SaaS Analytics Dashboard

Overview

The client had a growing B2B SaaS product where the analytics page had become unusable at scale. What once loaded in 2 seconds now took 11–15 seconds — and support tickets about it were piling up.

I was brought in to diagnose and rebuild the analytics layer without touching the core product database schema.

The Problem

The original implementation ran ad-hoc aggregate queries directly against the production transactional database. At low user counts this was fine. At 50 k daily active users, it created lock contention and occasionally brought down the whole app.

Key constraints going in:

• Could not modify the core product schema
• Had to support arbitrary date ranges up to 2 years of history
• Infrastructure budget was limited — no managed data warehouse

Solution

I built a separate analytics pipeline alongside the existing product:

Pre-aggregation layer — A nightly job (Cloudflare Workers Cron) reads from the production DB in read-only mode and writes summarised rows into a separate schema. The queries against this schema are cheap and predictable.

Redis caching — Common dashboard views (last 7 days, last 30 days) are cached with Upstash Redis. Invalidation fires when the nightly job completes.

Edge delivery — The API endpoint moves to a Cloudflare Worker, so responses are served from the nearest PoP with cache hits returning in single-digit milliseconds.

Stack

Layer	Tech
Frontend	React + TanStack Query
API	Cloudflare Workers
Cache	Upstash Redis
Analytics DB	Cloudflare D1 (SQLite at the edge)
Charts	Recharts

Results

• P95 load time: 11 s → 380 ms
• Server costs: down 40 % (edge compute vs. always-on server)
• Churn: client reported a measurable drop in the month after launch
• Zero schema changes to the production database