What to track when consent banners break your data, why attribution lies, and the three figures that decide.
Your figures under-report, in a stable way
Three practical consequences.
The number of sales attributed to a campaign is lower than the real number, sometimes markedly.
The gap varies little from one week to the next, which makes internal comparisons reliable.
And the channels are not affected equally: what is measured badly is undervalued in favour of what is measured well, generally brand search.
One frequent mistake: stopping a campaign that works because the tool does not see its sales, and increasing the budget of a channel that merely harvests sales already decided.
The three indicators to track anyway
Three figures, in this order of reliability.
The cost per purchase measured by the advertising account, compared only with itself, over equivalent periods.
The total turnover of the period, compared with a period with no campaign: that is the most honest measure, and the most ignored.
And the three-second retention rate, which depends on no consent and judges the video's opening.
A fourth, useful and free indicator: the volume of searches on the brand name, which rises when a campaign works, even with no measurable click.
The offline measures to bring back
Four old methods, useful again.
Asking at purchase how the person heard of the brand, in one single question.
Using a distinct code per campaign, not as a discount but as a marker.
Comparing sales in an exposed geographic area with an unexposed one.
And deliberately pausing a campaign for a week, then measuring the difference.
That last method is the most reliable of all, and the hardest to get accepted internally. Yet it answers the only question that counts: what happens if we stop.
What video reduces and nobody counts
Three savings invisible in an advertising dashboard.
The return rate, when the video honestly shows the product: often the largest gain, and it appears nowhere in campaign tools.
The volume of customer-service questions, which costs time and wages.
And the internal production cost, when a video replaces work done by an agency.
One consequence for the calculation: add those three savings before concluding that a campaign is not profitable. They are measured in other tools, and nobody thinks of bringing them together.
The measurement window
Three benchmarks.
A minimum of three days and a spend threshold per variant, below which the figures mean nothing.
A maximum of one week per test cycle, beyond which the environment has changed.
And a comparison over four full weeks to judge an effect on sales, because the buying cycle is longer than the advertising cycle.
Three questions, three different measures
| The question | What you look at | Over what period |
|---|---|---|
| Does the video hook | The 3-second retention rate | 48 h |
| Does it sell | The ad's cost per purchase | 7 to 14 days |
| Does it last | The day the cost climbs back | 3 to 6 weeks |
What skews a comparison
Changing the edit, the audience and the budget at once. One variable moves at a time, otherwise the result says nothing about the video.
What is kept when a video wears out
The opening, almost always. Reusing the three seconds that work and re-cutting the rest costs less than a fresh shoot.
What to decide before launching
Five lines, written into the campaign document.
The numbered objective, and the indicator measuring it.
The spend threshold per variant.
The comparison period chosen in advance.
The single person who decides, and their deadline.
And the review date.
One remark suited to this market: that document is not a formality. In a working culture valuing writing, it obtains a budget more easily than an intuition presented orally, and it avoids the argument about the reliability of the figures, since the method was accepted before the result was known.
One last marker, more important than any tool: impressions, likes and comments measure nothing commercial. They describe a reaction, not a decision, and a campaign can be widely watched without selling a single item.
