Analytics Without the Bloat: What We Actually Track and Why

We inherited a project where the previous team tracked every page view, click, scroll depth, hover, and mouse movement. Forty-seven million events for three hundred active users. Nobody had ever looked at the data.

This is the default outcome of "track everything, analyze later." Teams track everything, analyze nothing, pay escalating storage costs, and make decisions on gut feelings anyway.

Our approach: exactly twelve to fifteen tracked events per application. Each must pass the "decision test" -- if this number changes by 20 percent, what specific decision would we make differently? If we cannot answer that, we do not track it.

Our standard SaaS event set: signup completed, onboarding completed, core action performed (the one action defining product engagement), upgrade initiated, upgrade completed, churn risk signal (previously active user goes inactive), support ticket created, and two to three key feature usage events. Each connects to a specific business decision.

We use PostHog for its generous free tier, self-hosting options, and session replay. For simpler needs, Plausible for page analytics plus custom event tracking to our own PostgreSQL events table. The custom approach costs nothing and handles millions of events per month.

The privacy angle matters. GDPR and CCPA make comprehensive tracking a legal liability. Every behavioral event must be accounted for in privacy policies, data exports, and deletion requests. Tracking less is operationally and legally simpler.

We also avoid vanity metrics entirely. Page views, session duration, and bounce rate are interesting but rarely actionable. A founder watching page views go up feels good but learns nothing about whether the product works.

One more thing: review your tracked events quarterly. Business priorities shift. The events that mattered six months ago might be irrelevant now, and new questions might require new events. Treat your event list like code: review it, prune it, and keep it aligned with what your team actually needs to make decisions.

Delete 80 percent of your analytics events. Keep the ones tied to decisions you actually make. You will spend less, comply easier, and paradoxically understand your users better because every number in your dashboard means something.