Skip to content
UXE Award

Learnability: onboarding that teaches instead of touring

Learnability is one of the seven UXE criteria: how quickly a first-time user becomes a confident one. Why product tours rarely work, what does, and how to measure time-to-first-value honestly.

The UXE editors5 min read

Learnability is the criterion that asks how quickly a first-time user becomes a confident one. Teams usually answer it with a product tour — a sequence of tooltips pointing at things — and editors usually find that the tour is the weakest part of the entry.

Why tours rarely work

A tour teaches in the wrong order and at the wrong moment. It explains the interface before the user has a reason to care about any of it, then disappears exactly when the questions start. People dismiss tours for the same reason they skip manuals: they came to do something, and the tour is in the way.

Two approaches to the same first session

A tour

  • Teaches the interface, not the job
  • Front-loads everything, remembers nothing
  • Blocks the product while it runs
  • Can't be re-read at the moment of need

Teaching

  • Gets the user to one real outcome
  • Explains each thing where and when it's used
  • Never blocks — the product is usable throughout
  • Stays available when the question returns

Design for the first real outcome

The most useful question in onboarding design is: what is the smallest genuinely valuable thing this person can do in their first session, and how fast can they do it? For a design tool it's a first export; for an invoicing product it's a first invoice sent; for an analytics product it's one real question answered.

Four moves that raise a learnability score
  1. 01

    Name the first win

    One specific outcome, not "explore the dashboard".

  2. 02

    Remove what's in the way

    Defer every setting that isn't needed to reach it.

  3. 03

    Seed the state

    Sample data that behaves like real data, clearly labelled and easy to clear.

  4. 04

    Teach at the moment

    One line of help where the question arises, not in a tour.

Empty states are your curriculum

The empty state is the only teaching surface you can guarantee every new user will see. A good one names what will appear, explains what it's for, and offers the action. A great one shows an example of what a populated version looks like, so the person can picture where they're going.

It costs an afternoon and it lifts three criteria — learnability, usability and content design — which makes it the highest-leverage screen in most products.

Help at the point of need

  • Inline, one sentence. Beside the field that raises the question, not behind an info icon that opens a panel.
  • Examples, not definitions. "For example, INV-2026-014" teaches a format faster than a paragraph about formats.
  • Answer the question actually being asked. Next to a tax field, the question is usually "do I need this?" — not "what is tax?"
  • Let the user find it again. Help that only exists during onboarding fails everyone who returns after a fortnight away.

Confidence compounds

Learnability isn't only about the first session. It's about whether the product teaches its own patterns well enough that the tenth feature is easier than the first. Consistency does most of that work: if a list behaves the same way everywhere, a user who has learned one list has learned all of them.

The best onboarding is a product whose second feature costs nothing to learn because the first one taught it.
From the learnability criterion brief

Measuring it honestly

Four measures, done two ways
Compared onMeaningfulVanity
Time to first valueMedian minutes to the first real outcomeTime spent in onboarding
CompletionReached the outcome without support contactFinished the tour
ReturnCame back and did it again unaidedLogged in twice
HelpWhich questions stopped being askedArticle views

The most persuasive thing you can put in an entry is a change: "time to first invoice fell from 11 minutes to 4 after we removed the company-profile step from setup and asked for it later." One sentence, one decision, one measured consequence.

What to send

A genuinely new account — not a seeded demo — so we see what a first-time user sees. Your empty states. The moment of first value. And whatever you measured, with the method attached.

Learnability is weighted heavily in Product Experience and AI & Emerging Experience, where nobody has prior expectations to lean on. Enter a project.

Sample data that teaches instead of lying

Seeding a new account with example data is one of the most effective onboarding techniques available, and it goes wrong in a specific way: the sample data behaves differently from real data, so everything it taught has to be unlearned.

  • Make sample records obviously samples — labelled, and removable in one action.
  • Use realistic volume and realistic mess. Three perfect rows teach nothing about the product at 300.
  • Let every sample record be edited and deleted, so exploring is safe.
  • Never let sample data silently persist into production reporting.

The second session, which nobody designs

Onboarding design stops at the end of session one. But the moment that decides retention is the return: someone comes back two days later, having forgotten half of it, and has to pick up where they left off.

Designing the return
  1. 01

    Remember where they were

    Return them to the work, not to a generic home screen.

  2. 02

    Show what changed

    Anything that happened while they were away, briefly.

  3. 03

    Offer the next step

    One suggestion based on what they actually did, not a checklist of everything.

  4. 04

    Keep help reachable

    The explanation from session one should still be findable.

Questions we get asked

  • Are product tours always bad?

    Not always, but they're usually the weakest tool available. If you keep one, make it skippable, short, and resumable — and put the real teaching in the empty states and inline help where the questions actually occur.

  • How do we measure learnability?

    Time to first real outcome, completion of that outcome without a support contact, and whether people return and repeat it unaided. Onboarding completion rate measures your tour, not your product.

  • Should onboarding be personalised by role?

    If roles genuinely need different first outcomes, yes — but ask one question, not five. A long personalisation quiz before any value is a cost paid up front for a benefit that may never arrive.

  • What should we send in an entry?

    A genuinely new account so editors see a real first run, your empty states, and whatever you measured about time to first value.

  • onboarding
  • learnability
  • scoring

Written by

The UXE editors

The team that selects the work, uses it, scores the seven criteria and writes every verdict on this site.