How to write a measurement plan before you launch a campaign

Start from questions, not metrics

Write down the three to five questions the campaign must answer. 'Did this campaign generate qualified pipeline at or below our target cost?' is a question. 'Track everything in GA4' is not.

Questions keep the plan honest, because each one forces you to specify the data needed to answer it. Metrics chosen without a question attached are how dashboards end up with forty widgets nobody reads.

Name one primary metric

Exactly one metric decides whether the campaign worked. Everything else is diagnostic. Without that hierarchy, every review becomes an exercise in finding the chart that supports the preferred conclusion.

Pick the primary metric as close to revenue as your data allows: contribution margin, qualified opportunity value, new customer revenue. If you must use a proxy such as cost per qualified lead, write down the assumed conversion rate from proxy to revenue so the assumption is visible.

Define events and values precisely

For each metric, specify the event that produces it, where it fires, what parameters it carries and what monetary value is attached. Ambiguity here is what creates the classic 'our conversions doubled' moment that turns out to be a tag firing twice.

• Event name and trigger condition, written explicitly.
• Deduplication rule — one conversion per session, per user, or per order.
• Value: actual revenue, an agreed static value, or margin.
• Which conversions are primary for bidding and which are secondary signals only.

Agree attribution rules in advance

Platform-reported conversions, GA4 models and CRM records will never agree exactly, and that is normal. What matters is deciding in advance which source is authoritative for the primary metric, and what the others are used for.

A workable convention: CRM for commercial truth, GA4 for cross-channel comparison, platform data for in-flight optimisation. Write it down and stop relitigating it every month.

Set targets, guardrails and a decision rule

State the target for the primary metric, the guardrail metrics that must not deteriorate — average order value, lead quality, brand search — and the decision rule: at what point do you scale, iterate or stop.

Decision rules written before launch are the difference between optimisation and improvisation. They also protect the campaign from being killed in week one on noise.

Specify the reporting mechanics

• Who receives what, and how often.
• The date the data is considered final, allowing for conversion lag.
• The dashboard where numbers live, with one owner responsible for it.
• Naming and UTM conventions — boring, and the single most common cause of unusable data.

Validate before you spend

Run a live test of every conversion path yourself: click the ad, submit the form, place the order, and confirm the event appears with the right value in every system. Ten minutes of validation prevents a month of unusable reporting.

Then keep the plan with the campaign. When the post-mortem happens, the argument is about what to do next rather than about what the numbers mean.