The sales figure during a promotion is easy to obtain and almost useless on its own. Everything depends on what it is compared against.

The counterfactual is the whole problem

A promotion's value is the difference between what happened and what would have happened without it, and the second figure cannot be observed.

Every method of evaluation is therefore an attempt to construct a plausible estimate of an outcome that did not occur.

Different reasonable methods produce different estimates, sometimes differing enough to reverse the conclusion about whether the promotion paid for itself.

Comparing to previous periods is the weakest method

Measuring against the preceding weeks or the same period last year is simple and is what most reporting defaults to.

It attributes to the promotion every difference caused by weather, competitor activity, supply problems, pricing changes elsewhere and general demand movement.

In categories where those factors are large, the resulting figure describes the period rather than the promotion, and the two are not separable.

Holdout groups give a cleaner reading

Withholding the promotion from a comparable set of locations or customers produces a direct observation of what happens without it.

The comparison is valid only if the two groups are genuinely similar and nothing else differs between them, which requires care in how they are selected.

The method costs the sales that the excluded group would have generated, and that cost is why it is used selectively rather than as standard.

The measurement window must extend past the promotion

Because promotions pull purchases forward, a window ending when the promotion ends captures the increase and none of the subsequent shortfall.

Extending the window until sales return to their normal level gives the net figure, which is usually substantially lower than the promotional-period result.

Choosing where to end the window is one of the most consequential decisions in the analysis and is frequently made by convention rather than by evidence.

The level of measurement changes the conclusion

A promotion can lose money on the promoted item, gain at category level through additional visits, and lose again at store level if it displaced other purchases.

All three figures are correct, and the one that matters depends on what the promotion was intended to achieve.

Disagreements about promotional performance are frequently disagreements about which level to measure at rather than about the underlying numbers.