---
name: Endowed progress effect
slug: endowed-progress
category: Motivation
categories: [Onboarding, Retention]
evidence: mixed
page: /principles/endowed-progress/
updated: 2026-10-13
---

# Endowed progress effect

> People are more likely to finish a task when they feel they have already started it.

## What it is

The endowed progress effect is the tendency to work harder toward a goal when some progress has already been made, even if the head start was given rather than earned.

## Evidence

The canonical source is a single paper with one field experiment and two scenario studies. Nunes and Drèze (2006) distributed 300 loyalty cards at random at a professional car wash on the second and third Saturday of April 2004. One hundred and fifty cards needed eight stamps for a free wash and arrived empty (0 of 8, 0% progress). The other 150 needed ten but arrived with two stamps already affixed (2 of 10, 20% progress), so both groups had to buy exactly eight washes. Customers were told the free stamps were part of a special promotion that day, and the paper notes that no customer inquired further. Over the next nine months, 80 cards were redeemed, all between 5 June and 17 September 2004, giving records for 720 visits. The redemption rate was 34% in the endowed condition versus 19% in the plain condition, chi-square(1) = 8.1, p < .01. Those rates imply 51 of 150 against 29 of 150, a difference of 14.7 percentage points; derived from those counts the 95% confidence interval runs from about 5 to 25 percentage points, a risk ratio of 1.76 with an interval of roughly 1.2 to 2.6. The effect is real in that sample and loosely estimated.

The same experiment measured speed. Customers with the endowed card took on average 2.9 fewer days between washes, F(1, 636) = 5.2, p < .05, and across both conditions the gap between visits shrank by 0.5 days with each additional wash purchased, F(1, 636) = 18.38, p < .01, with no significant interaction, F(1, 636) < 1. So the head start shifted the whole curve up rather than changing its slope: it is endowed progress plus an ordinary goal gradient, not a new kind of acceleration. The authors' own stated limitation is that they only observed effort for customers who completed all eight washes.

The paper's other two studies are hypothetical and tell you what the effect is not. Study 2 put 146 undergraduates through paper-and-pencil restaurant loyalty scenarios and crossed the cash value of the endowment with the share of the task it covered. Interest in the programme rose from 4.74 with no endowment to 5.55 with two credits and 6.26 with five credits, F(2, 143) = 11.91, p < .01. The dollar value of the endowment ($12.50 versus $30.00) did nothing, F(1, 72) = 1.25, p = .27, while relative progress (one third versus one sixth complete) did, F(1, 72) = 3.79, p = .05. The driver is the fraction of the bar that is filled, not the value of what you were given, which rules out a sunk-cost reading. Study 3 surveyed 240 shoppers at a liquor store, 30 per cell, and found two boundaries: when progress was tallied in purchases and no reason was given for the endowment, the endowment did nothing (4.24 versus 4.06 for no endowment, p = .25), and a specious reason (“because you are here today”) worked just as well as a real one based on purchase history (6.31 versus 6.19 in points, p = .48; 5.33 versus 5.20 in purchases, p = .42).

The useful independent support arrives from a different lab in the same year. Kivetz, Urminsky and Zheng (2006) ran a café reward programme and randomly gave 108 customers either a plain ten-stamp card or a twelve-stamp card carrying two bonus stamps, both requiring ten purchases. The endowed group completed the ten purchases in 12.7 days against 15.6 days, about three days or 20% faster, t = 2.0, p < .05, medians 15 versus 10 days, Z = 2.1, p < .05. Two details matter more than the headline. First, the head start did not make anyone more likely to join: in a parallel test with 65 train travellers, stated likelihood of joining was 5.9 with the bonus stamps and 5.8 without, t = 0.1, not significant. Second, a test with 118 travellers separated endowed from earned stamps and found people felt worse about losing a card whose two stamps they had earned (2.1) than one whose two stamps were a gift (1.7, t = 1.9, p < .05), with no difference between the gift card and an empty card. Endowed progress moves people because the bar looks fuller, not because they are protecting an investment.

The rest of that paper supplies the mechanism and the main failure mode. Across 949 completed ten-stamp cards covering roughly 10,000 coffee purchases, the gap between purchases fell by 0.7 days from the first interval to the last, t = 2.6, p < .05, a 20% acceleration, cutting mean card completion from a projected 29.4 days to an observed 24.6. But after the first reward was collected, purchase rates reset: the first two intervals on the second card (3.1 and 2.7 days) were slower than the last two on the first card (2.2 and 2.1 days) and statistically indistinguishable from the first two intervals on the first card (3.2 and 2.8 days), p < .01. Progress motivates, and then it unwinds the moment you hand over the reward.

