Building a KPI & Metric Dictionary

The metric dictionary is the quiet hero of reliable reporting. It records, for every metric a report contains, what the metric means, where it comes from and who owns it. With a dictionary, two people produce the same number for the same metric; without one, they improvise, and the report inherits the inconsistency.

This guide covers how to build a KPI and metric dictionary, what each entry should contain, and how to keep it current. It is the companion to Data Quality for Reporting and What Is a Single Source of Truth?.

Most “data disputes” are definition disputes in disguise. A dictionary settles them in advance.

Why a metric dictionary matters

Every recurring report raises the same questions: what does this metric include, where does the number come from, and who owns it. Without a dictionary, each question is answered by whoever is closest, and the answers drift.

A dictionary fixes that by making the answer explicit and shared. It also does three quieter things:

• It speeds reporting, because writers stop hunting for the definition.
• It enables reconciliation, because there is one definition to reconcile against.
• It supports governance, because every metric has a named owner.

It is the single most leveraged document in reporting quality.

What each entry contains

A dictionary entry is short but complete.

With entries like this, two people produce the same number for the same metric, and a reader can see exactly what a metric includes and excludes.

Prioritizing what to document first

You cannot document everything at once, and you should not try. Start with the metrics that matter most.

• Frequency: the metrics that appear in the most reports.
• Ambiguity: the metrics people already argue about.
• Consequence: the metrics a decision depends on.
• Sensitivity: the metrics where an error is most damaging.

Document the frequent, ambiguous, consequential and sensitive metrics first. A dictionary of twenty well-chosen metrics delivers most of the benefit.

Where the dictionary lives and who owns it

A dictionary only works if it is findable and maintained.

• One location, linked from the report template and the style guide.
• One owner, ideally the same person who owns the style guide, so terminology and definitions stay aligned.
• Named metric owners, accountable for their metric’s accuracy and currency.
• A review cadence, at least annually, and after any definition dispute.

Without an owner, the dictionary becomes a stale document within a year, and the disputes return.

Keeping it current

Definitions change as the business changes, and the dictionary must change with them.

• Update on change – a new product, a changed process, a new system.
• Version the changes so reports can reference which definition applied when.
• Review annually for metrics that have quietly drifted.
• Retire obsolete metrics rather than leaving them to confuse.

A dictionary that lags the business is worse than none, because it provides false confidence.

A dictionary extract

A dictionary is best shown, not described. A short extract:

• Metric: Gross margin
• Definition: (Revenue – cost of goods sold) / revenue
• Formula: (Revenue – COGS) / revenue
• Source: Finance system, monthly extract
• Owner: Finance systems lead
• Cadence: Monthly, day 3
• Caveats: Excludes one-off restructuring costs; gross, not contribution, margin

Repeat the pattern for every metric a report contains. Twenty entries like this cover most reports and settle most disputes.

Rolling out the dictionary

A dictionary only helps if it is used.

• Start with the most-reported metrics, not everything at once.
• Link it from the report template, so writers see it where they work.
• Reference it in the style guide, so terminology matches definitions.
• Review it at least annually, and after any dispute.
• Name an owner per entry, so accuracy has an owner.

The dictionary is not a one-off document; it is a maintained asset, and it improves as disputes teach you which definitions need to be sharper.

Handling derived and composite metrics

Some metrics are built from others, and the dictionary needs to say so.

• Show the components, so a change in a component is traceable.
• Record the formula, so the metric can be recomputed.
• Flag the dependencies, so a restatement of one flows to the other.
• State the exclusions, so the metric is not read more broadly than intended.

A composite metric without its components recorded is a number no one can check – and the first to be questioned when it moves.

A minimal dictionary entry

At its smallest, an entry needs six fields.

• Name – the metric as it appears in reports.
• Definition – what it means, in one sentence.
• Formula – how it is calculated.
• Source – where the data comes from.
• Owner – who is accountable for it.
• Cadence – how often and when it is produced.

Everything else – caveats, dependencies, exclusions – is an improvement on those six, not a substitute for them.

Common mistakes

• Documenting everything at once. The effort stalls, and nothing gets finished.
• Definitions without formulas. Two people interpret the definition differently.
• No owners. Metrics drift with no one accountable.
• No review cadence. The dictionary falls behind the business.
• Living in one team’s folder. Other teams never find it, so they improvise.

Frequently asked questions

What is a KPI or metric dictionary?

A record of every metric a report contains: its name, definition, formula, source, owner, cadence and caveats. It is what makes two people produce the same number for the same metric.

What should a metric dictionary entry include?

The metric name, its definition, its formula, its source system, its owner, its refresh cadence, and any caveats or exclusions.

How many metrics should you document first?

Start with twenty or fewer of the most frequent, ambiguous, consequential and sensitive metrics. A short dictionary used widely beats a long one nobody reads.

Who should own the metric dictionary?

One person, ideally the same owner as the report style guide, supported by a named owner per metric who is accountable for its accuracy and currency.

How often should the dictionary be reviewed?

At least annually, and immediately after any definition dispute or business change that affects a metric.

Should the dictionary be a spreadsheet or a system?

A spreadsheet is enough for most teams, provided it is findable and owned. A dedicated system helps at scale, but the definitions matter more than the format.

How do you handle a metric with no clear owner?

Assign one. A metric with no owner has no one responsible for its accuracy, which is exactly how definitions drift. The owner need not be senior; they need to be accountable.

How do you handle a derived metric?

Record its components, formula, dependencies and exclusions. A composite metric without its components recorded cannot be checked, and will be the first figure questioned when it moves.

How many metrics should the dictionary cover?

Start with the ten to twenty most-reported. That covers most reports and settles most disputes; extend it as new metrics recur.

Who updates the dictionary?

The metric owner, with the reporting owner coordinating. Definitions change deliberately, are versioned, and are communicated – never changed quietly inside a report.

What are the essential fields in a metric entry?

Name, definition, formula, source, owner and cadence – six fields. Everything else improves the entry but does not replace those. Add caveats and dependencies as the metric matures.

How do you know which metrics to document first?

Those that appear most often, cause the most disputes, or feed the most decisions. Frequency and friction are the best guides.

How do you handle two teams that define a metric differently?

Pick the definition that best serves the decision, record it in the dictionary, and give both teams the same term. Where two definitions are genuinely needed, give them distinct names rather than one shared name.

Should the dictionary be public within the organization?

Yes, wherever possible. A dictionary only resolves disputes if people can find it, and findability is half its value.

What is the biggest cause of metric disputes?

Undocumented definitions and unnamed owners. Where both are recorded, most disputes resolve before they start.

Next step

Build a dictionary for your twenty most-reported metrics, name an owner for each, and link it from the report template. Download the KPI & Metric Dictionary Template to start, and see Reconciling Numbers Across Reports for how the definitions settle disputes.

Sources

• Data-quality and reporting practice: controlled metric definitions and the single source of truth, applied to report production.
• Harvard Business Review (Nagle, Redman and Sammon, 2017): the prevalence of inconsistent data definitions across teams.

Good-practice claims are cited from their sources; no statistic in this article is invented.