Hi-fi prototype

Onboarding for every level

Tradeoffs

1. Gamification. Streaks and rewards can increase engagement, but in investing they can reward activity over quality. Engagement had to be tied to learning, not just completion.

2. Removing Exchange and Market sections. Reduced beginner overload at the cost of less functionality for power users.

3. Personalization vs. complexity. Three levels require more design + engineering, but a one-size-fits-all experience becomes irrelevant for everyone.

User testing

I tested the redesign with 54 users across beginner, intermediate, and advanced levels. Everyone completed onboarding, then explored freely.

I tracked onboarding completion, first action after onboarding, early drop-off (especially after locked content), and task success for labs/experiments, plus hesitation moments before committing to an action.

A clear pattern emerged. Beginners were the most likely to follow the recommended path when it was explicit, but several still stalled after the intro test when the next step was not made obvious. Some users began exploring independently, reached locked or unavailable content too early, and lost momentum before they had experienced the value of the platform. Intermediate and advanced users were more likely to choose free exploration, but many hesitated before starting any meaningful activity because the platform did not clearly signal what was most relevant for their level.

When users were given level-appropriate tasks, the core tools themselves tested well. Beginners who entered the career path were able to complete the first module successfully. Intermediate users were able to locate and complete labs, and advanced users were able to set up experiments. The main problem was not whether users could use the product once inside a flow. It was how confidently the product moved them into the right flow in the first place.

Finding 1: Intermediate and advanced users often did not know where to start when they entered free exploration.
Solution 1: Suggest a career path and personalized content recommendations for all levels as the final step of onboarding, and highlight recommended content directly on the labs page.

Finding 2: Not all beginners started the career path after completing the assessment, even when that was the clearest next step for them.
Solution 2: Apply the Endowment Effect by redirecting beginners straight to their recommended career path after onboarding instead of asking them to choose again.

Finding 3: The intro test was not always a reliable signal of user knowledge or intent. Some participants moved through it too quickly or answered inconsistently, which created a risk of inaccurate personalization.
Solution 3: Make content adaptive to user behavior and progress in real time, rather than relying only on a single onboarding assessment.

Constraint and compromise

The biggest unresolved tradeoff was personalization quality versus feasibility. A truly adaptive onboarding system would require instrumentation, content tagging, recommendation logic, and ongoing tuning based on real usage data. In this concept, I kept the MVP more conservative: redirect beginners into a clear default path and use lightweight recommendations, accepting that some users would still receive imperfect routing until behavioral signals could be collected and iterated on.

Results

Because the redesigned platform was validated through prototype testing rather than launched in production, these are projected outcomes and risks based on observed behavior, moderated task success, and the friction points identified during research.

1. Higher first-action rate after onboarding. The redesign gives every user a clear next step instead of leaving them in open exploration immediately after assessment. Based on the testing pattern, I would expect the percentage of users who start a meaningful activity in their first session to rise to roughly 75-85%. The risk is that stronger routing can feel too prescriptive for intermediate and advanced users, so the system would need a clear “skip” path and a way to regain control.

2. Lower early-session drop-off. Redirecting beginners into a recommended career path and surfacing relevant content for intermediate and advanced users should reduce the number of users who leave after hitting unavailable or irrelevant content. A realistic projected reduction in early drop-off is 25-35%. The tradeoff is discoverability: hiding or delaying content can make the platform feel smaller, so it would need transparency about what exists and when it unlocks.

3. Faster time to value. One of the clearest problems in the original platform was hesitation before action. With adaptive onboarding, highlighted recommendations, and stronger guidance, the expected time from onboarding completion to first meaningful action should fall to under 2 minutes for most users. The compromise is that speed can come at the cost of exploration, so the interface would need to support both guided starts and safe free discovery.

4. Stronger completion of level-appropriate tasks. In testing, users were able to complete the core activities once they entered the right flow. The redesign should therefore increase completion rates for beginner modules, labs, and experiments by improving routing rather than changing the activities themselves. A realistic target would be 85%+ task completion for users who enter the recommended flow. The risk is measuring the wrong thing: higher completion does not guarantee better learning or decision quality, so success metrics would need to include understanding, not just finishing.

5. Better personalization accuracy over time. Replacing fixed labeling with adaptive content should reduce the damage caused by rushed or inconsistent intro-test responses. The expected result is not perfect classification on day one, but a gradual increase in relevance as the system responds to user behavior instead of relying on a single quiz outcome. The limitation is the cold start: without enough behavioral signal, the system will still misroute some users early, and it requires ongoing tuning and instrumentation to improve.

