Incrementality: measuring sales that wouldn't have happened

If a promotion ran and sales went up, how much of that increase would have happened anyway? That's the question incrementality answers. Incremental sales are the ones a promotion, display or campaign actually caused, measured against a credible estimate of what sales would have been without it. Everything else is sales you'd have had regardless.

The gap between total sales during an event and incremental sales is often large. Teams that report the first as if it were the second end up overfunding the wrong activities.

What is a baseline, and why does it decide everything?

A baseline is your estimate of what would have sold without the activity. Incremental sales equal actual sales minus the baseline. Get the baseline wrong and every conclusion built on it is wrong too.

The simplest baseline is the average of the weeks just before the event. It's easy, and it ignores seasonality. If a promotion on cold drinks starts in the first warm week of spring, a pre-period baseline will credit the promotion with the weather.

A stronger baseline uses control stores: similar stores that didn't run the activity during the same weeks. Whatever happened in the control stores, such as the season, the calendar or a local event, also happened in the test stores. The difference between the two groups is a much cleaner read of the activity itself.

How do you calculate incrementality with control stores?

Here's a hypothetical. A brand runs a four-week promotion in 25 stores. Before the promotion, those stores sold 1,000 units a month. During it, they sold 1,200. A naive read says the promotion added 200 units.

Now add controls. Twenty-five matched stores without the promotion sold 1,000 units a month before and 1,080 during the same weeks. Something lifted sales by 80 units in both groups. Subtracting that shared lift leaves about 120 units attributable to the promotion.

That 120 is the incremental figure. It's less exciting than 200, and it's far more useful for deciding whether to run the promotion again.

What else reduces the incremental figure?

Two things, mainly.

The first is the post-promotion dip. If shoppers stocked up during the deal, sales in the weeks after may fall below baseline. Those lost weeks should be subtracted from the promotion's lift. A promotion that pulls sales forward without adding new ones looks great during and poor overall.

The second is cannibalization. If the promoted item took sales from other items in your own line, the brand's gain is smaller than the item's. Measuring incrementality at the brand level, and not only the item level, catches this.

What would you do with this on Monday?

Pick the next promotion on your calendar and set up the measurement before it runs. Choose control stores in advance, matched on size, location type and recent sales of the item. Write down the metric you'll use and the weeks you'll include, including a few weeks after the event.

If a test isn't possible for a given event, at least compare against the same weeks last year in the same stores, adjusted for any change in store count. It's weaker than a control group, and much better than a pre-period average alone.

Then review past promotions with the same lens if you have the data. You may find some that looked strong on gross lift and weak on incremental sales. Those are candidates to redesign or drop. The testing logic is the same one behind A/B testing, and the broader topic of sales promotion gives useful background on the mechanics. Category context from NRS Insights helps you judge whether the whole category was moving at the time.