Metric Definitions: How to Build a Shared Metric Dictionary

Use the same fields for every metric so people learn where to look. Keep each entry short enough to read in a minute:

1. Name and plain-language meaning. One sentence a new employee could understand.
2. Formula. Numerator, denominator and time period, written out.
3. Inclusions and exclusions. What counts and what doesn't, stated explicitly.
4. Data source. The table, report or system the number comes from.
5. Grain and cut. By day, week or month, and by which breakdowns (region, product, customer segment).
6. Owner. The one person who answers questions and approves changes.
7. Known caveats. Data lags, backfills or known quality issues.
8. Last reviewed and version. So people can tell whether an entry is current.

Add an example calculation with made-up numbers. Seeing the arithmetic once prevents most misreadings, and it doubles as a test if you later automate the metric in a reporting tool.

The trouble almost always sits in the exclusions. Three examples show the pattern.

Active customer. Does it mean anyone with a paid contract, anyone who logged in during the last 30 days, or anyone who was billed this quarter? A customer on a paused plan might count in one definition and not in another. Pick one, and give the others different names.

Churn. Is it lost customers or lost revenue? Do downgrades count? Does a customer who cancels and returns in two weeks count as churned? Say whether you measure by logo count or by revenue, and state the time window.

On-time delivery. On time compared with the date you promised, the date the customer requested or the date your system originally scheduled? A customer-requested delay should not count as late, but only if you write that rule down.

Notice that each of these disagreements is really a business decision about what you want to manage. Resolving it takes a conversation, and the dictionary records the outcome. The operations KPI dashboard guide shows how definitions feed the page leaders actually read.

Run a short working session for each contested metric rather than circulating drafts by email. Bring the people who produce the number and the people who use it, and follow this sequence:

1. Show the competing numbers side by side, with the different formulas behind them.
2. Ask what decision the metric is supposed to inform, since that usually settles which definition is right.
3. Choose one definition for the official name. If another team needs the alternative, give it a distinct name, such as "billed customers" versus "active customers."
4. Name the owner and write the entry during the meeting.
5. Publish the entry and announce it in the channels where the old number circulated.

Expect some resistance, especially from whoever's number changes. Present the change as a rule for the future and note the old figures in a comparison table if you restate history, so trends remain interpretable. Keep the dictionary to the metrics that matter most, maybe twenty to start, rather than defining everything.

Put it where people already look: a shared wiki page, a spreadsheet linked from your dashboards or the documentation area of your reporting tool. Whichever you choose, keep one canonical copy, and link to it from every dashboard and report rather than pasting definitions in.

If you use a BI tool, encode the definition in the tool once, as a saved metric or a governed data model, so every chart calculates it the same way. Metabase is often used for teams that want to define reusable questions and models on top of their own database, and Tableau for organizations that want certified data sources and governed reporting, so consider your team's skills and size. A side-by-side comparison of BI tools can help you choose.

Whatever the tool, make the dictionary entry and the calculation match, and test it: recompute a known number by hand from the raw data and confirm the report agrees.

Definitions change as the business does, and uncontrolled changes are how trust in numbers erodes. Adopt a simple protocol:

• Propose in writing: what changes, why and which reports are affected.
• Owner approves, and consults anyone who relies on the metric.
• Version it. Add a new version number and date, and keep the old definition in the history.
• Decide about history. Either restate past periods using the new definition or mark a clear break in the trend, and say which.
• Announce it where people use the number, before the first report with the new figure goes out.

Review the dictionary quarterly, remove metrics no one uses and check that owners are still in their roles. For companies with a handbook and wiki, the company wiki structure guide shows where a dictionary fits, and the vendor scorecard template is a good example of a metric set that benefits from clear definitions.