Notes / Conversion Works
Did your discount create purchases—or just reduce the price?
Evaluate discount incrementality with a defined comparison, contribution per assigned customer, and follow-up that accounts for returns and later purchases.
By Edu Rigonato. Published Oct 14, 2026. 10 min read.
The coupon report looks encouraging. Orders arrived, customers used the code, and the campaign received credit for revenue. You can see that the offer was used. What you cannot see in that report is how much those same customers would have bought without it.
That missing comparison changes the decision. An accessory store might subsidize an order that was already going to happen. A promotional message might also bring someone back who buys without entering the code. Counting only redemptions can miss one effect while overstating another.
To assess discount incrementality, compare eligible customers assigned to a defined offer with an appropriate comparison group. Measure all their relevant purchases and consistently defined contribution, including people who never order. Then allow enough follow-up to see the later effects that matter to the decision.
This does not make discounts inherently good or bad. It gives you a way to decide what a particular offer accomplished before making it more generous.
What does a redeemed discount actually tell you?
Redemption tells you that an order used a code. It does not identify the order’s individual counterfactual: what that person would have done without the offer. You cannot observe both versions of the same purchase at the same moment.
Imagine a fictional accessories merchant sending a promotion to customers eligible for a replacement travel pouch. Some already need one. Others might browse because of the message. Others may not buy under either condition. The checkout does not label which situation produced each order.
This is why the denominator matters. If you analyze only coupon users, you select people who responded to the treatment. Customers who ignored the message or bought without the code disappear, even though their behavior belongs in the commercial assessment.
Frequently asked questions
Can I compare coupon users with customers who paid full price?
You can describe those groups, but the comparison does not isolate the coupon’s effect. People choose whether to redeem, and their needs or purchase readiness may differ. For a causal test, preserve the original eligible-group assignment and include nonpurchasers in the main outcome.
Does a no-email holdout isolate the discount?
It usually answers the combined campaign question when the other group receives a message and an incentive. To isolate the added incentive, design a comparable-message contrast. Document other differences, including subject lines and creative, rather than assuming the discount is the only changed element.
Should the discount be a separate cost in contribution?
Not if the revenue figure already deducts it. Define net revenue and variable costs explicitly, including the treatment of refunds, returned goods and shipping. Apply identical accounting rules to each arm and avoid counting the same reduction twice.
How long should the test run?
Separate enrollment time from follow-up. Enrollment must support the needed comparison; follow-up must cover the outcomes relevant to the decision. Returns and repeat purchases may require longer observation. A fixed number of days borrowed from another store cannot resolve that choice.
What if the store cannot support a credible holdout test?
Use descriptive results to identify questions and constraints, and label them as descriptive. You can still reconcile costs, inspect offer overlap and improve measurement. Avoid presenting a before-and-after promotion total as a causal estimate simply because a randomized comparison is unavailable.
Explore Conversion Works