The Endowed Progress Effect: Two Free Stamps Nearly Doubled Completion

In 2006, Joseph Nunes and Xavier Drèze ran an experiment at a car wash and published it in the Journal of Consumer Research. It remains one of the cleanest demonstrations in behavioural economics, because the manipulation was almost nothing and the effect was large.

The short answer

Three hundred customers received one of two loyalty cards: an 8-stamp card, or a 10-stamp card with 2 spaces already stamped. Both required exactly 8 purchases to earn a free wash. Over nine months, completion was 19% on the plain card and 34% on the pre-stamped one, a 78.9% relative increase from a change that cost nothing but ink.

Why identical effort produces different behaviour

People do not evaluate progress in absolute terms. They evaluate it as a fraction of a goal, and they are more motivated by a goal they have already started than one they have not.

On the plain card, a customer with zero stamps is at 0% of 8. On the pre-stamped card, the same customer is at 20% of 10. The remaining work is identical. The framing is not.

This connects to the goal gradient effect, first described by Clark Hull in 1932 and replicated in modern retail settings: effort accelerates as the goal approaches. If you can make someone feel they have already begun, you move them onto the part of the curve where motivation is rising.

It also connects to the Zeigarnik effect, Bluma Zeigarnik's 1927 finding that uncompleted tasks stay active in memory in a way completed ones do not. A card with two stamps on it is an open loop. A blank card is not yet a task at all.

The economics of the change

The experiment measured completion. Turning that into money requires a few assumptions, stated openly: a $10 wash at $3 cost of goods, and a free wash awarded on completion.

Margin per completed card is 8 paid washes at $7 contribution, minus the $3 cost of the free one, so $53.

Incremental completers: 22.5. Incremental margin: $1,192.50, which is $7.95 per customer enrolled.

N_PER_ARM = 150
RATE_PLAIN, RATE_ENDOWED = 0.19, 0.34
WASH_PRICE, WASH_COGS, PURCHASES = 10.0, 3.0, 8

margin_per_completer = PURCHASES * (WASH_PRICE - WASH_COGS) - WASH_COGS

for label, rate in (("plain", RATE_PLAIN), ("endowed", RATE_ENDOWED)):
    completers = N_PER_ARM * rate
    print(f"{label:>8}: {completers:>5.1f} completers, "
          f"margin ${completers * margin_per_completer:>9,.2f}")

incr = N_PER_ARM * (RATE_ENDOWED - RATE_PLAIN)
print(f"\nincremental completers {incr:.1f}")
print(f"incremental margin     ${incr * margin_per_completer:,.2f}")
print(f"per customer enrolled  ${incr * margin_per_completer / N_PER_ARM:,.2f}")

Output:

   plain:  28.5 completers, margin $ 1,510.50
 endowed:  51.0 completers, margin $ 2,703.00

incremental completers 22.5
incremental margin     $1,192.50
per customer enrolled  $7.95

The change pays for itself if printing the modified card costs less than $7.95 per customer. It costs a fraction of a cent.

One caveat worth stating: this counts completers only. Non-completers in both arms still bought washes, and the study design does not let us separate their spend cleanly. The true incremental figure is therefore somewhat lower than $1,192.50, because some of the extra completions came from customers who would have bought several washes regardless and simply finished the card. The direction and the rough magnitude hold. The precision does not.

Applying it without lying

The mechanism only works if the endowed progress is real progress toward a real goal. Three common implementations, in descending order of honesty:

The car wash version is the honest one. The customer really was given two stamps. Nobody was told they had done something they had not.

The dishonest variants work in the short run and fail specifically, because a user who discovers that a progress bar was fictional stops trusting every other number in the product.

Where to use it

Onboarding checklists are the obvious case. Mark the steps the user has genuinely already completed, including signing up, verifying an email and anything else that happened before they saw the list. Most products start these at zero even when the user has done three things.

Loyalty cards and punch cards, exactly as in the original study. If your card requires 8, print 10 and give 2.

Profile completeness, referral milestones and tier progress all work the same way. The rule is the same everywhere: give real credit for real actions, then let the goal gradient do the rest.