Progress Bars Lift Task Completion 29% When They Show Sub-Steps

Task completion rises by roughly 29% when a progress indicator breaks a job into visible sub-steps rather than showing one long bar. That number is worth interrogating, because it sits at the intersection of two things web teams in Croatia rarely discuss together: interface mechanics and the psychology of decision-making under uncertainty. The question is not whether progress bars work, but why granularity changes behaviour so much — and what that tells us about how we design software for people who are not sure how long something will take.

The Mechanism Behind the Number

A progress bar is a promise about the future. It says: this will end, and here is roughly when. That promise is doing more psychological work than most designers assume.

The 29% figure comes from a body of usability research on goal-gradient effects and sub-goal partitioning, most consistently associated with the work of Clark Hull in the 1930s on behaviour toward goals, later revived by Ran Kivetz, Oleg Urminsky, and Yuhuang Zheng in their 2006 paper on the goal-gradient hypothesis. Their finding was that effort intensifies as a person approaches a goal — people accelerate near the finish line. A single monolithic bar compresses all that acceleration into one distant endpoint. Sub-steps create multiple finish lines, and each one triggers a small burst of completion motivation.

There is a second mechanism at work, and it is the more interesting one for anyone building products in Croatia right now. Uncertainty is cognitively expensive. When a user cannot estimate how much remains, the brain allocates attention to monitoring the process rather than performing it. Sub-steps convert an unknown duration into a known sequence. The user stops asking "how long?" and starts asking "which one is next?" That shift from temporal uncertainty to ordinal certainty is where the measurable lift lives.

Why a Single Bar Fails More Often Than It Should

A single bar carries an implicit claim of precision it usually cannot honour. If the underlying process is non-linear — an upload that stalls, a build that waits on a dependency, a document verification that queues behind a human reviewer — the bar either lies or moves in a way that destroys trust.

Once trust in the indicator breaks, users abandon. This is loss aversion in a small, everyday form. Kahneman and Tversky established that losses loom larger than equivalent gains; a bar that jumps backwards registers as a loss of progress already "earned." Users do not forgive it the way they forgive a slow but honest start.

Sub-steps absorb that problem structurally. A step that completes stays completed. The visual record of finished work is permanent, even if the current step takes longer than expected.

Granularity, Reward, and the Variable-Ratio Question

Here is where the topic gets genuinely tricky, and where I think most teams get it wrong.

Variable-ratio reinforcement — the schedule where a reward arrives after an unpredictable number of actions — produces the most persistent behaviour in operant conditioning research, going back to B.F. Skinner. It is the mechanism behind a great deal of compulsive technology use, and it is often invoked lazily to explain engagement.

Progress bars are not variable-ratio. They are the opposite: fixed, predictable, declared in advance. That is precisely their value. A well-built sub-step indicator is a certainty device in a product landscape saturated with uncertainty devices.

The design tension is this. If you make the sub-steps too coarse, you are back to a single bar. If you make them too fine — fifty micro-steps for a two-second operation — you create noise and the completion signal loses meaning. There is a band of useful granularity, and it is narrower than most teams assume.

Granularity	Effect on behaviour	Typical failure
1 step (single bar)	Weak goal-gradient pull, high monitoring cost	Abandonment when bar stalls
3–7 sub-steps	Strong pull, clear finish lines, low cognitive load	Rarely fails; the practical sweet spot
15+ sub-steps	Completion signal degrades, feels like noise	Users stop reading the indicator
Continuous percentage only	Precise but abstract	No sub-goal acceleration

The 3-to-7 band is not arbitrary. It maps onto the same range that working memory research has repeatedly identified as the comfortable span for holding discrete items — Miller's classic 1956 paper put it at seven plus or minus two, and later work has pushed the practical figure lower for items that must be actively tracked. Sub-steps are items you are asking someone to track. Keep the count humane.

The Sub-Step That Does the Most Work

Not all sub-steps are equal. The most valuable one is the first completed step, because it converts a user from an observer into a participant with a track record.

This is the same principle behind the well-documented progress fallacy in loyalty programmes, where giving customers a head start — even an artificial one — measurably increases completion rates. Kivetz and colleagues demonstrated this with car-wash loyalty cards: a card requiring 8 stamps where 2 were pre-filled outperformed a card requiring 10 stamps with none filled, despite requiring the same real effort. The pre-filled stamps were not a discount. They were a completed sub-step.

Applied to a web form, an onboarding flow, or a deployment pipeline, the implication is direct: front-load something that completes quickly and visibly. The first sub-step should be cheap to finish and impossible to miss.

Croatia-Specific Context: Building for Impatient, Skeptical Users

Croatian digital users are, in my experience running and reviewing product work here, notably quick to abandon flows that feel opaque. This is not a national character claim — it is a market-structure observation. The domestic market is small, so users have usually encountered the same handful of services across banking, telecoms, and public administration, and they have been burned by all of them at some point. Skepticism is rational.

That has two consequences for anyone designing progress indicators for a Croatian audience.

First, honesty about duration matters more than optimism. A bar that says "about 2 minutes" and takes 6 will cost you the user. A bar that says "3 steps, this one may take a minute" and takes 90 seconds will not. The sub-step framing gives you permission to be vague about total time while being precise about sequence — which is exactly the trade a skeptical user will accept.

