Create a Local Marketing Dashboard With Definitions People Can Reproduce

A dashboard is not reliable merely because its charts update automatically. If one chart counts all form attempts and another counts deduplicated qualified enquiries, a shared label such as “leads” hides a major difference. Metric definitions make the dashboard reproducible and prevent a design that looks polished from becoming a source of inconsistent decisions.

Create a metric dictionary

For every displayed measure, write the plain-language definition, source, counting unit, inclusion rules, exclusions, date basis, refresh timing, owner, and known limitations. A definition should be precise enough that a new analyst can recreate the number without asking the original author what they meant.

Distinguish counts, rates, and ratios

A count describes a quantity, a rate divides one count by another, and a ratio compares two quantities that may use different units. Display the numerator and denominator for important rates. Do not sum percentages across pages or branches; calculate the overall rate from the underlying totals using the intended weighting.

Design around decisions

Group metrics by the decision they support: search visibility, website behavior, enquiry handling, lead qualification, and commercial outcomes. Do not put every available measure on the first screen. A chart should answer a question, identify a change worth investigating, or show whether an agreed action was completed.

Label dates and comparisons explicitly

Show the reporting period, comparison period, time zone where relevant, and whether data is complete. Avoid vague labels such as “last month” in a downloaded report that may be opened weeks later. Mark partial periods and explain when a source has a known processing delay.

Example of a dashboard conflict

A fictional agency dashboard shows 73 leads, while the CRM shows 49 qualified enquiries. The correct response is not to force the numbers to match without investigation. The dashboard may count calls, form submissions, duplicates, or events from another period. Add a metric note and trace the source logic before changing either total.

Control visual interpretation

Use clear units, readable axis labels, and consistent definitions. A truncated axis can exaggerate a small movement; a percentage without its base count can make a tiny sample look dramatic. Include a caveat beside the chart when the limitation materially changes the interpretation rather than burying it in a distant footnote.

Maintain the dashboard over time

Version the metric dictionary, record changes to event logic, and assign a review owner. When a source platform changes a report or dimension, verify the effect before editing the dashboard. Retain the prior definition when historical comparisons depend on it, and annotate any break in the series.

Questions readers often ask

Should every metric have a target?

Not necessarily. Some metrics are diagnostic or data-quality checks. If a target exists, explain its origin and do not present it as an industry benchmark unless properly supported.

Who should own a metric definition?

Name an accountable role or person who can approve changes and coordinate with the source-system owner.

Decision note

If two teams use the same metric name for different things, rename one of the measures rather than hiding the difference in a footnote. Clear labels are part of data quality.

A quick acceptance test

Ask a colleague to select one card and explain, without opening the source system, what is counted, which dates are covered, and what would make the value invalid. Then ask them to reproduce the result from the documented export. If they cannot do both, the card needs a clearer definition or a more direct source link. This small exercise checks whether the dashboard communicates a reliable measure rather than merely displaying a number.

Review question

What evidence would change the conclusion? Name the extra record, technical test, or later observation that would settle the remaining uncertainty. The summary should make it possible to distinguish a verified observation from an open diagnostic question.