Build a weekly performance review

The day is fixed so the series is comparable and so a partial day never lands inside the window. Today is always partly filled in — a range ending today is reporting a day that is not over, and comparing it against whole days manufactures a decline.

The window is fixed for the same reason and one more: a window chosen after seeing the numbers is a window chosen to support a conclusion. Pick a length long enough that one bad day cannot carry it, and then never move it.

Fixing both has a second benefit that is easy to miss. When the procedure is constant, a week with no finding is information. When the procedure varies, a week with no finding just means nobody looked hard enough yet.

Ninety seconds, every week, before any interpretation.

Connection status. A connection whose last sync finished with errors under-reports, because some accounts landed and some did not. A connection that has never synced answers with nothing at all.

Restatement. Last week's figures are not frozen. Platforms keep revising a trailing window and each sync takes the revision, so a number you wrote down on Monday can be a different number on Friday. The connection carries a restated mark when a sync rewrote a lot of stored history, and that mark is the thing to look for rather than the difference itself.

Estimates. A total spanning several currencies is converted at published rates into whichever currency carries the most spend, labelled an estimate, and stamped with the date of the oldest rate involved. Where a rate is missing the total is refused rather than guessed. Know which kind of figure you are holding before you compare it.

Skipping this step is how a restated fortnight becomes a crisis meeting.

Blended spend and results for the window, with the per-channel breakdown, and the same for the window before.

Read the blend against your break-even line rather than against last week's blend. The line comes from your own margin — break-even ROAS and break-even CPA compute it from your figures with no account attached — and it is the only comparison that says whether the operation is worth running.

Then read the mixture. A blend is a weighted average over whatever mixture existed, so if the spend shares moved, most of the change in the blend is composition rather than performance. That check takes ten seconds and prevents most wrong conclusions. Read a blended report honestly is the longer argument.

One rule that matters here more than anywhere: a ratio with a zero denominator comes back empty rather than as a zero, and the result says so. An empty cost per result means there was nothing to divide by. It is not a free result and it is not a catastrophe.

Which campaigns carried the spend in the window, and which spent most while producing least?

Concentration decides whether there is anything to do. If most of the movement sits in one campaign, the account's aggregate is that campaign wearing a disguise and the response is narrow. If it is spread, the cause is usually upstream of the ad accounts entirely and no campaign-level change will reach it.

Rankings come back highest spend first, each carrying a marker saying whether the figures were natively comparable, converted in order to compare them, or left raw. A raw order across currencies is not a ranking, and the marker exists so nobody reads one as though it were.

This is the question that makes the review compound, and it is the one most often skipped.

Every applied change is a row on the record with what it moved, from what, when, and which authority cleared it — you, a credential, or a rule your workspace wrote. Read last week's rows against this week's counts on the same entities.

The arrangement that record sits inside, stated once: your connections read your accounts, a change is written as a draft on the record with the state it would change from and the checks it passed, applying it is gated on approval — by you, or by a rule you wrote — and no platform write reaches a customer account until that platform has granted write access. Applying a cleared change to a platform is in private beta.

Where a platform exposes its own change history, it also says who changed what and which fields moved, which is how you tell a change of yours from somebody else's on an account several people touch.

One change, at most, and one written prediction.

The prediction is the point. Two lines — what you expect to happen and by when — written before you look again. Without it, next week's review becomes a story constructed to fit whatever the numbers did, and the whole exercise stops being measurement.

A change per week is also a pace the machinery pushes you towards anyway. A budget may not move more than a hundred per cent of its current value in one change, applied increases on a connection are summed over a rolling twenty-four hours against the account's own daily ceiling, and every change to how a campaign delivers costs a stretch of unsettled delivery. Fewer, larger, deliberate changes read better and survive the limits.

Ask for the rows in one call rather than walking campaigns one at a time — a single question covering the window across every connected account returns one result, and iteration is both slower and the fastest way to meet the rate limit.

Use the account brief for context rather than for the week: it covers the last ninety days with a comparison against the previous twenty-eight, the objective mix, spend concentration, the rankings and the weekday rhythm, and it is computed rather than live, so say when it was computed before quoting it.

And write the review somewhere that survives. The performance review template is the structure as a document, which matters mostly because it stops the format drifting week to week — and a drifting format is the same problem as a drifting window.

Currency and FX in ad reporting is question one in detail, and Set up UTM conventions that last is what makes question three answerable at all.