Why the same number has three different answers

This is a placeholder draft. The structure and argument are here as a starting point — rewrite it in your own voice with your own examples before publishing. Set draft: false in the meta block above when it is ready.

---

It usually surfaces in a meeting. Someone quotes a figure, someone else has a different one, and the next twenty minutes go on establishing whose spreadsheet is right instead of on the decision the meeting was called to make.

The instinct is to find the error. Usually there isn't one.

Three answers, three questions

Trace each figure back and you tend to find that they are measuring genuinely different things that have come to share a name:

• Different cut-off. One counts to the last day of the month, one to the last completed cycle, one to whenever the export ran.
• Different inclusion rule. One includes internal transfers, one nets them off, one excludes anything still marked provisional.
• Different correction policy. One is restated when late records arrive, one is frozen at publication, one is silently overwritten.

None of those is a mistake. They are three reasonable answers to three questions nobody wrote down.

Why it does not fix itself

Each version has a constituency. The operations figure is the one the site managers trust because they can see where it comes from. The finance figure is the one that survives audit. The board figure is the one that has been reported externally and cannot now be changed without explanation.

So the argument recurs monthly, is never resolved, and quietly costs more than any of the systems involved.

What actually resolves it

Not a new dashboard. A dashboard built on top of the disagreement inherits it, and adds a fourth figure.

1. Pick the metric that causes the most argument. One. Not a governance programme.
2. Write down the definition — cut-off, inclusions, exclusions, correction policy — in language the people arguing would both recognise.
3. Agree who owns it. One named person who can change the definition, and a record of when it changed.
4. Compute it once, in one place, and have every report read from there.
5. Keep the other versions available and labelled, rather than deleting them. People stop trusting a single figure that appeared overnight and made their familiar one vanish.

Then repeat for the next metric. This is slower than buying a tool and it is the part that determines whether the tool would have worked.

What good looks like

• Two people quoting the same figure without checking which file it came from.
• A definition someone can point to when a number is questioned.
• Meetings that start at the decision rather than at the reconciliation.
• Any new reporting or model work standing on figures that already survived this argument.