Measure the decision, not the dashboard

A dashboard can contain many accurate numbers and still fail to answer the question in front of the team. Start with the decision: what would you do differently if this page, campaign or change worked?

Separate three stages

Stage	Useful question	Do not assume
Discovery	Did the intended audience reach the page?	Every visit represents a different person.
Useful action	Did someone complete the task the page supports?	A button click means the task was completed.
Outcome	Did that action lead to the result the team values?	Every later sale was caused by the first page.

Define the event in operational terms

“Engaged reader” is not a complete definition. Specify the event and the rule: a completed calculator run, a successfully saved enquiry, or another observable action. A scroll event may be useful for interface analysis, but it is not automatically evidence that the person understood the material.

Check the implementation with a known action. Confirm the event name, destination project, time, page address and whether the same action is counted twice. A chart with data is not enough to prove that the data describes the intended behaviour.

Keep the denominator visible

If you report a rate, show what is divided by what. Sessions, users and events are different units. Removing a segment from the numerator while leaving it in the denominator can make a tidy chart misleading.

Compare the same kind of period

Match weekdays, traffic sources and audience conditions where possible. Note launches, outages, consent changes and bot filtering. If the tracking method changes, a break in the line may reflect measurement rather than a change in demand.

For a planned change, use the experiment card. For a page not yet written, start with a bounded brief rather than deciding the success metric afterwards.