Case study Metrics generalized Edtech · B2B SaaS

Make a course catalog feel like a path.

A learning platform redesign that replaced a one-size-fits-all content dump with structured learning paths for both employees and customers.

RoleSenior Product Designer, contract
CompanyLearnUpon
TimelineNov 2023 to Mar 2024
UsersInternal employees & enterprise customers
2 daysFaster onboarding time-to-competency
↑ completionEnd-to-end completion materially up
1 frameworkThree cohorts served by one IA
WCAG AAAccessibility baseline verified in design QA

Onboarding and completion measured against pre-redesign cohorts; exact figures generalized under NDA. Accessibility checked against WCAG AA success criteria during design QA.

In 30 seconds
  • The shift: from "here is all the content" to "here is your path," with milestone-based paths mapped to role and goal.
  • The fix that mattered: weighted milestone progress replaced a technically accurate bar that testers refused to trust.
  • Results: onboarding roughly two days faster and completion materially up; fully self-directed browsing is still the unsolved tail.
Learning platform overview: path-first home surface with milestone progress
FIG. 01 The path-first home surface. Named paths tied to role and goal replace the search-driven catalog as the primary entry point.
01 · The problem

The worst of the catalog and the worst of the LMS.

The platform served a flat catalog of 200+ unranked lessons identically to a brand-new hire, an experienced engineer, and a paying customer. Completion was poor, support tickets high. Two cohorts drove the failure: the New Hire ("just tell me what I have to finish this week") and the Upskiller ("I don't need lesson one. I need lesson seven. Why is there no way to just go there?"). Most users abandoned between lesson 3 and 5, and the structural cause wasn't content; it was the absence of any visible progress horizon.

Cohorts and quotes from moderated tests with new hires and customer upskillers during discovery; quotes lightly edited for length.

The insight

This wasn't about better lessons. It was about replacing "here is all the content" with "here is your path through the content."

02 · The shift

Paths over catalogs, milestones over checkboxes.

Before: the catalog 200+ LESSONS, NO PRIORITY ORDER After: the path Milestone 1what to do today Milestone 2done enough this week Milestone 3productive SKIP TEST ON EVERY ENTRY: I ALREADY KNOW THIS
FIG. 02 The structural shift: an unranked catalog becomes named milestones with a visible finish line, and a skip test guards every entry.
Sign inany cohort Pick a role pathnamed, goal-mapped new hire Milestone 1 › 2 › 3guided, weighted progress, visible finish line upskiller Skip test"I already know this" Lesson sevenstraight to the gap
FIG. 03 One IA, two journeys: guided onboarding for the new hire, a tested shortcut for the upskiller. The skip test was a structural addition, not a polish pass; our first design wrongly assumed paths would self-select.
03 · What testing changed

The progress bar that lied.

Lesson-count progress, rejected: every lesson moves the bar equally a 2-minute video counts the same as a 30-minute lab, so nobody trusts the number Weighted milestone progress, shipped: the bar reflects effort Module 2 of 5: 60% complete. Trust in the number was the unlock.
FIG. 04 Why the first bar lied: technically accurate, emotionally wrong. Weighting by effort made the number trustworthy. The other mid-project pivot: customer testing surfaced that self-serve learning happens on phones during commutes, forcing a layout, accessibility and offline-state rework.
04 · Results & retrospective

The results, including the miss.

Onboarding

New employees reached productive status roughly two days earlier on average.

Completion

End-to-end completion rose materially, largest gains on the previously worst-performing paths.

Scalability

The framework absorbed next-quarter content additions without new design work.

Weakest score

Optional, non-required courses: without a path forcing structure, users still struggled to choose. The catalog problem is unsolved for fully self-directed learning.

What I'd argue about with my past self. True personalization is unfinished: we shipped role-based paths, not learner-based ones, and a pre-assessment that builds a custom path is still the highest-leverage next move. On gamification, we resisted points and badges because milestone progress worked by reflecting real effort; I'd defend that call again, but less smugly.

Screens and details generalized to respect NDA.

Next case study

Top Eleven × Bundesliga: the collectible gallery

Read next All work