Progress Bars and the Zeigarnik Problem
The famous unfinished-task effect is shaky; the endowed progress research is stronger.
Drumlin, a compliance tool for care homes, adds a profile completeness bar. It sits in the sidebar and reads 40% complete, with a list of what is missing. Completion climbs from 34 percent to 51 percent in a month. Six weeks later, support notices something: people are filling fields with placeholder text to clear the bar. The staff directory now contains nine entries named TBC.
The bar worked on the number and not on the goal. To understand why, it helps to separate two ideas that usually get mashed together under one name.
What a progress indicator has to do
Zeigarnik, and the trouble with it
Bluma Zeigarnik reported in 1927 that waiters recalled unpaid orders better than settled ones, and that participants interrupted mid-task remembered those tasks better than completed ones. It is a lovely story and it is cited constantly in product writing. It is also one of the findings where you should be careful. Replication attempts over the following decades produced mixed and often null results, and the effect appears sensitive to how motivated participants were and how the interruption was staged. Treat it as a suggestive idea with an uncertain evidence base, not a lever.
The stronger body of work is the goal gradient: Effort rises as a goal gets closer, which is why progress bars work.. Clark Hull described in the 1930s that rats ran faster as they approached food. Kivetz, Urminsky and Zheng revisited this with humans in 2006 in the Journal of Marketing Research, using coffee cards. Customers bought coffee at an accelerating rate as they got closer to the free drink. In the same line of research, Nunes and Drèze published the endowed progress effect in 2006: a car wash loyalty card requiring 8 stamps, handed over with 2 already stamped, was completed more often than a card requiring 6 from zero. Identical work remaining, different completion rates.
What the goal gradient work implies for a progress indicator
- Start above zero
- If the user has already done something real (signed up, verified an email, connected a calendar), count it. A bar that opens at 0% after they have done work is both demotivating and inaccurate.
- Make the remaining steps countable
- A percentage with no visible list is a mystery. 3 of 7 with the three named is a plan. The acceleration effect needs a visible finish line.
- Only count steps that matter
- Every item on the bar is a promise that doing it improves the user's outcome. Padding the list with low-value fields is how you get nine people called TBC.
A bar that measures your form
- Profile 40% complete
- Add a profile photo to reach 50%
- 6 fields remaining
A bar that measures the user's readiness
- Ready for your March inspection: 3 of 5
- Add 2 missing staff DBS dates so the Safeguarding report can run
- Once medication records are linked, we can generate the CQC summary
The right column ties each remaining item to a consequence the user already cares about. That is what stops placeholder answers. Somebody typing TBC into a DBS date field knows they are breaking their own inspection report, and there is no reward for clearing the bar because the bar is not the point.
Worked example
Drumlin's rewrite
They delete the percentage. The sidebar now says: Inspection readiness, 3 of 5 checks passing, with two named failures and a link to each. Completion of genuinely useful fields rises more slowly than the old vanity number did. Placeholder entries stop appearing. When the team reports the change, they show both the old metric and the new one so nobody reads the slowdown as a regression.
There is a fair criticism to hold onto. Progress mechanics are the easiest thing in product to turn into manipulation, because they create an itch you control. BJ Fogg's behaviour model is worth a look here: he frames behaviour as motivation, ability and a prompt arriving together, and the honest use of a progress indicator is mostly on the ability side. It reduces uncertainty about what is left. The moment it is working mainly by discomfort, you have built a nag with a gradient.
Quick check
What did the car wash study show about a card needing 8 stamps with 2 pre-stamped?
The takeaway
Progress indicators earn their keep by making remaining work clear and consequential, not by creating an unfinished-task itch.
Try this tomorrow
Take your completion meter and rewrite every remaining item as a sentence naming what the user gets once it is done. Delete any item you cannot write that sentence for.
Answer the check above, then bank the day.
Where this comes from
- The Goal-Gradient Hypothesis Resurfaces (2006), Ran Kivetz, Oleg Urminsky and Yuhuang Zheng
- The Endowed Progress Effect (2006), Joseph Nunes and Xavier Drèze
- Tiny Habits, BJ Fogg
- Hooked, Nir Eyal
User Psychology is one of six tracks. These lessons summarise and build on the work above, they do not reproduce it. Buy the books, they are better.