# A 30-Day Playbook to Stand Up Web Analytics You'll Actually Use

> Installing an analytics snippet is easy. Getting to dashboards your team trusts and acts on is the hard part. Here's a week-by-week plan to go from zero to a working analytics practice in 30 days.

Source: https://oakdata.co/blog/web-analytics-30-day-playbook

---

Installing an analytics snippet takes five minutes.

Building an analytics practice your team actually trusts and acts on takes a bit longer. The gap is usually the same in every company: the data exists, but nobody believes it, nobody owns it, and nobody has wired it into how decisions get made.

This 30-day playbook closes that gap. Each week builds on the last, so by the end you have clean data, the events that matter, real sessions to learn from, and a review ritual that keeps it all honest.

## Week 1: Install the snippet and verify the basics

Start by getting data flowing and proving it's correct. Skip the verification step and you'll spend the next three weeks doubting every number.

1. **Drop in the snippet.** One script captures pageviews, autocaptured clicks, web vitals, errors, and sessions out of the box. Use the [Next.js quickstart](https://oakdata.co/docs/quickstart/nextjs) (or the React / plain-script versions) and add your `oak_pub_…` key.
2. **Confirm pageviews land in real time.** Open your site, click around, and watch sessions appear in the dashboard. Check that referrers, paths, and devices look sane.
3. **Sanity-check sessions and bots.** Make sure a real visit groups into one session and that bot traffic is tagged, not inflating your counts.
4. **Verify web vitals and errors.** Trigger a console error on purpose and confirm it shows up. You want to trust this channel before week 3.

By Friday you should be able to say, with confidence, "these numbers are real." That confidence is the whole foundation.

## Week 2: Instrument the events that matter and add identity

Pageviews tell you what got visited. They don't tell you whether anyone did the thing your business depends on. This week you make the important moments measurable.

Prioritize a short list, not everything:

- **Activation moments** - signup, first key action, the aha step. Capture them with `oak.capture("signed_up")`.
- **Revenue-adjacent steps** - checkout started, plan selected, upgrade clicked.
- **Friction signals** - failed form submits, empty searches, rage clicks worth naming explicitly.
- **Identity** - call `oak.identify(userId)` once a user is known so sessions stitch across devices and visits.

Resist the urge to track 50 events on day one. Five well-named events you understand beat fifty you'll never look at. For a deeper take on choosing the right ones, read [how to instrument the events that matter](https://oakdata.co/blog/how-to-instrument-events-that-matter).

## Week 3: Turn on session replay and fix the top friction

Numbers tell you *what* happened. Replay tells you *why*. This week you stop guessing.

1. **Enable session replay.** It rides on the same snippet - no new install, just flip it on.
2. **Watch your highest-intent flows.** Pick the funnel you care about most and watch ten real sessions end to end. You will see things no chart surfaced.
3. **Find the top point of drop-off.** Cross-reference where the event funnel from week 2 leaks against what replay shows people actually doing there.
4. **Ship one fix.** One confusing button, one broken validation, one slow step. Fix it, then re-watch to confirm it's gone.

Replay is also the fastest way to debug. When an error spikes, you can watch the exact session that produced it instead of reconstructing it from a stack trace. More on that in [debug faster with session replay](https://oakdata.co/blog/debug-faster-with-session-replay).

## Week 4: Connect your agent and set up the weekly ritual

The last week turns a dashboard into a system. Two pieces: make the data readable by your AI agent, and make reviewing it a habit.

1. **Wire up the MCP server.** Point your AI agent at the [OakData MCP server](https://oakdata.co/docs/api/mcp) with an `oak_sec_…` key. Now you can ask "which page lost the most conversions last week?" in plain language and get an answer backed by your real data.
2. **Save a few standing questions.** Top drop-off, worst web vitals, new errors, biggest week-over-week swing. Let the agent run them for you.
3. **Set a 30-minute weekly review.** Same time every week. Review the swings, watch two or three sessions, ship one focused change. Re-measure next week.

This is the loop that keeps analytics alive instead of decorative.

## The five things to check every week

Once the loop is running, keep your weekly review tight:

- **Activation rate** - are new visitors reaching the key moment?
- **Funnel drop-off** - where is the biggest leak right now?
- **Web vitals** - is the experience getting faster or slower?
- **New errors** - anything broke since last week?
- **Biggest swing** - what changed most, and do you know why?

Five questions, thirty minutes, one change shipped. That cadence is what separates teams who have analytics from teams who use it.

## Why this order works

You verify before you build, instrument before you watch, watch before you automate, and automate before you ritualize. Each week earns the trust the next one depends on. Most analytics setups fail because they skip straight to dashboards nobody believes.

Run this once and you'll have something rare: an analytics practice your whole team actually opens on a Monday.

---

*[OakData](https://oakdata.co/) gives you the whole loop from one snippet - pageviews, events, identity, web vitals, errors, and session replay, readable by your team and your AI agent over REST and MCP.*