Finlens

Retail investing has never had more information (screeners, terminals, YouTube analysts, subreddits) and never less clarity. A first-time investor researching one stock faces candlesticks they can't read, ratios they can't rank, and opinions that all contradict each other. The tools built for professionals assume the judgment the retail investor doesn't have yet.

So most people resolve the anxiety socially: they buy what a friend bought, what a influencer shouted, what was already up 40%. The cost of that shortcut is invisible until it isn't.

Solving it mattered because the missing layer isn't data, it's an analyst: something that reads the financials, the valuation, the technicals, and the sentiment, then commits to a call and shows its work. That layer is exactly what LLMs made buildable.

The category splits into two camps. Data platforms, screeners, brokers' research tabs, TradingView, hand you every number and no conclusion. Recommendation media, newsletters, influencers, Motley Fool-style picks, hand you a conclusion and no evidence you can interrogate. Simply Wall St comes closest to the middle with its visual snowflake, but stops short of a committed, personalized call.

Two research findings shaped the design. First, retail users don't abandon research tools because they lack data, they abandon them at the moment the tool asks them to synthesize. Second, trust in an AI verdict tracks with how gracefully it shows uncertainty: a confident number with a visible confidence score reads as honest; an unqualified rating reads as marketing. Both findings pushed the same direction: verdict first, evidence attached, uncertainty visible.

1. I.

   Verdict first, evidence one scroll away.

   Every stock page opens with the call, BUY, 92% confidence, medium risk, +28% expected, and every claim beneath it is a named, tappable piece of evidence. The user should never wonder what the product thinks.

2. II.

   Charts you touch, not read.

   Every chart answers the hand: crosshairs and tooltips on hover, timeframes that re-draw, stacked views that decompose. Interaction is comprehension, the Simply Wall St lesson, taken further.

3. III.

   The profile is the lens.

   Onboarding captures goal, risk, horizon, and markets, and every verdict, discovery feed, and portfolio nudge is computed through it. Two users see two different Finlenses, honestly labeled.

The IA follows the investor's loop, discover, research, decide, monitor, with the stock workspace as the spine: one page, seven evidence tabs. Two artifacts below: the sitemap of the platform, and the journey from onboarding to a monitored portfolio.

What went well:leading with the verdict. Every earlier draft that hedged, scores without calls, dashboards without conclusions, tested as “another data site.” The moment the design committed to BUY / HOLD / AVOID with confidence attached, the product had a personality and a promise.

What I'd improve:I'd design the losing states earlier, a verdict that turns out wrong, a portfolio in drawdown, an AI note that ages badly. Trust in a finance product is built in the red weeks, not the green ones, and that surface deserves first-class design attention.

The lesson that carries: for AI products, the courage to commit, visibly, auditably, with uncertainty on the label, is the differentiator. Anyone can chart the data. The product is the judgment.