Analytics Console Redesign
A data-dense operator console that had grown slow to load and slow to change, rebuilt on the App Router with a real design system and a streaming data layer.
Overview
The existing console was a client-rendered SPA where every view fetched on mount, the charts re-rendered on every keystroke, and adding a screen meant copying the last one. Load time on the main dashboard was over four seconds on a mid-range laptop.
I moved the app to the Next.js App Router, split the data layer into typed Server Component fetches with a tagged cache, and put the interactive filters into small Client Components behind Suspense so each panel streams independently. The heavy table got virtualization and a debounced, deferred filter path.
Alongside the rebuild I introduced a token layer and a shared component library so the four squads working on the product stopped rebuilding the same table and card from scratch.
Outcomes
- LCP on the main dashboard went from ~4.2s to under 1.5s in the field
- INP moved from 'poor' to 'good' at p75 after cutting client-render cost
- New analytics screens now take hours, not days, on the shared components
- Zero-downtime migration — no feature freeze during the move to App Router
Tell me about your product
Share a short brief — problem, timeline, and stack — and I'll reply with a candid view on fit, effort, and the fastest path to a shippable milestone.