A marketing site wired to a custom analytics stack

Scrolled Brain is a speed-reading and focus platform aimed at students and knowledge workers. The founder came to us with a marketing site that was technically functional but converting poorly, and an analytics bill on PostHog that was eating more than fifteen percent of monthly revenue. Worse, the data PostHog was collecting did not actually answer the questions the team needed to make product decisions. The brief had two halves: rebuild the marketing site as a serious conversion surface, and build a custom analytics stack that costs a fraction of what they were paying and tells them exactly which onboarding step is bleeding users.

“PostHog was costing us more per month than our hosting and answering the wrong questions. I needed to know which onboarding step was killing conversion, not how many people scrolled past the fold.”

Generic product analytics tools are priced for a use case most early-stage products do not actually have. They charge per event, which means you instrument less than you should because every track call has a cost. They give you a hundred dashboards you will never look at and none of the three you need. For a focused product with a small number of critical funnels, a custom analytics stack — built on Firestore and a small Next.js dashboard — costs almost nothing to run and answers the actual questions an order of magnitude faster. The marketing site was the same insight in a different shape: stop optimising for vanity metrics and start measuring sign-ups.

The marketing site is Next.js 16 with the App Router, statically generated where possible and edge-rendered where not, with hero copy and pricing structured around the three concrete benefits of the product rather than the technology behind it. The custom analytics stack is shipped as a small browser SDK — @partnersinbiz/analytics-js — that batches events to a serverless ingestion endpoint, deduplicates by session, and writes to a Firestore collection with composite indexes for funnel queries. A small internal dashboard in the same Next.js codebase visualises funnels, retention, and live event streams. Mobile Lighthouse came in at 94 on the marketing site at launch, the analytics stack was in production from day one, and the sign-up conversion rate on the new homepage settled at 38% — almost three times the previous baseline.