Metrics

KPI definitions and a review cadence that holds

Two reports can disagree while both are technically correct: different time windows, different filters, different exclusions. The disagreement is not a reporting failure so much as a definition that was never written down. This guide covers what a metric needs before it is useful, how to choose a review cadence, and when to retire a measure.

Why the numbers disagree

  • Different windows. One report counts deals closed in the month, another counts deals created in the month and closed by now.
  • Different filters. One excludes internal accounts, test records, or renewals; the other does not.
  • Different source fields. A stage name in the CRM, a status in the finance system, and a spreadsheet maintained by hand all claim to describe the same thing.
  • Different timing. One refreshes overnight, the other on demand, so the two numbers describe different moments.
  • Different authors. Two people build similar reports independently and neither is wrong, because neither is defined.

The definition sheet

One page per measure, and one page in total if the set is small. Six fields are enough:

  1. Name and question. What decision this measure informs. A metric that informs no decision is a candidate for retirement.
  2. Formula. The arithmetic in plain words, including what is counted once and what is excluded.
  3. Source. The system and field the number comes from, so a disagreement can be traced to a place rather than an opinion.
  4. Window. The period it covers, and whether it is a point-in-time or period measure.
  5. Owner. The person accountable for the number being right and for the response when it moves.
  6. Threshold. What level triggers a conversation, and what the conversation is.

Choosing the review cadence

Three rhythms cover most operating needs, each with a different kind of decision:

  • Short cycle. Operational measures: activity, stage movement, delivery status. Decisions are about the next week — who is doing what, and what is stuck.
  • Monthly. Performance measures: conversion, margin, pipeline coverage, delivery cost. Decisions are about allocation and correction.
  • Quarterly. Strategic measures: growth, retention, capability. Decisions are about priorities and what to stop.

The test of a cadence is whether the meeting ends with decisions and owners. A review that reads the dashboard aloud and moves on is a status meeting, and the measures will drift back to being decorative.

An illustration

The example below is hypothetical and describes a pattern rather than a client account.

Retiring a measure

Measures accumulate. A metric that was useful during one phase keeps being reported after the decision it informed has been made. Retiring one is not a failure: it keeps the set small enough to be read, and it stops the monthly pack growing into something nobody reads closely. The definition sheet is the place to record what was retired and why, so it is not reintroduced by accident.

Where this fits

Definition sheets and review cadence are the substance of KPI and performance systems, one area inside the Operations & Systems capability. Where the measures come from CRM reporting, dashboard definitions are handled in the CRM consulting scopes.

KPI definitions: common questions

How many KPIs should a business run on?

Fewer than most dashboards suggest. A workable set is the handful of measures that predict the outcome you are managing, each with an owner. Where a metric has no owner and no decision attached to it, it is a number the business collects rather than one it manages by.

What if two teams need different definitions?

Then they are two measures and should be named differently. The same name carrying two formulas is what produces the meeting where the first ten minutes go on deciding which number is correct.

Does the cadence have to be weekly?

No. Match the rhythm to how fast the measure can change and what decision it informs. Operational numbers tend to need a short cycle; strategic ones move too slowly for a weekly review to say anything useful.

What do we do when a metric looks wrong?

Trace it to its source before changing the definition. A number that disagrees with expectation is either a data problem or a finding; changing the formula first loses the finding and leaves the data problem in place.

Start the Conversation

Bring us the opportunity, challenge, or next step. We'll start by understanding where you are, where you want to go, and what it will take to get there.