Where does this number come from? Minimal traceability for your reports

A number nobody can trace back to its source is not data: it is a claim. It may well be correct — it often is — but it rests on trust in a person, not on a path you can show. The decision in this article is a concrete one: what it takes — and what it does not take — for any number that supports a decision to be traceable from the report all the way back to its origin, without launching a technical project to get there.

What tracing a number actually means

Every figure in a report has made a journey: it left a system, was extracted at a specific moment, got filtered, was combined with other things, went through a formula and landed in a cell. Tracing it means being able to walk that journey backwards, and it comes down to four questions:

• Which system did it come from, and through which query or export?
• When was it extracted?
• What was filtered out or excluded along the way?
• Which formula or rule produced it?

If any of the four has no written answer, the number depends on the memory of whoever built it. Note what this is not about: whether "sales" includes credit notes is a different conversation, the one about definitions. This is more basic — being able to show the path the number travelled, whatever it means.

Where the path breaks

The path of a number almost never breaks through bad faith. It breaks through the accumulation of manual steps nobody sees:

• Intermediate copies. The data gets exported to a spreadsheet, tweaked "for a second", and the tweak is recorded nowhere. The final figure no longer matches any source.
• Filters from memory. "I always remove the test orders" is a perfectly reasonable rule that lives in one person's head. The day someone else builds the report, the number changes and nobody knows why.
• Parallel versions. Once the file is called report_final_v3_GOOD, the question of which version fed the decision made two months ago has no answer left.
• Chains of people. One person extracts, another combines, a third presents. Each sees their own stretch; nobody sees the whole path.

Each of these steps looks harmless on its own. Together, they make the honest answer to "where does this number come from?" become "you'd have to ask the person who built it". That sentence is the tell: there is no traceability.

The minimal record for a number

The fix does not require new software. For every report that supports decisions, a short record is enough — a shared document sitting next to the report itself works fine — answering the four questions:

• Source: the system and the specific query or export each figure comes from.
• Moment: date and time of the extraction.
• Filters: all of them, including the "obvious" ones nobody writes down because "that's how it's always done".
• Calculation: the formula or rule, written in a sentence someone who didn't build it can understand.
• Author: who produced the figure, for whatever the record doesn't cover.

The record gets filled in while the report is being built, not afterwards. Reconstructing from memory weeks later is precisely the problem the record exists to prevent. And if filling it in takes more than a few minutes, that is information too: it means even the person building the report doesn't quite know where their numbers come from.

Fewer steps, not more documentation

The temptation at this point is to document everything. That is a mistake: every manual step in the path is a point where it can break, and documenting ten fragile steps does not make them any less fragile. The second move, as important as the record, is to shorten the path. If a figure passes through three spreadsheets before reaching the report, the question is not how to document all three — it is which ones are actually needed. Every intermediate copy you remove is a part of the record you no longer have to maintain, and an error that can no longer happen.

The ideal version — the figure travels from the source system to the report without hands in between — is a destination, not a requirement. Minimal traceability works in the real world of exports and spreadsheets too: it only asks that every step be written down.

Where to start, without a project

You do not need to chase every number in the company. Traceability is owed to the figures that support decisions: the ones quoted in the monthly meeting, the ones that trigger a purchase, a hire or a price change. The rest can wait.

The first step fits in an afternoon: pick the most-quoted number in the business and trace it end to end, today. The exercise teaches more than any desk review, because it exposes the intermediate copies, the filters living in someone's memory, and the stretches only one person knows. Then write the record for that figure and its report neighbours, and set one simple house rule: no number goes into the management report without its record. The rule enforces itself, because whoever presents a number without a path is the one who has to defend it.

Three questions to know whether you can trust a figure

1. Can someone trace this number end to end without calling the person who built it?
2. Are the filters and the calculation written down, or do they live in someone's memory?
3. If two people calculated it independently, would they land on the same number?

When the answer to all three is yes, discussing the business gets easier: the conversation is about what to do, not about whose number is the right one. Tracing the path of critical figures is, in fact, one of the first things we look at in a data & systems assessment: it says a great deal about how information flows — or fails to flow — through a company.