The Learner Curriculum Model
Product Design for High-Stakes Test Prep
Every test-prep program the company launched shared one underlying learning model: a short diagnostic that places the learner, a study pack that teaches daily, and full-length practice tests that benchmark readiness. I defined and documented that model as the canonical product spec across programs, including the single design call that mattered most — that studying and testing must feel like different products.
The Problem
Test-prep products fail in a specific, repeatable way: they conflate studying with testing. When practice questions carry a timer and a score, learners feel judged during the part that's supposed to teach them. When practice tests offer hints and instant feedback, learners walk into the real exam having never experienced its actual conditions. Both failures come from one screen being ambiguous about which mode it belongs to.
Research & Discovery
Mapped three distinct entry states — a returning taker who knows their score, a new taker with a fixed exam date, and an explorer with neither — and found each needs different information collected and a different first session
Identified the session report shown before the paywall as one of the highest-leverage surfaces in the product, since it's where a learner decides whether the assessment understood them
Traced the underlying question behind three separate open design threads across programs and found they were the same one: what does a score mean, and what should the learner do next
Strategy
Three Components
Diagnostic places the learner, study pack teaches daily, practice tests benchmark readiness. Same shape across every exam.
Formative vs Summative
Studying and testing must feel like different products. Identify a screen's mode before designing it.
Disjoint Item Banks
Practice-test items never appear in the study pack, or the benchmark measures memory instead of readiness.
Reorganize, Don't Replace
Customers own their curriculum and standards. The platform restructures it into a learner-first system.
Key Decisions
Made 'which mode does this screen belong to' the first question in any design review. Nearly every scoping mistake in test prep traces back to a screen that was ambiguous about whether it was teaching or measuring
Positioned the platform as reorganizing a customer's existing curriculum into a learner-first structure rather than replacing it. Credentialing bodies own their content and their standards; a product that implies otherwise doesn't get bought
Allowed a concrete prior score to skip the diagnostic. Making a returning learner who already knows their band sit through a placement test is friction that buys nothing
Sequenced the content pipeline as import-then-decompose: bring customer lessons in as-is for launch and modernize images without changing intent, then decompose into reusable tagged artifacts afterward. Full knowledge-graph tagging before a binding launch date would have been the wrong trade
Maintained an explicit list of items requiring customer sign-off before scope could lock — diagnostic length and composition, which items are eligible for which component, per-item metadata, and what the session report shows before the paywall
Results & Impact
3
Model Components
3
Onboarding Paths
8wk
Reference Plan
- Defined the canonical three-component model — a ~20-item diagnostic for placement, a study pack sized several times a full test for daily learning, and full-length practice tests for periodic benchmarking — generalized so the same shape held across every exam program
- Established the formative/summative split as the highest-leverage design decision: study packs run untimed with immediate per-question feedback, an always-available assistant, adaptive item selection, and per-topic mastery; practice tests run with an enforced exam-matching timer, a fixed form, no assistant until post-test review, and a single score with percentile
- Set the hard rule that practice-test items must be disjoint from the study-pack bank, because reusing items turns a readiness benchmark into a memory check
- Designed three persona onboarding paths — returning test taker, new test taker with a date, and explorer — each collecting different inputs and routing differently, including skipping the diagnostic entirely when a concrete prior score already exists
- Specified the universal core loop and an eight-week reference plan that compresses for shorter timelines, giving every program a default study schedule rather than a blank calendar
- Captured the model as a reusable skill so any scoping or design review on any program could be checked against it, rather than relying on the spec being remembered