GA4 ships with hundreds of dimensions, dozens of stock reports, and an Explore module that can build almost anything — which is exactly why it overwhelms someone who just wants to know if their app is working. You don't need most of it. For a vibe-coded site in its first months, five things tell you nearly everything: active users, traffic source, engagement, your one key event, and revenue. Each maps to a single number you can watch weekly, and the rest of GA4 can wait until you have a specific question.
1. Active users — are real people showing up
Start with the headcount that isn't inflated. Active users counts people who had an engaged session, not every request that hit your site. That distinction matters because raw "visits" include bots and instant bounces. Watch active users week over week; the trend is the signal, not any single day's number. In GA4 it's the default headline metric on the Reports snapshot and the Acquisition reports.
2. Where they come from — source and medium
A user count with no origin can't guide a decision. GA4's Traffic acquisition report attributes each session to a default channel — direct, organic search, organic social, referral, paid — derived from source and medium. This is the metric that tells you whether your launch tweet, your SEO, or word of mouth is actually driving people, and therefore where to spend the next week of effort. If organic search is climbing, that's the compounding channel; pair GA4 with Google Search Console to see the exact queries behind it.
3. Engagement — are they doing anything
Plenty of people "visit" and leave. GA4 measures the opposite of that with engaged sessions and engagement rate: a session counts as engaged if it lasts longer than 10 seconds, fires a key event, or has at least two page or screen views (Google Analytics Help — engaged sessions). Engagement rate is engaged sessions ÷ total sessions. A low rate on a launch spike usually means the traffic was curious, not interested — a content or targeting problem, not a volume one.
4. Your one key event
This is the metric that separates a real measurement setup from a vanity one. In 2024 Google renamed GA4 "conversions" to key events — events you mark as especially valuable, like sign_up or purchase (Google Analytics Help — key events and conversions). Pick the single event that means "this person got the point of the product" and mark it as a key event. Then read it as a rate against active users, not a raw tally, so it stays meaningful as traffic grows.
Mark exactly one to start. A funnel with one clear key event you trust beats ten half-instrumented ones you don't.
5. Revenue — if you charge
If money changes hands, track it as the purchase event with value and currency so GA4 can total revenue and tie it to channels. Even pre-revenue, define the event now so the day you turn on payments, your funnel is already measurable end to end. Revenue closes the loop on the other four: it tells you not just that people show up and engage, but that the engagement is worth something.
What you can safely ignore for now
The flip side of a short list is permission to skip the rest. In your first months you don't need demographic breakdowns, the tech (browser/device) reports, custom Explorations, audiences, or attribution modeling — they answer questions you don't have yet, and tuning them is time not spent on the product. Bounce rate, which GA4 reports as the inverse of engagement rate, is also redundant once you watch engagement directly. Set up the five above, mark one key event, and leave the rest of GA4 closed until a specific question forces you to open it. The cost of ignoring a report is zero; the cost of drowning in all of them is that you check none.
Map the five to one normalized view
The catch is that these five don't live in one place. Active users, engagement, and your key event sit in GA4; revenue may live in your billing system; organic source detail lives in Search Console; paid performance lives in Meta. Reading them honestly means reconciling tools that count differently — which is why GA4 and Meta numbers rarely match.
The durable fix is to record the two events that matter most — your key event and purchase — as first-party owned events as well as GA4 key events, so the count survives ad blockers and consent declines (see owned events vs provider events). Then normalize all five into a single schema. That's what VibesAnalytics does: your agent connects GA4, Meta, Firebase, and Search Console over MCP, the five metrics land side by side, and a daily summary flags which one moved and what to do about it. For the GA4 wiring itself, start with how to set up GA4 on Next.js.
Five metrics, one screen, checked weekly. That's the whole job in the first few months — and it's small enough that you'll actually do it.