Measuring UX

From a goal to a metric

“How do I even measure this?” is rarely a metrics problem — it's a translation problem. Google's HEART framework names the five things a product experience can be good at; its Goals → Signals → Metrics process walks each one down from a fuzzy ambition to a number you can actually instrument. Pick a dimension and watch the descent. One rule before you start: choose the one or two dimensions that matter for this work — not all five.

Happiness

Attitudinal — how users feel about the experience.

Reach for Happiness when the question is about perception and satisfaction rather than what people did. It's the one HEART dimension you can't read from logs — it has to be asked.

Worked example — a redesigned onboarding.

Goal

Users feel confident and satisfied, not confused or frustrated.

New users feel confident they've set the product up correctly.

Signal

Users rate the experience well and say they'd recommend it.

Users rate the setup experience highly right after finishing it.

Metric

A satisfaction score — CSAT, a post-task rating, SUS, or NPS.

Post-setup CSAT ≥ 4.2 / 5, sampled in-product on the final step.

Watch out

Attitudinal data comes from surveys, so it carries sampling and self-selection bias — the delighted and the furious answer most. Track the trend across releases, not a single hero number, and never let a good satisfaction score stand in for whether people actually succeeded.

Engagement

Behavioral — the depth and frequency of interaction.

Reach for Engagement to measure whether people find ongoing value in the core experience. Most useful for habitual, discretionary products where coming back is a choice — less so for tools people are required to use.

Worked example — a note-taking app.

Goal

Users find enough value to interact often and go deep, not just show up.

People reach for the app to capture thoughts as a habit, not once a month.

Signal

Users return within a period and take more of the valuable action per visit.

The number of notes a person creates per active week climbs and holds.

Metric

A rate tied to the value action — actions per user per period, not raw time.

Average notes created per active user per week.

Watch out

More time in the product isn't automatically good — it can mean people are lost. Anchor engagement to a value-bearing action (a note created, a track played), not to time-on-screen or session length, which reward friction as easily as delight.

Adoption

Behavioral — new users or new uptake of a feature.

Reach for Adoption right after a launch or a feature ship, when the question is whether people are starting to use the thing at all. It counts first-time uptake within a window — the opening move, not the long game.

Worked example — a newly shipped export feature.

Goal

New users (or existing users meeting a new feature) successfully start using it.

Existing users discover and try the new export at least once.

Signal

People complete the key first action that counts as 'started'.

A user opens the export flow and successfully produces a file.

Metric

A count or rate of first-time completion within a defined window.

% of active users who complete an export within 14 days of launch.

Watch out

Adoption measures the splash, not the swim. A great launch spike says nothing about whether the feature earned a place in people's routine — always pair adoption with retention, or you'll celebrate a number that quietly evaporates by week three.

Retention

Behavioral — whether users keep coming back over time.

Reach for Retention when the concern is durability: are the people who showed up last period still here this period? It's the truest test of real value, and the hardest metric to fake with a good launch.

Worked example — a subscription product.

Goal

Users keep returning over time instead of drifting away and churning.

New subscribers are still active past their first month, not one-and-done.

Signal

Users active in one period are still active — doing something real — in the next.

A signup cohort still performs a core action in their second month.

Metric

A cohorted return rate — N-day / N-week retention, or its inverse, churn.

Month-2 retention rate of each monthly signup cohort.

Watch out

Define 'active' honestly. Counting a passive login or a background ping as retention flatters the number and hides churn in plain sight. Tie it to a meaningful action, and always read it by cohort — a blended rate mixes loyal veterans with a leaky front door.

Task success

Behavioral — can people complete key tasks well?

Reach for Task success when there's a specific job the design exists to get done. It's the classic usability lens — effectiveness, efficiency, and error rate on a defined task — and the one you can move most directly with craft.

Worked example — an e-commerce checkout.

Goal

Users can complete the key task quickly, correctly, and without giving up.

Shoppers who start checkout finish it without hitting a wall.

Signal

Users reach the end state without errors, backtracking, or abandonment.

Shoppers reach the confirmation screen without erroring out on payment.

Metric

Completion rate, time-on-task, error rate, or step-level funnel drop-off.

Checkout completion rate, plus error rate on the payment step.

Watch out

Task success needs a defined task and a defined finish line — decide both before you measure, or the number means nothing. Completion rate from analytics tells you that people struggle; a handful of usability sessions tell you why. You usually need both.