Making Data Legible
Designing an analytics platform from zero — from the first metric to the weekly report Kibit's team runs on.
Context & challenge
The challenge
We built Kibit from a blank page — so there was no analytics to redesign. The numbers the business would run on didn’t exist yet. The hard part wasn’t drawing charts; it was getting product, growth and engineering to agree on what to measure and how to count it.
Objective
The objective
Stand up Kibit’s analytics from zero:
- Define — what success looks like for a new subscription app.
- Instrument — every event with engineering.
- Ship — a report that works for a 5-second glance and a deep dive.
The solution
A weekly report that reads like an editorial, not a dashboard
It unifies four sources (GA4, Firestore, RevenueCat, stores) into one weekly snapshot.
A “read of the week.” Each report opens with an auto-written line — “Views grew +21%; conversion at 0% missed the 2–3% goal.” The takeaway lands before any chart.
Metrics with built-in targets. KPI cards color against their goal, not just last week. You read health before the number.
Six lenses, one language. Summary (5-second read), Content (analyst depth + CSV), Acquisition (GA4 funnels), Monetization (paywalls by origin), Feedback (ratings + real messages), Team.
One color system. Teal = healthy, red = off — everywhere, always vs. the target.
Editorial, on purpose. Serif for human quotes, sans for data, calm warm surface. It doesn’t look like a BI tool — so people actually read it.
Research
Before designing a screen, I designed the questions
With no data to study, I worked two fronts at once: stakeholders (what decisions do you make weekly?) and engineering (what can we actually capture?).
Two reading speeds
Leadership needs 5-second KPIs; the team needs deep filters and export. One screen can't do both.
Same word, same number
“Conversion” meant different things to different people. Defining metrics mattered more than visualizing them.
A number needs a target
“0% conversion” means nothing until you see it against a 2–3% goal.
The funnel is the spine
Onboarding → subscription drove everything, so it leads the report.
Role
I led the analytics end to end — from metric definition to shipped platform
Role: Lead UX/UI · Product Designer. Timeline & tools: 5 months · Claude Design (AI) · Illustrator · GA4 · RevenueCat · App Stores · Google Sheets.
Metrics
Built the KPI model from scratch.
Events
Mapped the event taxonomy with engineering.
IA
Split the report into six lenses around real decisions.
Data viz & UI
Charts, goal-based color, and the full launch-ready design.
Outcome
Kibit's first shared source of truth
The team went from scattered numbers to one weekly report everyone reads — and, more importantly, a shared language. “Conversion,” “activation,” “a title that works” now mean the same thing to everyone, because the definitions and targets live inside the tool. A framework that didn’t exist five months earlier became the default way the team sees the product.
What I learned: the hard part of a metrics product isn’t the dashboard — it’s the definitions underneath, and giving every number a target. Designing the language of the data was as much product design as designing the screens.
Defined
A metric framework and event taxonomy built from a blank page.
Instrumented
Four data sources unified into one weekly snapshot.
Shipped
A report serving the 5-second glance and the deep dive, in one language.