A metric dictionary: stop debating what the numbers mean

Start with the arguments you already have

Collect the recurring disagreements from commercial meetings. Which number gets challenged? Which team produces a different total? Where does a manager ask what a label really includes? These are better starting points than attempting to document every field in every system.

Choose a short list of commercially important measures. For each, identify the business question and the people using it. “New customer” may mean a first paid order to finance and a newly created account to marketing. Both events can be useful, but they need different labels if people are expected to make consistent decisions from them.

Define more than the calculation

Record the name, plain-language meaning, inclusion rules, exclusions, unit, time basis and source. Name an owner who can resolve ambiguous cases. A formula alone cannot explain whether cancelled orders are included or whether the date refers to order creation, payment or fulfilment.

The UK Government Data Quality Framework emphasises fitness for purpose. Apply that idea to definitions: a metric is useful when it fits the decision being made. Do not force one number to serve incompatible purposes. A sales activity report and a finance reconciliation can legitimately show different values when their definitions are explicit and their relationship is understood.

A hypothetical qualified-lead definition

Imagine a B2B provider that wants to report qualified leads. Its first definition is simply “a good enquiry”, which invites inconsistent judgement. A more usable definition might require a relevant organisation, a need within the service scope and a recorded next step accepted by sales. The business must decide whether those conditions are appropriate.

The dictionary also records when qualification occurs and what happens if an enquiry is later rejected. This is an illustrative definition, not a universal sales standard. The point is to make judgement reviewable. Teams can then discuss whether the criteria are helpful instead of silently counting different populations under the same heading.

Check scope before combining systems

A website event, a CRM contact and an invoice represent different kinds of records. One customer may generate several events, several contacts may belong to one organisation and one opportunity may create multiple invoices. Joining those records without understanding their relationships can produce convincing but incorrect totals.

Google Analytics’ data-compatibility documentation shows that not every dimension and metric combination is valid even within a single platform. Across systems, the need for care is greater. Include record grain in the dictionary: one row per event, person, organisation, opportunity or transaction. Explain the matching keys and how duplicates or unmatched records are treated.

Use a minimum definition template

Keep the document readable enough for business owners to review. Technical details can sit in a linked implementation note, but the commercial meaning must be understandable without reading a query.

• Metric name and the decision it supports.
• Plain-language definition and calculation.
• Included and excluded records, with examples.
• Unit, currency where relevant and reporting timezone.
• Date used to assign the reporting period.
• Source system, record grain and matching rules.
• Refresh timing and known limitations.
• Business owner, technical owner and approval date.
• Change history and the treatment of historical comparisons.

Govern changes without freezing learning

Definitions will evolve. A company may introduce a new qualification stage or separate subscription income from project work. Decide whether a change will restate history, create a new series or mark a break in comparison. Quietly changing the formula while keeping the chart title is the most confusing option.

There is a tradeoff between a highly detailed dictionary and one people actually maintain. Start with the measures that drive important decisions and expand as needed. Review definitions when systems or commercial processes change. The goal is not bureaucratic perfection. It is shared language that lets a meeting move from “whose number is right?” to “what should we do next?”