims what happened after a click or an impression. Changing it does not change the business, it changes every number in the report at once: conversions, cost per result and ROAS. Two reports of the same campaign under two different windows are not comparable, and nothing in the layout warns anybody about it.
There are three things worth printing on the first page of a report that almost no report prints.
- Which window was applied, separating click from view. A conversion credited to somebody who saw the ad and never clicked is a far stronger claim than one credited to a click, and both land added together in the same column.
- Whether it is the same window as last month. When somebody changes it in the account settings, earlier history is not always recalculated, and suddenly you have an improvement that exists only on the sheet.
- Which model shares out the credit. Every platform documents its own in its own help pages, and none of them shares credit the way the other four do.
If the report is going to carry a budget decision, fix the window at the start of the quarter and leave it alone until the end. Switching windows halfway through to see a campaign in a better light is the cleanest way to lose the ability to compare anything for six months.
Why the month is not the sum of the weeks
This is the classic phone call on the 3rd: the client has added up four weekly reports and the total does not match the monthly one. Neither document is wrong.
Conversions are booked to the day of the click or the view, not to the day the person paid. Because the window lasts days, a purchase made today can land on a day of last week, and that weekly report has already gone out. That is why a freshly pulled figure keeps rising for several days and then settles, and why checking yesterday's performance first thing in the morning is the worst habit available.
Three other sources of mismatch get blamed on attribution and have nothing to do with it.
- The ad account's time zone. The account's day and your analytics day can start at different hours, and on campaigns that spend overnight that moves spend from one day into the next.
- Reach does not add up. Somebody who saw you on Monday and on Thursday is one person. A monthly reach obtained by summing four weekly reach figures is always inflated.
- Currency and exchange rates. An account billed in another currency is converted at the day's rate, and the period total will not match the sum of the converted parts.
ROAS is not profit, and this is where a client is lost
ROAS is attributed revenue divided by spend, and it deducts nothing: not the cost of the goods, not returns, not shipping, not tax, not your own fee. A client who reads that number as profit and then looks at their accounts does not conclude they used the wrong metric, they conclude you sold them smoke.
The honest conversion is the break even ROAS, and it comes out of their gross margin, not yours. At a 50 % margin you need twice as much revenue as spend just to stand still; at a 25 % margin, four times. The same number can therefore be a success in one business and a slow bleed in the one next door, which is why the margin question belongs in the contract conversation and not in the month four review.
Two more adjustments move the figure without anybody mentioning them. The first is whether the revenue the pixel sends includes VAT, the value added tax charged in Spain, because that money is not yours and still inflates the numerator. The second is returns, which arrive weeks later and are never deducted from the report of the campaign that brought them. To run that calculation on your own figures before showing anyone, there is the CPM, CPC and ROAS calculator, and the definition of each acronym with its traps is in the ad metrics guide.
Everything in this article happens in one place in GoFeed.
Try it freeThe report you can defend in front of a client
A defensible report has four blocks and none of them is optional.
- What we spent and what we got, with the window written beside it.
- What we are comparing it against, with the same objective, the same length and the same seasonality as the reference period.
- What we changed during the period, with dates. A creative swap on the 12th explains half the graph and costs one line to write down.
- What we do not know. Untagged traffic, conversions that never reached the pixel, whatever the platform modelled.
The fourth block buys the most trust and is the one almost nobody includes. Saying up front which part of the measurement is estimated avoids the awkward conversation next month, when the client crosses your report against their real sales and finds the gap on their own. Put the business figure next to the platform figure, even when it is uglier, and explain why they differ. The full mechanism behind that gap is in the UTM and attribution guide.
Where these figures come from
The five platforms do not report alike, and reconciling that is half the work of putting them together. Meta, LinkedIn, TikTok and Pinterest answer a normalised query; Google has its own query language and its own field names, so its numbers are translated into the same columns as the rest, which is what lets a brand's month be read in one go. Google also brings two things none of the others have: the real search terms and the keyword planner.
Each ad account hangs off its own brand rather than off the workspace, so one brand can never drag another one's spend into a report. Exactly what each platform allows inside the dashboard is in the ads dashboard guide, and how an account is connected is on the ads page.
One last thing about cadence. A monthly report reads better than four weekly ones, because the attribution window has already closed over almost everything it counts. If somebody needs to look every week, have them look at delivery rather than at results: that is where the early signal is, and it is the one that says whether something has to change before the month ends.