A post-purchase survey is usually treated as measurement. It’s also an intervention. People who are asked what they intend to do become a little more likely to do it, which is useful for first use and a problem for anyone reading the survey as a neutral count.

The findings

Vicki Morwitz, Eric Johnson and David Schmittlein found in 1993 that asking people once whether they intended to buy a car or a personal computer made them more likely to buy one. Asking repeatedly made low-intent people less likely to buy Published. Utpal Dholakia and Morwitz then followed customers of a financial services firm who had been asked to take a satisfaction survey. Compared with customers who weren’t asked, they later bought more, defected less and were more profitable, and the effect grew for months and was still there a year later Published.

How big it really is

The field calls this the question-behavior effect, and like much of behavioral science it looked bigger in its early years. A 2016 meta-analysis by Chantelle Wood and colleagues pooled 116 published tests and found a small effect (d = 0.24). Corrected for the likelihood that null results went unpublished, the estimate fell to d = 0.15. The effect was larger for behaviors that are easy and socially desirable, and in student samples Published.

Using a product you’ve already bought is easy and desirable, which is the favorable case. So expect a nudge, not a transformation, and don’t design a survey around a big effect you haven’t measured.

A survey is also a nudge. Hold some buyers back, or you can’t tell how much.

Using it well

• Ask about use, not satisfaction. “Have you used it yet?” measures time to first use (chapter 3). A follow-up for those who haven’t, “When do you plan to?”, doubles as the if-then prompt.
• Ask once. Morwitz’s finding on repeated questioning is a warning. A buyer who says “not yet” twice doesn’t need a third survey; they need a first move and a reason it’s easy.
• Keep a holdout. Send the survey to a random 80% and compare use and reorders with the 20% who didn’t get it. Otherwise you can’t separate what you learned from what you caused.
• Only ask what you’ll use. A survey that exists only to nudge, with answers nobody reads, is manipulation dressed as research.

Do this

• Rewrite your post-purchase survey to lead with the use question from chapter 3, add the plan question for “not yet” answers, and set the 20% holdout before it goes live.