6. More meaningful retention signals. The redesign shifts success away from superficial engagement and toward measurable learning behaviors: starting a path, completing a module, entering a lab, reviewing feedback, and returning for the next level. If launched, these would be the primary indicators of success, because they reflect actual progress rather than passive visits. The tradeoff is that even “good” engagement metrics can be gamed, so the product would need guardrails that reward reflection and learning outcomes instead of encouraging risky activity.

Economics & Business model

Investure operates on a freemium-to-subscription model with a B2B channel. The free tier delivers one complete beginner career path, daily challenges, and guided simulation - enough to demonstrate the full learning loop and build a return habit before monetisation. The upgrade trigger is natural: the user completes the first path and wants the next one, or completes two intermediate simulations and sees advanced scenarios locked behind the threshold. The ceiling is hit at the moment of highest motivation, not before it.

Individual subscription unlocks all career paths, labs, simulations, bias pattern tracking, and annotated trade history. Professional subscription adds experiment log, benchmark comparison, historical replay, and falsification check - tools professionals currently pay for separately across trading journals, simulation environments, and backtest platforms. Investure consolidates them with learning context built in, which is the differentiator that justifies the price point.

The B2B channel is the highest-margin opportunity. The intermediate and advanced flows are structurally identical to professional development tools - a wealth management firm could use the lab and experiment system to onboard analysts, train advisors on behavioral decision-making, or assess junior portfolio managers against their own decision history. This requires no product modification. The core metric across all tiers is session depth, not daily active users. A user who completes three labs generates more defensible retention signal than one who opens the dashboard daily without executing. The product is built around this distinction, and the business model tracks it the same way.

What would I do differently?

− Instrument the MVP before launch, not after. We launched and then used Hotjar to understand behavior. The funnel data showed where people dropped off, but not why. If I had set up proper event tracking from day one, the behavioral data would have been richer and the think-aloud sessions would have started with real hypotheses to test, not open questions.

− Run a copy audit before the information architecture. The case identifies jargon as a threshold condition - one unknown term ends the session. The design response was to add a glossary and simplify navigation. But I never systematically audited every label, heading, and action in the interface for language that would fail a beginner. Starting over, that audit happens before the IA is finalized, because the copy and the structure are the same problem.

− Design one system, not three separate experiences. Three user levels led naturally to three parallel design tracks. The result was complexity that was hard to manage and harder to test. Starting over, I'd design a shared component system first - one that works structurally for all levels - and differentiate only what genuinely needed to be different. Most of the difference between a beginner and an advanced experience is in the content and the degree of guidance, not in the interface itself.

What will I test next

− Run user tests for advanced and intermediate users (simulations and labs) adapted for different type of content (situational task, advanced concepts exploration etc.)

− The core differentiator of this platform is feedback that distinguishes luck from skill. I'd track whether users who receive reasoning-coherence scores make structurally different decisions after 10 labs vs 1 lab - not just whether they feel more confident.

− I'd run a structured interview with 5–8 financial advisors or analysts to test whether the experiment log and reasoning-quality feedback genuinely replace their external spreadsheets - or whether the professional context requires features not yet designed.

What is next for Investure?

− Develop the adaptive content system - the mechanism that adjusts difficulty based on real-time performance rather than the initial test. This is the architectural prerequisite for long-term retention and for the personalization promise made throughout the design.

− Evaluate every step of user journey against engagement metrics - including the habit loop, streak mechanic, and dashboard design.

What did this teach me about UX

− Feedback quality matters more than feedback quantity. The current platform says "order added." That is technically feedback. But it tells the user nothing about whether their reasoning was sound, whether the outcome was process or luck, or what to do differently. Designing feedback that explains rather than just evaluates is the hardest and most important design challenge in this space.

− Gamification without understanding produces harm. Robinhood's confetti and Duolingo's streaks both backfire when applied to consequential decisions. This project forced me to think about where motivation design ends and manipulation begins - and to build the constraint that every engagement mechanic must be tied to demonstrated understanding, not just task completion.

− One size for everyone is a choice to serve no one. Every competitor audit finding came back to the same root cause: the platform was built for an average user who doesn't exist. Personalization has a real engineering cost - but the alternative isn't neutral, it's a product that is actively wrong for every user it touches.