The Dashboard Is Winning. Is the User?

The number can be correct and still tell an incomplete story

Duolingo defines its streak as consecutive days on which a learner completes a lesson. Its help center describes the feature as a way to encourage regular practice. That is a coherent purpose: practicing regularly matters to language learning, and continuity is something a product can observe.

GOODSCOOL JOURNAL / FIG. 01

Read full figure transcript

FIELD STUDY / 01 GOODScool · ORIGINAL TEACHING EXAMPLE What a streak can show—and hide THE DASHBOARD SEES 7 days Current learning streak M T W T F S S Recorded activity: yes THE PRODUCT STILL NEEDS TO ASK Learning quality ? Can the learner explain the idea? Can they apply it to a new problem? Did the streak prompt rush the work? These questions need their own evidence. Illustrative dashboard: seven consecutive days is a count of recorded activity, not a direct measure of comprehension or useful progress. This example is schematic and contains no usage data from GoodScool or any other product.

But completing a lesson is not the same measurement as remembering a new word a week later, understanding a conversation or confidently speaking with another person. A streak can be perfectly accurate about attendance without answering those questions.

This distinction matters because the visible measure can become the user's immediate goal. The learner may return because they want to learn. They may also return because they do not want to lose a counter. Both motivations can coexist; neither should be assumed from streak data alone.

Product teams face the same ambiguity. More daily visits might reflect genuine usefulness. They might reflect a successful reminder campaign, additional friction in a task, or a habit loop that has become separate from the original job. The number records an event. It does not independently explain the person's reason for being there.

Measurement can change the thing being measured

There is experimental evidence for a tension that many product teams recognize intuitively. In six experiments published in the Journal of Consumer Research, Jordan Etkin found that measuring enjoyable activities could increase how much people did while simultaneously reducing how much they enjoyed them. The proposed mechanism involved shifting attention toward output and away from the activity's intrinsic appeal. The experiments included activities such as walking and reading; they do not establish that every streak, scoreboard or productivity dashboard has the same effect. Etkin, The Hidden Cost of Personal Quantification.

The lesson is not that measuring is harmful. Measurement made people do more in those experimental settings. That matters when the activity is difficult to sustain or when consistency has genuine value.

The more useful lesson is that a metric can be both helpful and costly, depending on which outcome we care about. A reading app could plausibly increase completed pages while making some readers feel obliged to finish rather than enjoy or understand. That example is a hypothesis to test, not a documented result for a specific app.

The challenge for product teams is therefore bigger than choosing a better number. They must notice when displaying the number changes the behavior that generates it.

The product team is part of the same feedback loop

Imagine a subscription learning product reviewing its weekly numbers. Return visits are improving, and a daily-goal feature appears to be popular. A team could reasonably interpret those signals as reasons to invest further.

Now imagine its next decisions: make the goal easier to maintain, give it more prominent placement, add more reminders and celebrate each completed session. These are proposed decisions in a hypothetical product, not an account of an experiment or its results.

Each change might make the tracked behavior more likely. But the questions become sharper: Are users returning to practise something difficult? Are they becoming more capable? Do they feel in control of their time? Are reminders helping them or interrupting them?

An analytics dashboard cannot answer those questions merely by showing a higher return rate. The team has begun designing around a signal, while the experience it wanted to improve may be drifting out of frame.

This is the central risk of metric-led product design: the proxy becomes easier to optimize than the underlying human outcome. And once that proxy influences both interface decisions and user incentives, the loop can reinforce itself.

GOODSCOOL JOURNAL / FIG. 02

Read full figure transcript

FIG. 01 / EXPLAINER ILLUSTRATIVE / ORIGINAL GOODScool ARTWORK Figure 1. How a visible score can shape the next action. Illustrative mechanism; no platform-specific behavioral outcome is asserted.

The metric is not the mission

Economist Charles Goodhart warned that a statistical relationship may become less reliable when it is put under pressure as a means of control. The insight originated in economic policy, not app design, so using it here is an analogy rather than proof about a particular product. Goodhart, 2006.

The familiar product version is simple: when teams are rewarded for one number, they may produce more of the behavior that number recognizes—even if it is only part of what users value. When users see that same number displayed as a target, they may adapt too.

That is not a reason to abolish KPIs. A product that refuses to measure whether it helps anyone is not more humane; it is simply harder to evaluate.

Google researchers Kerry Rodden, Hilary Hutchinson and Xin Fu developed the HEART framework to connect user-centered goals to signals and metrics. Its categories include happiness, engagement, adoption, retention and task success. It is a useful reminder that engagement is one dimension of experience, not a synonym for quality.

If a user visits a banking app less often because a payment problem is finally fixed, fewer sessions might accompany a better experience. If someone completes a hard course rather than repeatedly opening its dashboard, time spent in the dashboard may fall as capability grows. These are illustrative possibilities, not claims about actual banking or learning products.

The right metric depends on the user's job, not the team's ability to generate a chart.

Give every important metric a contract

Before adding a prominent score or celebrating a movement in a KPI, write down what the measure promises—and what it cannot promise. A metric contract can be short enough to fit into a product review.

Consider a hypothetical learning app:

This is an original GoodScool teaching tool, not a validated measurement framework or the result of an actual field experiment. Its purpose is to make product judgment inspectable.

The most revealing part is usually the fifth row: what it cannot establish. A team may know exactly how a score is computed yet have no shared agreement about which larger outcome it represents.

It also matters where the number appears. A private progress indicator helps a user reflect differently from a leaderboard comparing them with strangers. A reminder may support a user-selected commitment; a warning that implies failure may pressure them to protect a score even when their circumstances change. Those differences are design hypotheses worth researching, not effects that can be inferred from a screenshot.

There is a case for streaks, badges and visible progress

Some goals are hard to pursue without structure. A small visible achievement can make effort feel less abstract. Progress indicators can also tell people where they stand and what remains.

Research on gamification does not support treating every badge or leaderboard as either universally effective or universally manipulative. An experimental study by Michael Sailer and colleagues found that different game elements affected different aspects of psychological need satisfaction. The design elements and context mattered. Sailer et al., How Gamification Motivates.

So the design question is not whether a streak should exist. It is whose goal the streak serves, how much authority it receives and what choices the user retains.

For a learning product, that might mean letting people choose their practice commitment, seeing progress without being publicly ranked, pausing or adjusting a goal without disproportionate punishment, and finding evidence of competence beyond attendance. These are design options to evaluate, not features proven here to increase learning.

A useful measure should make the user's original goal easier to pursue. It should not quietly replace that goal with the maintenance of the measure itself.

What to ask at the next dashboard review

Choose one number your product team watches closely: daily active users, streak continuation, lessons finished, sessions per user, messages sent or time spent.

Do not start by asking how to increase it. Ask three questions first:

1. What did the person come here to accomplish? State the job in their terms, not as an analytics event.
2. What behavior does this number reward? Include the easiest way a person could improve the score without improving their actual outcome.
3. What would make us reconsider it? Identify a second signal, a qualitative check or a possible harmful effect that would challenge the interpretation.

Then inspect a few real journeys—with appropriate consent and privacy safeguards—before proposing a nudge or redesign. If the number and the experience disagree, treat the disagreement as information, not as an inconvenience to explain away.

Return to the learner at 11:57 p.m. A streak can honestly say they showed up. That may be encouraging, and it may be valuable. The more consequential questions remain: what did they understand, what can they do now, and did the product help them learn on their own terms?

A good product knows what it counts. A better product remembers why it started counting.