Guide

How long to run a campaign before you can judge it

Every dashboard will compute a cost per result from one result. The number is arithmetically correct and worth nothing, and acting on it is how campaigns that were about to work get switched off. The question that comes first is not what is the number but is there enough here to read a number at all. That has an answer in plain arithmetic, and it is three thresholds deep.

The three thresholds, in the order they apply

The order matters as much as the numbers. These are the thresholds our own engine uses, rendered here from the engine itself — so this page cannot drift from what the product does.

1. Enough results
10 results in the window
Reached, nothing else is consulted. A cost per result is spend divided by results, and what fixes its precision is the number of results — not how many people saw the ad.
2. Enough delivery
5,000 impressions in the window
Below this, zero results is the expected reading, not evidence. A spend threshold cannot stand in for this one: the same amount of money buys wildly different amounts of delivery depending on the CPM.
3. Enough spend to call zero a verdict
3× your cost ceiling, e.g. $90 at a $30 ceiling
Only applies when results are still zero. Stopping earlier kills every campaign with a slow start; with no ceiling set, there is no basis for stopping at all and the honest output is silence.

Why results are checked before delivery

Put the impression threshold first and search campaigns never become readable. Search buys few impressions at a high CPM and converts a large share of them; an account can spend for a month, collect results well past the threshold that fixes precision, and still sit below an impression count that was only ever a proxy for has delivery started. Once the results are there, that proxy has nothing left to add: a cost per result is spend divided by results, and its precision is set by the count of results, not by how many people saw the ad. The last row of the table below is that case.

Worked examples

Invented inputs, real engine. Each verdict below is produced by calling the same function the product calls, so the wording changes if the thresholds ever change. Note that a cost per result is computable in most of these rows — being computable is not the same as being readable.

Situation	What the window holds	Can it be judged?
First day live	$18 spent · 900 impressions · 0 results ceiling $30 · no result to divide by	Not yet: 4,100 more impressions before the question can be asked at all.
Delivering, still nothing back	$60 spent · 12,000 impressions · 0 results ceiling $30 · no result to divide by	Not yet: $30 more would have to be spent before "nothing" counts as waste.
Still nothing back, later	$120 spent · 20,000 impressions · 0 results ceiling $30 · no result to divide by	Decidable: 3× the ceiling has been paid and nothing came back.
A few results, above the ceiling	$150 spent · 20,000 impressions · 4 results ceiling $30 · reads $38 per result	Not yet: 4 results so far, 6 more before the cost per result means anything.
Enough results to read	$400 spent · 22,000 impressions · 16 results ceiling $30 · reads $25 per result	Decidable: 16 results is enough to read a cost per result.
Search-shaped: few impressions, many results	$500 spent · 3,000 impressions · 40 results ceiling $30 · reads $13 per result	Decidable: 40 results is enough to read a cost per result.
Nothing back, and no ceiling was set	$120 spent · 20,000 impressions · 0 results ceiling none set · no result to divide by	No answer: with no cost ceiling there is no yardstick for how long zero results is too long.

The row with no ceiling is the one people find hardest to accept. Nothing came back and delivery was plentiful, yet there is still no verdict — because too long is measured in money, and the only yardstick for how much money is what a result is worth to you. Without that, the honest output is silence, not a stop.

What this replaces

The usual rule of thumb is a time box: give it a week, give it two. Time is the wrong unit. A week at a small daily budget can hold less evidence than a day at a large one, and the same week on search and on social produce delivery counts that differ by an order of magnitude. What a decision needs is a count of results, and a floor of delivery underneath it to tell an empty campaign apart from a bad one.

The other common rule is a spend threshold on its own — stop it once it has burned some amount with nothing to show. That one fails in a specific direction: the same amount of money buys wildly different amounts of delivery depending on the CPM you happen to be clearing, so a fixed spend threshold stops expensive-inventory campaigns before they have been seen enough to have failed. Spend is the right yardstick only for the narrow question of when zero results has been paid for long enough, which is why it sits third and applies to that case alone.

Check it against your own export

You can run the counts above on an Ads Manager export without connecting anything: our CSV check reads the file in the browser you already have open and sends it nowhere — no sign-up, no account connection, no email address. If you are not sure which columns your export needs to make results and impressions readable at all, that is the column-by-column guide.

Check a CSV export →What Loop Marketing does

Or run it on connected accounts instead of an export — start a free trial. No credit card to begin.