Reporting
Give every dashboard metric a definition
A practical metric brief: decision, population, formula, time window, freshness and owner. Make the number understandable before adding another chart.
A dashboard becomes more useful when people agree what its numbers mean. Start with a decision and a written definition, then choose the visualization.
A familiar label can hide different calculations
“Conversion rate” can describe several different ratios. One team may divide accepted enquiries by sessions, another by visitors, and another by form starters. None of those labels tells the reader which population was used.
Start with the question the report should answer. If the decision is whether a form needs attention, the relevant denominator may differ from the one used to compare acquisition channels. Write that choice down rather than letting the chart silently make it.
Use a compact metric contract
A metric contract is simply a short shared definition. It need not become a large documentation project. Place it close to the report and make someone responsible for keeping it current.
For a synthetic example, define “form completion ratio” as confirmed test-form submissions divided by test-form starts during the selected period. This example is a proposed definition, not a platform default or a claim about a real dataset.
- Decision: what action could this number inform?
- Population: which people, sessions or records are included?
- Formula: what are the numerator and denominator?
- Time: which period and time zone are used?
- Exclusions: are tests, duplicates or invalid records removed?
- Freshness: when did the source last update?
- Owner: who reviews a changed definition?
Make missingness visible
An empty source and a measured zero are different states. If a connector has not refreshed, a dashboard should not quietly imply that nothing happened. Explain whether data is missing, delayed or genuinely zero according to the source.
Comparison periods also need context. A short partial period can be useful, but label it as partial and compare like with like. When definitions change, document the effective date so an apparent movement is not mistaken for a business trend.
Review with the person who makes the decision
Ask someone to read the report and explain what they would do next. If they cannot identify the period, population or meaning of the main metric, improve the definition and layout before adding more visual detail.
The checklist is an editorial method for building clearer reports. It does not certify data quality. Source reconciliation and implementation tests still need their own evidence.
- Can the reader explain the metric without opening a separate chat?
- Can another analyst reproduce the calculation from the definition?
- Are limitations visible at the point of decision?
- Is there a clear next check when a number looks wrong?