Chapter 3 — Transforming the numbers

The moment when something doesn't add up

The work rarely starts with an adjustment. It starts with an anomaly. A P&L category behaving differently. A number stable for months, and then, suddenly, a break. 100. 100. 100. Then 150. The number is comparable. It respects the framework. And yet, it raises questions. If it's an expense, it's a problem. If it's a revenue going up, or an expense going down, there's sometimes a tendency to dig less. That's human. You know what will be challenged. You anticipate the questions to come. But this difference in treatment is already a signal.

Understanding before correcting

When a variation draws attention, you have to understand. Not in order to correct automatically. But to know what the number is really saying. So you dive into the detail. The accounting entries. The invoices received. The miscellaneous entries booked in accounting. The descriptions. The reconciliations. This work is rarely smooth. The detail is never simple to read. You have to reconstruct the path. Sometimes, the explanation is clear:

• a poorly classified entry,

• an invoice charged to the wrong perimeter,

• a business reality that isn't reflected in the standard reading.

Sometimes, everything is correct. And that conclusion is also part of the analysis.

Analysis as investigative work

This investigative work is essential. But it is neither mechanical, nor exhaustive. The volume is too large. The dimensions too numerous. The combinations infinite. So you have to choose. You look at:

• the most visible movements,

• the line items most exposed to questions,

• the perimeters where a misreading would have the greatest impact.

These are angles of analysis. Not full coverage. Every month, you have to start over.

When correcting becomes deciding

At the end of this analysis, a decision has to be made. Not:

is the number compliant?

But:

what reading enables steering?

Leave the number as it is. Or step in to improve its readability. This can mean:

• reclassifying an entry to better reflect the business reality,

• isolating an item to avoid a misleading reading,

• occasionally adjusting a presentation to clarify a message.

In every case, this isn't a simple technical fix. It is a choice of reading.

The impossibility of tracking everything

These decisions pile up. Some are one-off. Others become recurring. Ideally, you would, every month, go back over the readings previously retained in detail, check that they remain relevant, confirm that they still tell the right story. In practice, this work is done quickly. Reluctantly. Not because it would be pointless, but because it delays the analysis of the present. Going back over past readings in detail means slowing down an already constrained cycle. So you make trade-offs. You check what's essential. You secure what will be exposed. And you move on.

What shifts silently

Gradually, something shifts. The number no longer rests solely on the normalized data. It also rests on:

• past analyses,

• readings already validated,

• the experience of those who know why an adjustment was made.

The number becomes more readable. But less autonomous. It holds because someone knows.

When readability depends on interpretation

The adjustments are not structural corrections. They aren't meant to make the numbers comparable. That work has already been done. They serve to make the numbers readable, understandable, steerable. But they introduce a new dependency. The understanding of the number no longer lives solely in the system. It begins to live in the analyses and in the people who carry them. The number is ready to be aggregated. But the analytical path is already starting to narrow.