A row verdict is one action instead of a column of numbers: cut, too early, scale. The most expensive mistakes in advertising happen exactly at this step: people pause a campaign whose tags simply do not arrive, or a helper campaign that last touch undervalues. That is why verdicts follow rules that are the same for Breakdown, Entities and the “Overspend against the norm” card in the What to do now section: the report and the card cannot disagree. Breakdown has one addition of its own — see “When a source’s leads fell at once”.
What is compared
Spend is the row’s ad spend.
Result is the metric of the project’s main goal, if KO knows it and can count it. The goal is set by Main goal of the project in the brief of the Context section: leads, sales, revenue or keeping customers. If no goal is set, or it cannot be counted, the result is leads. Sometimes the goal is rare (repeat purchases, for example) and the rows with spend have no results for it in the window. Then verdicts use leads, and the Findings button says so.
The norm is the cost per result that every row is compared with. It has two sources:
- The section plan. If Plan is on in Breakdown and it has targets for both spend and the result, the norm is planned spend divided by the planned result.
- The table average — in every other case, and always in Entities. It is the spend of all rows with spend divided by their result. Rows without spend (organic, direct) stay out of the norm: they have no cost.
The table average is a reference point, not a target. If everything is expensive, every row looks normal against it. That is why, when a plan is on, KO also checks the whole table against the plan — see “Findings about the table” below.
”Worse than the norm” is a test, not a comparison of two numbers
At the norm, a row should have brought spend ÷ norm results. It is worse than the norm when it
brought so few that its cost is more than 20% above the norm. A smaller gap is still the
norm: the cost per result moves up and down in good campaigns too.
The second condition matters more than the first. With small numbers the difference is chance: 2 leads where the norm expected 4 is not a verdict yet. KO takes the 95% interval for the row’s number of results and tests the value most favourable to the row — the upper bound. A row is proven worse than the norm only if even at that value its cost is above the norm by more than the tolerance. Looks worse but not proven means “too early”, not “cut”.
The width of the interval depends on how many rows get a verdict, that is, how many rows have spend. For a single row it is the usual 95%. But among forty campaigns, one will look “above the norm” just by chance, and it gets “cut” only because there are many rows. So the more rows are compared, the wider the interval for each one: a random miss does not become a verdict.
KO treats zero the same way. Zero leads on spend of one norm can still be chance; on spend of several norms it is not.
The order of verdicts
The verdict is the first rule that fires, from top to bottom.
- The row is proven worse than the norm. What happens next depends on assists — leads
where the row was one of the touches, not the last one:
- no results by last touch and none with assists → check tracking. A campaign that spends and has not brought a single lead in any role more often loses its tags than works for nothing. Pausing it before checking is exactly the mistake the verdict is for;
- assists are not counted for this row → above norm. Its role in the touch chain is not visible, and the hint suggests checking whether the row brings people who buy later;
- with assists the row is still proven worse than the norm → cut;
- with assists it is no longer proven → don’t cut blindly. It is expensive by last touch, but the row takes part in other leads, and turning it off may remove them.
- Worse than the norm, but not proven → too early. There is too little data; take a longer window.
- Clearly cheaper than the norm: even the lower bound of the interval is not worse than the norm, and there are noticeably more results than the norm expected → scale. This is a place for more budget.
- Everything else — the row is within the norm and gets no badge.
Assists are counted where lead attribution counts them: in a Breakdown by tags they are. Where there are no assists, there is no “don’t cut blindly”: the row gets “above norm” instead of “cut”.
When a source’s leads fell at once
“The order of verdicts” looks at one window and does not see what came before it. That is why “check tracking” has a second path — through the engine finding “Why a source lost its leads” in the What to do now section. The engine splits the drop in leads of a paid source into clicks, visits and the visit-to-lead rate. If leads fell in the same week in all campaigns of the source, while visits stayed the same as before, leads seem to lose their tag.
While such a finding is open and recent, Breakdown does not trust the cost of this source’s rows:
- instead of cut, above norm, don’t cut blindly and too early the row gets check tracking. Scale and rows within the norm do not change;
- this applies to the source row and to every row under it, its campaigns for example, if the source is the first level of the breakdown. A breakdown whose first level is the campaign does not see it: one campaign tag can belong to different sources;
- the section window has to cover at least one week after the step. On a window that ends before the step, the row cost is honest, and the verdict follows “The order of verdicts”.
The hint on such a badge is different: “Leads of the source fell in all campaigns at once, while visits stayed the same — leads seem to lose their tag. Until this is checked, the row cost is based on lost leads”. Below it are the week of the step, how many leads are missing against the earlier pace and which verdict the row would get by cost alone.
Entities has no such override: there the verdict goes by cost only.
The campaign status in the ad account does not change the verdict. The faded no spend badge on a campaign that is active in the ad account but spent nothing in the period is a note about a mismatch between the account and the data, not a judgement of the campaign.
Overpayment and “Where to look”
Overpayment is the money a row spent above the norm: spend − result × norm. Rows worse
than the norm have it; for “scale” and rows within the norm it is zero. For “don’t cut blindly”
the hint shows a second overpayment — counting assists — and it is smaller than the first. A row
that the step in a source’s leads turned into “check tracking” has no overpayment: its cost is
based on lost leads.
Where to look sorts the rows in groups by urgency: check tracking → cut → above norm → don’t cut blindly → too early → scale → within the norm. Inside a group rows go by overpayment, the most expensive at the top.
Findings about the table
A row cannot tell what is going on with the whole table, so the Findings button next to the period dates works out four findings on paid spend — the spend of the rows that have any:
- The numbers are not reliable — a noticeable share of paid spend sits on “check tracking” rows. The norm is then based on incomplete data, and the verdicts of the other rows are wrong with it. This is the only red finding. Its text says where the rows come from: they have no results at all, their source’s leads fell at once while visits stayed the same — or both.
- Too early to judge — most of the paid spend sits on “too early” rows. In this window most of the budget cannot be judged yet.
- Everything is expensive, not single rows — only with a plan. The cost per result across all paid rows together is above the plan by more than the same 20% tolerance — or there are no results at all.
- Overpayment sits in a few rows — the overpayment of “cut” and “don’t cut blindly” rows together is a noticeable share of paid spend. The finding says how many rows make 80% of it: there are usually few of them, and that is where to start. The finding and the “Overspend against the norm” card use the same threshold, so they appear together.
What the verdict does not know
- It is not a forecast. Overpayment is what has already been spent above the norm, not what comes back once the row is off: the platform will redistribute impressions on its own, and the cost of the other results may go up.
- The verdict covers the section window. A short window gives more “too early”; a sales cycle longer than the window lowers the result of every row at once.
- A row has to be something you can turn off. In Breakdown verdicts appear only when the first level is a tag; a breakdown by device or country gets none. In Entities — only for entities with ad spend, campaigns for example.
Related
- The Breakdown section — where to turn verdicts on and how to read the Findings button.
- The Entities section — verdicts on campaigns from the catalog.
- The What to do now section — the “Overspend against the norm” card, built on the same rules.
- The Context section — the project’s main goal, which decides the result.