Progress Bars Beat Badges on Task Completion in 640 Trials

A progress bar and a badge can both be described as feedback. But one tells you where you are in a task; the other tells you what you have already done. That difference matters more than most interface design discussions admit. A recent line of experimental work — 640 trials across several task types — found that a visible progress bar produced higher completion rates than a badge system, even when the badge was awarded at the same point in the sequence. The question worth asking is not which decoration users prefer, but what each element does to the way a person represents the remaining work in their head.

What a Progress Bar Actually Communicates

A progress bar is a ratio made visible. It answers a question the mind keeps asking during any extended task: how much is left? That question is not idle. It is the input to a cost-benefit calculation that runs continuously, and it is sensitive to framing. In their work on goal gradients, Clark Hull and later Kurt Lewin observed that the pull toward a goal strengthens as the distance to it shrinks. A progress bar externalizes that distance, and it does so in a form that updates.

Contrast this with a badge. A badge is a discrete token awarded after a threshold is crossed. It is binary: earned or not earned. Between thresholds, the badge provides no information about proximity. The user is either inside a rewarded interval or outside one, and the interface offers no signal about how far into the interval they have traveled. In the 640-trial comparison, this informational silence is the most plausible explanation for the completion gap. A badge says keep going and something will happen. A progress bar says you are 40% through, and here is what 40% looks like.

The Endowed Progress Effect

The most cited evidence for the bar's advantage comes from Joseph Nunes and Xavier Drèze, who studied loyalty-card completion in a car wash setting. Customers given a card requiring ten stamps, but issued with two stamps already applied, completed at roughly double the rate of customers given an eight-stamp card with no head start. Both groups needed eight purchases. The difference was that one group could see itself as already underway. A progress bar reproduces this effect structurally: it never starts at zero in the user's perception, because the first increment appears the moment they act. A badge system, by contrast, begins with an empty shelf, which is a different psychological starting position.

Reward Schedules and the Shape of Persistence

Behavioral psychology offers a useful distinction here, and it cuts against the instinct to make rewards frequent. Variable-ratio reinforcement — the schedule in which a reward arrives after an unpredictable number of responses — produces the most persistent behavior in laboratory animals. B.F. Skinner's pigeons pecked longest under intermittent, unpredictable payoff. This is often cited in discussions of engagement design, and it is worth being precise about what it implies.

Intermittent reinforcement sustains behavior that is already established. It is less effective at establishing behavior in the first place, and it is notably poor at supporting tasks that require sustained attention to a defined endpoint. A progress bar is closer to a fixed-interval or continuous signal: predictable, proportional, and legible. For a task with a known end — finishing a form, completing a reading, working through a checklist — predictability outperforms surprise. The 640-trial result is consistent with this: users completed more tasks when the feedback was continuous and proportional than when it was discrete and celebratory.

Where Loss Aversion Enters

Kahneman and Tversky's work on loss aversion suggests a second mechanism. A partially filled progress bar represents accumulated effort that would be forfeited by abandoning the task. The bar makes that investment visible, and visible investment is harder to walk away from than invisible investment. A badge does something similar only at the moment of award. Before the badge is earned, there is nothing to lose, because nothing has been represented as owned. After it is earned, the badge is a possession, but the task is already complete. The bar operates during the window when abandonment is still a live option, which is precisely when the intervention is needed.

The Design Tension: Precision Versus Motivation

There is a counterargument worth taking seriously. Progress bars can be demotivating when they move too slowly, and they can be gamed when the denominator is arbitrary. A bar that sits at 12% for a long stretch communicates effort without communicating payoff. Some researchers have found that perceived slowness in a progress indicator reduces persistence more than the absence of an indicator altogether. The bar is not magic; it is an information channel, and like any channel it can carry bad news.

The 640-trial comparison does not settle this. It establishes that a bar outperforms a badge on completion, not that every bar outperforms every badge. The practical lesson is about granularity. A bar that advances in visible increments — even small ones — preserves the proximity signal that drives the goal gradient. A bar that stalls does not. Badges, meanwhile, retain one advantage the bar lacks: they are shareable and socially legible in a way that a percentage is not. For tasks where the outcome matters to other people, the badge's discreteness is a feature rather than a bug.

Reading the Evidence Into Practice

For anyone designing a sequence of tasks — a course module, a research protocol, a reading plan — the implication is to instrument the middle of the task, not just the ends. Badges cluster at completion. Progress bars live in the interval where people actually quit. That interval is where the behavioral evidence points, and it is where most designs are thinnest.

The forward-looking question is not whether to use bars or badges but how to make the bar honest. A bar whose denominator is real, whose increments are visible, and whose endpoint is defined gives the user something a badge cannot: a continuously updated estimate of the distance remaining. The next useful experiment is not another head-to-head comparison. It is a test of bar granularity — how fine the increments must be before the proximity signal stops doing its work, and whether that threshold shifts across task types and populations. That is a question the current literature has barely begun to answer, and it is the one that would actually change how these interfaces get built.