Chapter 1 — Collecting the numbers

Before understanding, you have to gather

Before talking about analysis, before talking about consistency, before even talking about numbers, you first have to manage to gather them. This moment is rarely visible. It is rarely valued. And yet, everything begins there.

Data that exists… but not in the same place

The numbers exist. They are produced. They are recorded. They meet real constraints. But they live in different systems. Accounting, payroll, sales, banking, sometimes even files received by email.

Each source is legitimate. Each source does exactly what is asked of it. But none of them was designed to be read alongside the others.

What accounting doesn't carry

Even at the heart of finance, accounting data doesn't carry everything. A General Ledger, for instance. It is compliant. It is exhaustive. It fulfills its regulatory role. But on the P&L, there is sometimes neither customer nor supplier.

The data is correct. But it is blind to certain readings.

When detail becomes a problem

Sometimes, the detail exists. But the choice is made not to let it through. Payroll is a classic example. The detail is too high in volume. The confidentiality stakes are real. There is no desire to expose salaries in accounting. The result:

• payroll costs are aggregated in accounting,

• the detail is tracked elsewhere,

• often in a separate file.

The data exists. But it is deliberately fragmented.

Aggregating to hold up… and rebuilding elsewhere

The same mechanism appears with sales. When volumes grow, entries are aggregated in accounting. This is rational. It is necessary. But this aggregation makes the detail disappear:

• by customer,

• by product,

• by channel.

To analyze sales, you then need another source. Another export. Another file. Another system. Accounting holds up. But the analysis is displaced.

When formats become a topic in their own right

Collection doesn't just run into sources. It runs into formats. Growth through acquisition. Legacies of different accounting systems. Outsourced international accounting. Each country. Each firm. Each piece of software. And with them:

• files with or without headers,

• empty lines before the data,

• fields that don't repeat,

• column names specific to each system,

• structures in rows here, in columns there.

Nothing is wrong. Nothing is standard. But everything is different.

Data produced to be closed, not to be reused

As you move forward, an impression sets in. Accounting tools don't seem designed for the data to actually be reused downstream. The exports exist. But as if they were thought of for one-off analyses. Exceptional ones. As if, after accounting, there was nothing. This feeling often resonates with field conversations. Profiles deeply rooted in accounting sometimes find it hard to imagine that accounting data could serve any purpose other than the one it was produced for. It isn't a refusal. It is a difference in perspective.

Collection as the first point of fragility

At this stage, nothing has been analyzed yet. Nothing has been normalized yet. Nothing has been interpreted yet. But the system is already fragile.

It depends on:

• manual exports,

• heterogeneous formats,

• multiple sources,

• implicit decisions about what passes through… and what doesn't.

The data exists. But it is not yet a whole.

When gathering becomes a job in itself

Collection is not a technical step. It is the moment when you discover that the numbers were never designed to live together. Each source optimizes its own objective. And that is perfectly legitimate. But before any financial reading, before any consistency, before any analysis, you must first manage to bring a common perimeter into existence.

That is where the system begins.