Two field experiments, 20 years old, in two shops, with 108 and 300 participants, both by the authors who proposed the effect or a contemporaneous team, and no preregistered replication since. That is the honest state of the direct evidence, and it is why this belongs in the mixed column rather than the strong one. The direction is consistent and both field tests used real money and real behaviour, which is worth more than a dozen scenario studies; the magnitude is not something you should plan a roadmap around.

The digital analogue has been tested much harder and comes out worse. Villar, Callegaro and Yang (2013) meta-analysed 32 randomised experiments on progress indicators in web surveys. A constant progress indicator did not reduce drop-off at all, log odds ratio 0.072, p = .365, significant in only 2 of 18 studies, with meaningful heterogeneity (Q = 33.203, p = .011, I² = 48.8). Where a small completion incentive was promised, the indicator made things worse: log odds ratio 0.137, p = .159, rising to 0.157, p = .025 with two outliers removed, an increase in drop-off by a factor of about 1.17. Front-loaded indicators that move fast early and slow later — the closest thing to an endowment in an interface — were the only version that helped, and only weakly: not significant overall across 7 studies, and log odds ratio -0.212, p = .02 (odds of drop-off 0.80) once one outlier was excluded. Back-loaded indicators that start slow were clearly harmful, log odds ratio 0.447, p = .001, 1.56 times the odds of dropping off. The lesson is that a filled-in bar is not a motivator by itself. What the first part of the bar implies about the rest of the task is what people respond to.

## When it holds

- Relative progress, not absolute value. A head start worth $30 did no more than one worth $12.50; one third complete beat one sixth complete. Give a bigger share of a shorter goal, not a bigger gift.
- A stated reason, when progress is counted in the real unit. Endowing stamps or purchases with no explanation produced nothing. Endowing points worked without a reason, which Nunes and Drèze attribute to medium maximisation: an abstract currency makes both the gift and the return on effort look larger.
- Early, not late. Zhang and Huang (2010) report that at low progress people are asking whether the goal is attainable, and attributing that progress to themselves rather than to circumstances signals a harder goal and lowers commitment; at high progress they are asking whether the goal is worth it, and self-attributed progress signals value and raises commitment. Read against endowed progress, which is by construction situational, that puts the head start at the start of the journey and earned progress near the finish.
- A single, credible endowment. Nunes and Drèze flag that a recurring endowment may lose credibility the way a permanent discount does, and that an absurd gift (90 of 100 purchases) sets an unbelievable target. Both are untested.
- A goal that stays put. The acceleration in both field studies comes from the goal gradient. If the finish line moves, there is no gradient to ride.
- Likely absent or reversed when the endowment does not change the visible fraction remaining; when the head start is the only progress the user will see for a long time, so the bar front-loads and then stalls; when a completion incentive is in play and the indicator mainly tells people how much is left; and immediately after a reward, where both field studies predict a reset rather than momentum.

## How to apply

- Count work the user genuinely did and show it as done. Account created, email verified, payment method connected: these are completed steps, and starting the checklist at three of eight rather than zero of five is an endowment that needs no fiction.
- Express the head start as a fraction of the whole, because the fraction is what moves people. “2 of 10 done” or a bar filled to 20% carries the effect; “you received two free stamps” on its own does not.
- Say why the progress is there, in one line, next to it. The research is blunt that even a trivial reason works, which is the argument for using a true one: “because you signed up today” is both true and sufficient.
- Keep the remaining steps few and visible, so each completed one visibly shortens the distance. The 0.5 days per step acceleration in the car wash data is the goal gradient doing the work after the endowment starts it.
- Order the steps so the cheap ones come first. Front-loaded progress was the only progress indicator shape that reduced abandonment in the survey meta-analysis.
- Named patterns that carry this: profile completion meters, loyalty cards and point balances issued with a starting balance, and setup checklists with the signup steps pre-ticked.
- If progress is denominated in points rather than steps, expect the endowment to read as larger than it is. That is the medium maximisation finding, and it is a reason to be conservative with point grants rather than a licence to inflate them.
- Plan for the reset. After someone earns the reward, their rate drops back to beginner levels, so the next cycle needs its own starting state rather than an empty card.

