Which figures decide, which only diagnose, and how to compare honestly two campaigns that do not play in the same language.
The figures that decide
Two, and they are enough to settle it.
The cost per purchase attributed to the video, compared with your other creatives over the same period and the same budget.
And the conversion rate of the page the video sends to, before and after it went live. If the page converts better, the video is working, even if its views are modest.
A rule that avoids false victories: always compare at equal budget and over the same duration. A video receiving three times more money mechanically gets better raw figures.
The figures that only diagnose
Three useful indicators, but they decide nothing.
Three-second retention: it judges the hook, not the product.
Click rate: it judges the promise, not the video.
And comments: they show what was not understood, which is gold for the next video.
Those three figures serve to repair, never to conclude. A video can have poor retention and sell correctly; the opposite is even more frequent.
Keep something running permanently
That is what makes every later measurement readable.
Leave one creative running continuously, on a small budget, never touching it. It becomes your benchmark: when it drops, it is the market or the season; when it stays stable and the new one drops, the new one is the problem.
Two conditions for that control to be useful: do not change it, and keep it the same audience.
What survives the campaign
Three assets remain when the budget stops, and nobody counts them.
The raw files, reusable for the whole duration of the rights.
The lessons: which hook worked, which objection kept coming back, which format was watched to the end.
And the relationship with the creators who delivered cleanly: the second shoot with the same person costs less and fails less often.
Note those three things in a single document, or you will buy the same lesson again next year.
What you read, and when
| Indicator | What it tells you | When to read it |
|---|---|---|
| Three-second view | Whether the hook holds | From 1,000 views |
| Full view | Whether the edit holds | At 5,000 views |
| Click to the page | Whether the promise is clear | After two days |
| Purchase | Whether the product follows | After a week |
What distorts a reading
Comparing two videos on different budgets, or judging a creative over three days. A test is read over seven full days, with the same budget on each side.
What you do with each weak segment
A weak hook is replaced, a weak middle is shortened, a weak ending is rewritten. You do not throw out a whole video for one segment.
What to refuse to measure
Two temptations, and they waste time.
The direct comparison between an English-language and a French-language campaign in absolute values: the population pools are not the same size, and a one-to-four ratio crushes any reading. Each version is judged against its own base and its own history.
And the promise of perfect attribution: between tracking blocks, in-store purchases and private sharing, part of the sales will never be linked to the video. Decide with the two figures at the start, and accept the margin of uncertainty rather than paying for a tool claiming to remove it.
