Who Decides What the Number Means? Write It Down Before AI Does

Emily Goh ยท 2026-11-22

It is Monday morning and the CEO forwards an AI summary to the leadership channel. "Pipeline coverage is 2.1x, down from 3.4x last month." The sales leader replies within minutes: coverage is fine, the AI must be counting it wrong. Finance says the 3.4x was never right in the first place. Marketing asks which pipeline the summary even used. By lunch there are four threads and no decision.

The AI did exactly what it was asked. It picked a definition of pipeline coverage, because nobody had written one down.

So who decides what a metric means? A named person, usually the leader whose team is judged by that number, decides, and the decision gets written down in one shared place. That is metric definition governance: every important metric has a written definition, an owner who approves it and a record of every change. Do it before AI tools start answering questions about your data. If you do not, the AI will choose a meaning for you, and it will not tell anyone which one it chose.

AI does not argue. It assumes.

When two people disagree about a number, at least the disagreement is visible. Someone says "that is not how we count it" and the meeting stops.

An AI assistant does not stop. Ask it for pipeline coverage and it will find a pipeline field, a quota field and a date range, then divide one by the other. It might include deals in the first stage. It might count renewals. It might use this quarter's quota or the full year's. Each of those choices is reasonable on its own. Together they produce a number that no team recognises, delivered in a confident paragraph with a chart.

This is the risk we covered in what happens when an AI copilot briefs the CEO on bad data. The data does not have to be dirty for the answer to be wrong. The definition only has to be missing.

What "write it down" actually means

Most companies have definitions. They live in a slide from last year's kickoff, a comment in a dashboard, a Slack thread and the head of a long-serving analyst. That is not written down. That is scattered.

A usable definition answers five questions in plain words:

  1. What it measures. "Open pipeline divided by remaining quota for the current quarter."
  2. What counts and what does not. "New business only. Excludes renewals and deals in the Discovery stage."
  3. Where it comes from. "Opportunity amount and stage from the CRM, quota from the Finance plan."
  4. Who owns it. One named person who approves changes.
  5. When it last changed and why. "Discovery stage removed on 3 November, approved by the VP of Sales."

We showed how much a single missing line can cost in the definition nobody writes down. Closed Won is a good place to start, but the same five questions work for MQL, pipeline, churn, ARR and anything else that ends up on a board slide.

Who should own the definition

The owner is not the person who builds the report. It is the person who has to stand behind the number when someone challenges it.

Pipeline coverage usually belongs to the sales leader. MQL usually belongs to the marketing leader. Revenue and ARR usually belong to Finance. RevOps often holds the pen, writes the drafts and keeps the list tidy, but RevOps should not be the one quietly deciding what Sales means by pipeline.

Two teams can still use different definitions of the same word. Marketing and Sales may both talk about qualified leads and mean different stages. That is allowed, as long as both versions are written, both have owners and every report says which one it uses. What is not allowed is a definition nobody owns, because then nobody can say whether the number is right. That is the heart of business data accountability: every number traces back to a rule, an owner and the records behind it.

Why it has to happen before AI, not after

You could wait until the AI gets something wrong and fix it then. The trouble is that you may never notice.

A wrong number in a spreadsheet tends to get caught because people know the spreadsheet is fragile. A wrong number in a polished AI summary tends to get forwarded. It reads well, it arrives fast and it looks finished. By the time someone questions it, it may already be in a board pack or a hiring plan.

Writing definitions first also makes the AI more useful. If the agreed definition of pipeline coverage sits in a document the tool can read, the answer starts from your rule instead of a guess. And when someone asks "how was this calculated?", there is a real answer: this definition, approved by this person, on this date.

What you can do this week

  1. List your top five metrics. The ones that show up in leadership meetings or board slides. Say pipeline, pipeline coverage, MQL, Closed Won and ARR.
  2. Find every current definition. Search old decks, dashboards and Slack for each one. Paste what you find side by side. Do not fix anything yet.
  3. Name one owner per metric. Ask the leader directly: "When this number is challenged, are you the person who answers for it?"
  4. Draft the five answers. For each metric, write what it measures, what counts, where it comes from, who owns it and when it last changed. Keep it to one page.
  5. Get one approval. Send the draft for a single metric to its owner and ask them to approve or correct it. One approved definition beats five drafts.
  6. Tell your AI tools where the definitions live. If your assistant can read documents, point it at the page. If it cannot, add the definition to the top of any AI-generated report.

Where Gohvio fits

Gohvio is built around written, owned definitions. You upload your own guardrails, the definitions your teams actually use, as a document. A person approves them, and each change is recorded with who changed it and when. Gohvio then checks your CRM data against those guardrails and shows what is broken, what it does to pipeline and forecast, and the rows behind every finding. It is read-only, so it never changes your CRM.

That means when a number is questioned, you can point to the rule, the person who approved it and the records that do or do not follow it.

Stop debating your data. Start making evidence-based decisions.

Frequently asked questions

What is metric definition governance?

It means every important metric has a written definition, one named owner who approves changes, and a record of each change with who made it and when. Everyone, including AI tools, works from the same written rule.

Who should own a metric definition?

The leader who has to stand behind the number when it is challenged. Pipeline coverage usually belongs to Sales, MQL to Marketing, and revenue or ARR to Finance. RevOps can draft and maintain the definitions, but should not decide them alone.

Why write metric definitions before using AI on CRM data?

AI tools do not ask which definition you mean. They pick one and present the answer confidently. Without a written definition, nobody can tell whether the AI's number is right.

What should a metric definition include?

Five things: what it measures, what counts and what does not, where the data comes from, who owns it, and when it last changed and why.