Analytics measurement plan: events, funnels and running costs

Start with a decision rather than a dashboard

Write one question the team can act on. For example, a hypothetical service might ask whether visitors who start a quote request can complete it, and which step needs investigation when they stop. Name the person who owns that decision and the evidence they need. A page-view total can describe activity without explaining the next action. Keep a measurement brief short enough that the people building the journey and the people reading its reports can agree on what success means.

Separate acquisition, product use and observed behaviour

A website acquisition question concerns where an audience arrives and what happens afterwards. A product question can concern onboarding, repeated actions or a sequence inside an account. A replay or heatmap can help investigate a particular interaction without explaining the whole audience. These jobs can share a platform, or need connected tools. Choose the required question first, then compare the actual package. Avoid counting the presence of a funnel or replay label as proof that both products collect the same evidence.

Make a small event dictionary

For each important action, record a name, the exact trigger, permitted properties and the expected count. Decide whether a quote request is counted when a person presses a button, when validation passes or when the server accepts the request. Those are different events. Google Analytics recommended events require configuration and the prescribed parameters; they are not sent automatically. Use the selected platform’s documented event model; do not rename an existing event casually and assume historical reports will remain comparable.

Define identity and the reporting boundary

State whether the question concerns a browser, device, person, organisation or account. Explain which identifiers the implementation may use and who approves them. Check what happens before and after sign-in, when a person changes devices and when collection is unavailable. Include the intended date range, time zone, filters and exclusions in the report brief. An apparent difference between two tools can come from different measurement boundaries. It does not by itself establish that one tool is more accurate.

Verify the normal path and an exception

Use synthetic actions that are permitted for the evaluation. Complete the intended journey once, repeat a step, trigger a validation failure and review the resulting events. Compare the names and properties with the dictionary, then inspect the aggregate report or funnel that will support the decision. Google Analytics provides DebugView and Realtime views for event verification; other platforms have their own workflows. A visible network request alone does not prove that the report represents the intended event once. Retain the expected and observed results.

Agree the collection and access rules before rollout

Ask the responsible privacy, security and product owners to approve which data may be collected, which fields must be excluded or masked and who can read it. Specify how a change to a form or page will be reviewed. Include retention, export, deletion and access handover in the implementation scope. A masking feature appearing in a product description does not establish the behaviour of your configuration. This is an implementation checklist, not a determination that a particular collection arrangement satisfies legal requirements.

Price the retained measurement plan

List the units used in the quote: events, sessions, monthly people, replay recordings, retention, seats, exports and optional products. Describe a normal month and a busy month. Keep a temporary trial credit separate from the ongoing allowance. If reports depend on a warehouse or connector, include its operating cost and owner. Ask what happens at the limit: collection may stop, a feature may become unavailable or additional usage may be charged according to the selected terms. Confirm the actual arrangement before enabling production traffic.

Compare the tools around the agreed evidence

Use Google Analytics vs Adobe Analytics for a website-measurement shortlist, Mixpanel vs Amplitude for product journeys, Matomo vs PostHog for their different deployment and product scopes, or Hotjar vs Microsoft Clarity for behaviour-investigation requirements. Retain the event dictionary, expected results, package assumptions and dated quote with your comparison scorecard. Revisit the brief when the journey or buying decision changes.

Frequently asked questions