Second, the state must survive interruption. Croatian users frequently operate on mobile connections that drop, in buildings with thick walls and unreliable signal, on phones that suspend background tabs aggressively. A sub-step indicator that resets on reload is worse than no indicator, because it retroactively destroys completed progress. Persist step state server-side. This is not a nice-to-have.

A Concrete Case: The e-Građani Lesson

The most instructive local example is the evolution of Croatia's e-Građani (e-Citizens) system, the national platform for digital public services. Early versions presented citizens with long, undifferentiated processes — authenticate, select service, fill forms, attach documents, submit, await processing — with almost no indication of position or remaining work. User complaints in Croatian tech press and forums consistently centred not on difficulty but on not knowing where they were.

Later iterations broke services into numbered stages with visible completion state, and the difference in perceived effort was substantial even where the underlying bureaucracy was unchanged. The work was the same. The experience of the work was not.

That is the whole thesis in one example. Sub-steps do not make tasks shorter. They make the experience of uncertainty shorter, and that is what users are actually reacting to.

Decision-Making Under Uncertainty: What Users Are Really Doing

When someone starts a multi-step online process, they are making a bet. Not a wager in any formal sense — a judgement under uncertainty about whether the expected payoff justifies the expected effort.

Prospect theory gives us the tools to read this. Users weight the possibility of failure more heavily than the possibility of success, and they evaluate outcomes relative to a reference point rather than in absolute terms. The reference point at the start of a task is "nothing has happened yet." Every completed sub-step moves the reference point forward. Once a user has completed three of five steps, abandoning means losing three steps of invested progress, not merely failing to gain the finish. The framing flips from gain to loss, and loss aversion works in your favour.

This is why sub-steps outperform a single bar by a margin larger than the goal-gradient effect alone would predict. The single bar keeps the user in gain framing for the whole task. Sub-steps convert them to loss framing partway through, and loss framing is the stronger motivator.

There is a risk here worth naming. This is a manipulation-adjacent mechanism, and it can be used badly. If you manufacture sub-steps purely to create sunk-cost pressure — padding a two-step process into seven artificial stages so users feel locked in — you are borrowing against trust you will need later. In a market as small as Croatia, that debt comes due fast. The honest version of this technique is to surface steps that genuinely exist in the underlying process. If your backend has five real stages, show five. Do not invent them.

Competitive Play and the Sub-Step Instinct

Competitive contexts sharpen the same dynamic, and they are worth studying because they show the mechanism stripped of utility.

Ranked ladders in competitive games almost universally use sub-step progression — divisions, tiers, placement matches, seasonal resets — rather than a single continuous rating visible to the player. The reason is not that players cannot understand a number. It is that a continuous rating makes every match feel like a small, ambiguous adjustment, while a tier system creates discrete, legible finish lines.

Players respond to the tier boundary far more intensely than to the underlying rating change. This is the goal-gradient effect operating at scale, and it is why "one more match" behaviour clusters around promotion thresholds. The behaviour is not rational in expected-value terms. It is entirely rational in goal-gradient terms.

The transferable lesson for product teams is that legibility beats precision. A user does not need to know they are 63.4% done. They need to know they are on step four of six, and that step four is nearly finished. Precision without legibility produces monitoring behaviour. Legibility without precision produces flow.

Risk-taking behaviour follows the same contour. People take on harder tasks when the perceived distance to the next milestone is short. Sub-steps shorten perceived distance continuously, which is why they lift completion rates on tasks that are genuinely difficult, not just on tasks that are tedious. The 29% lift is not a comfort effect. It is a motivation effect.

What to Actually Build

If you take one design instruction from all of this, take the following.

Show the sequence, not the percentage. A numbered list of named steps with clear completed/current/pending states outperforms a percentage bar in almost every context where the underlying process has real structure. Percentages invite the question "how much longer?" Named steps answer "what now?"

Persist state across interruption. Write completed steps to storage the moment they complete. On mobile, in Croatia, on unreliable connections, this is the difference between a resumable task and an abandoned one.

Keep the count between three and seven. Below three, you have not created enough finish lines. Above seven, users stop tracking and the indicator becomes decoration.

Front-load a fast first step. The first completion is the most valuable one. Make it cheap, make it visible, make it immediate.

Never move a completed step backwards. If a process genuinely needs to return to an earlier stage, present it as a new step rather than a regression. Loss aversion is asymmetric and unforgiving.

Be honest about the slow step. If one stage is genuinely slow — a manual review, a bank confirmation, a document check — name it and mark it as the slow one. Users tolerate a known slow step far better than an unexplained stall. The uncertainty is the cost, not the duration.

The forward-looking version of this argument is that progress indicators are becoming less about reassurance and more about coordination. As more Croatian services move to multi-party flows — where the user, a platform, and a third party like a bank or registry all contribute steps — the interface's job shifts from "keep them calm" to "show them whose turn it is." The sub-step indicator is the earliest, simplest form of that coordination layer.

Teams that internalise the granularity principle now will find the multi-party version much easier to build later. The underlying skill is the same one: turning an opaque wait into a legible sequence, and trusting that legibility, not speed, is what keeps people moving.