## What to avoid

- Inventing completed steps. Crediting a user with work they never did is the anti-pattern: credit only steps the user genuinely completed. Note that the car wash study did not cross this line. The stamps were a real gift, openly described as a promotion; nobody was told they had already washed their car twice.
- Progress bars that move without work behind them. Villar and colleagues call the ethics of deliberately front-loaded indicators into question precisely because of their deceptive nature, and clarify that a varying-speed indicator is only acceptable when it mirrors the actual response process.
- A head start that papers over a long task. If the bar reaches 20% at signup and then does not move for twenty minutes, you have built the slow-to-fast shape that increased drop-off by 1.56 times the odds in the meta-analysis.
- Resetting progress on error. Clearing a half-finished form sends the user back to 0%, which is the one state the whole effect exists to avoid.
- Endowments that serve only the business. A head start toward a goal the user did not choose is a nudge toward your target, not theirs.
- In the EU, the legal line is the Unfair Commercial Practices Directive 2005/29/EC. Article 6(1) makes a practice misleading where false information, or a presentation that deceives the average consumer even when factually correct, is likely to cause a transactional decision they would not otherwise have taken; the main characteristics of the product, including the results to be expected, are explicitly covered. Annex I lists practices that are unfair in all circumstances, with no balancing test: item 31 prohibits creating the false impression that the consumer has already won or will win a prize or equivalent benefit when there is no such benefit or claiming it requires paying money, item 19 prohibits advertising a prize promotion and then not awarding the prizes described, and item 21 prohibits marketing material that gives the impression the consumer has already ordered something when they have not. A fabricated head start toward a reward that does not exist, or that quietly requires more purchases than the bar implies, sits squarely in Annex I territory.

## Measuring it

- Primary metric: completion rate of the goal per cohort, measured on the same real requirement in both arms. The whole point of the manipulation is that the work required is unchanged, so an A/B test that also changes the requirement measures nothing.
- Secondary: time to completion, and the interval between steps as a function of steps remaining. A genuine goal gradient shows intervals shrinking as the goal nears; if they do not, you have a progress display, not motivation.
- Backfire metric: drop-off at the first step after the endowed one. If the endowment buys a click and loses the session, the bar has front-loaded a task that then stalls.
- Second backfire metric: behaviour in the cycle after the reward. Both field studies predict a reset, so a programme judged only on first-reward completion will look better than the customer relationship actually is.
- Watch complaints and support contacts that mention the progress display. Users asking why they are at 20% is the signal that your reason is missing or not believed.

## Live examples

- World App opens the setup list at 75% with the three things signing up already did ticked off, so the one remaining step reads as finishing rather than starting.
- Cleo AI ticks “create your profile” the moment the profile exists, presenting a three-step setup as already a third done.
- Deel credits the four tasks signup already covered as done, so the admin meets the checklist with one open item instead of five.

## References

- Nunes & Drèze (2006). The Endowed Progress Effect: How Artificial Advancement Increases Effort. Journal of Consumer Research. — https://doi.org/10.1086/500480
- Kivetz, Urminsky & Zheng (2006). The Goal-Gradient Hypothesis Resurrected: Purchase Acceleration, Illusionary Goal Progress, and Customer Retention. Journal of Marketing Research. — https://doi.org/10.1509/jmkr.43.1.39
- Zhang & Huang (2010). How Endowed versus Earned Progress Affects Consumer Goal Commitment and Motivation. Journal of Consumer Research. — https://doi.org/10.1086/655417
- Villar, Callegaro & Yang (2013). Where Am I? A Meta-Analysis of Experiments on the Effects of Progress Indicators for Web Surveys. Social Science Computer Review. — https://doi.org/10.1177/0894439313497468
- Directive 2005/29/EC on unfair business-to-consumer commercial practices, Article 6(1) and Annex I (items 19, 21, 31). — https://www.legislation.gov.uk/eudr/2005/29/annex/I

## Similar principles

- Goal gradient effect — /agents/goal-gradient-effect.md
- Zeigarnik effect — /agents/zeigarnik-effect.md
- Sunk cost fallacy — /agents/sunk-cost-fallacy.md
- Labor illusion — /agents/labor-illusion.md
- Implementation intentions — /agents/implementation-intentions.md