Separating discount effects from advertising effects

To estimate the discount effect separately from advertising, plan comparable groups where price and media vary independently. When everyone receives both, neither a sale-week sales change nor an ad-attributed total identifies their separate effects.

Define the question and outcome

Define the question and outcome

Choose the product set, customers or locations, dates and outcome before the sale. Units, net sales and contribution answer different questions. Record the actual price, offer conditions, media exposure and availability for each group.

Decide whether you need the effect of advertising during the discount, the effect of discounting without campaign ads, or whether advertising works differently at the discounted price. Keep attributed-sales reports separate: attribution credits purchases under its rules, while incrementality concerns outcomes caused by marketing against a credible comparison.

Assess a four-cell design

Assess a four-cell design

Where the retailer can implement it fairly and lawfully, four planned combinations can address all three questions:

Within the discounted-price column, the difference between ad and no-ad groups estimates the advertising effect during the promotion, if assignment and execution make the groups comparable. Within the no-ad row, the difference between price groups estimates the discount effect without those ads.

Comparing the two advertising effects estimates whether price changes the effect of media. If they interact, there is no single, context-free share of the combined result to assign to each.

Choose an assignment unit that limits spillover between groups. Check baseline demand, other campaigns, stock, local conditions and whether customers could see media or prices intended for another group.

The creative must accurately describe the price available to its audience. Do not create inconsistent or misleading consumer offers merely to complete a test grid. Confirm whether independent price and media variation is possible before committing to this design.

Use a narrower comparison when necessary

Use a narrower comparison when necessary

If the discount must run everywhere, compare suitable groups with and without campaign ads. With sound assignment, exposure control and comparable conditions, this estimates the ad effect under the common discount. It does not identify the discount’s own effect. A price comparison while media remains common answers a different question.

A before-and-after chart can show timing but cannot isolate either cause on its own. Demand, competitors, other media and availability may also change. A model-based analysis may be useful if its assumptions, covered population and uncertainty are made clear; it does not turn an uncontrolled comparison into an experiment.

Check execution before reporting

Check execution before reporting

Verify that each group received its intended price and media. Record stock gaps, creative changes, overlapping activity and any sales or exposure that crossed group boundaries. Present the outcome, comparison method and uncertainty together.

If the planned separation failed, report observed promotional sales and ad-attributed purchases on their stated bases. Do not assign a causal share to the discount or advertising without a credible counterfactual.

Key Considerations for Valid Incrementality Testing