Most teams measure what is easy to see, then take costly decisions from figures that say nothing about the sale.
Views and likes say almost nothing
They measure circulation, not decision.
A video can be seen by tens of thousands of people who will never buy, because it appeals to a wider audience than yours. Another can reach few people and convert strongly, because it speaks exactly to those who were hesitating.
Those figures serve one purpose: comparing two videos with each other in the first three seconds.
What predicts a winning variant
Retention at the start, not average duration.
Look at how many people are still there at the third second: that is what separates a hook that works from one that loses everybody before the subject. Then look at the completion rate, which says whether the opening promise was kept.
Those two indicators are read before cost per purchase, because they are available earlier and they explain the rest.
Cost per purchase is the dashboard
It is the only figure that compares comparable things.
It takes in everything the other indicators ignore: the price of traffic, the quality of the audience, the landing page, the basket. A video can have modest figures everywhere and still be the most profitable.
Always compare at equal spend, never at equal duration: a creative launched on Friday and another on Monday did not live the same week.
The product page video is measured differently
It is judged neither in views nor in cost per purchase.
Three figures are enough: the add-to-basket rate of the page carrying it, the return rate linked to a disappointed expectation, and the number of support questions on the points it covers. If those three indicators move in the right direction, it has paid for itself.
Give a creative a fair test
Same audience, same budget, same moment.
A video launched alone is compared with a memory. A video launched with half the budget did not have its chance. And a video judged after a few hours has produced only noise.
Set a spend threshold in advance below which no decision is taken, and apply it to all of them, including the one you prefer.
What to measure, and what each figure says
| The indicator | What it reveals | Its limit |
|---|---|---|
| Three-second retention | The strength of the hook | Nothing about conversion |
| Cost per acquisition | Advertising efficiency | Also depends on targeting |
| Product return rate | The accuracy of the promise | Measured weeks later |
What distorts almost every comparison
The ad budget. A video pushed with ten times the spend looks better without being better, so the comparison only means something at identical budget and targeting.
What to decide before launching a test
The success criterion and its duration. Without a threshold set in advance, every result is justified after the fact, and no learning happens even when the numbers are good.
What can never be measured
A real part of the effect appears in no dashboard.
Somebody sees a video, does not click, searches for your name three days later and buys directly. That purchase will be attributed to nobody, and it exists all the same.
The practical answer is not to give up measuring, it is to track the volume of searches on your name and direct traffic too, month after month. When those two curves rise while you are producing, part of the answer is there.
Sources
Checked on 26 August 2026. Thresholds and rates change: where this guide and the official source disagree, the official source is right.
