Concept projectA self-initiated brief, not client work. Product names are fictional.
Lumen · Product analytics SaaS
Analytics the whole team can read.
A product analytics platform that answers questions in plain language instead of making people build charts.
The challenge
Product analytics tools are powerful, but most people on a team never open them. Building a funnel means knowing the event names, the filters and the chart types. So questions go to one analyst, and decisions wait.
Goals
- Designed so anyone on the team can get an answer without learning a query builder.
- Designed to show the few numbers that matter on one screen, with the detail one click away.
- Built to stay fast with years of event data and hundreds of seats.
Approach.
- 01
Start from questions
We collected the questions product teams actually ask each week and designed the home screen around answering them.
- 02
Ask, then refine
A plain-language “Ask” box drafts the chart; every filter it chose is shown as an editable chip, so nothing is hidden.
- 03
One component system
Charts, tables, filters and empty states share one set of tokens, so new reports look finished on day one.
What we built.
A multi-tenant web app with a live overview dashboard, a funnel explorer, plain-language questions that turn into editable charts, and weekly digests on mobile.
Biggest drop: trial to teammate invite. Ask why
Which signup channel retains best after four weeks?
✦Organic search keeps the most people after four weeks: 41% are still active, compared with 22% from paid social.
Monday digest
Your week in Lumen
Teammate invites fell on Tuesday after the onboarding change shipped.
Concept screens, designed and built in HTML and CSS. All names, figures and data shown are fictional.
Design decisions.
Show the query
AI-drafted charts always show the filters they used as chips you can change. Trust comes from being able to check the work.
One accent, many greys
A single blue marks the thing you are looking at. Everything else steps back, so a dense dashboard still reads in a glance.
Numbers you can say out loud
Rounded figures with the change since last week next to them, exact values on hover.
How it would ship.
The plan we would run for a project like this, with a demo every week.
- Weeks 1–2Question research, data model, key screens
- Weeks 3–4Design system, auth, workspaces, roles
- Weeks 5–10Dashboards, funnels, Ask, digests
- Weeks 11–12Billing, onboarding, launch prep
Stack
- TypeScript
- React
- Next.js
- PostgreSQL
- ClickHouse
- Stripe
- Claude API
Deliverables
- Product design and design system
- Web app (multi-tenant)
- Event ingestion API
- Plain-language query assistant
- Billing and plans
- Weekly email digest