Chapter 4 - When Aggregation Destroys the Analytical Path

A situation I know well

There are scenes I've experienced more than once.
The numbers are out.
The reporting is ready.
The closing went well.
The meeting starts, the indicators scroll by, and then, at some point, something doesn’t add up.
We were ahead last month.
This month, we are behind YTD compared to the forecast.
And yet, the month that just passed wasn’t catastrophic.
The question comes.
Sometimes from the CEO.
Sometimes from an operational director under pressure regarding their objectives.

« What really explains this discrepancy? »

This is not an attack.
It’s a legitimate question.
But at that precise moment, I already know what is at stake.

Why aggregation is necessary

After these meetings, there is always a moment of reflection.
A moment where I remind myself why, despite everything, we do things this way.
The financial data, raw, is not usable.

It arrives in the form of thousands of lines, entries, isolated movements.
At this level, there is neither reading nor management possible.
If I laid out this data as it is on the table,
no one could make anything out of it.
Not management.
Not the operational staff.
Not me.

So, like all finance teams, we aggregate.
We group the entries by account.
We consolidate by entity, by supplier, by type of expense.
We transform an unmanageable volume into something readable.
It is at this moment that the P&L starts to exist.
That trends appear.
That discussion becomes possible.

I have never seen a finance team aggregate for convenience.
We aggregate because we need to take a step back.
Because we have to produce numbers within tight deadlines.
Because we need to provide a comprehensible, shareable, defensible vision.

Without this aggregation, there is no common language.
No basis for discussion.
No management.

What aggregation makes disappear

As we aggregate, we move away from detail.
This is not a conscious choice.
It’s an unintended consequence.
We move from entries to accounts.
From accounts to categories.
From categories to large aggregates.
Each step makes reading simpler.
But each step also puts a little more distance with what actually constitutes the figure.

In the moment, this is not visible.
The P&L is clear.
The totals are consistent.
Nothing seems to have been lost.
And yet.

When it comes time to understand a discrepancy in depth,
I realize that I can no longer simply trace back to the number.
The link still exists somewhere,
but it is no longer direct.
It is no longer carried by the system itself.

When the system works… until it doesn’t

As long as the questions stay within the expected framework,
the system works.
But as soon as a question strays slightly off the expected path,
I know almost immediately:
the system will not follow.
I can no longer start from the displayed number
and naturally unfold the reasoning down to the data.

So I rebuild.
I go back to the extractions.
I redo groupings.
I reapply adjustments.
Each analysis becomes a particular case.
Each question follows its own course.

The system is not wrong.
It is simply optimized to answer the questions it already knows.

Why finance continually recreates files

When the system no longer allows for exploration,
one must create a space elsewhere.

So I open a file.
I reconstruct a view tailored to the question.
This file works.
It allows for answers.
But it is specific.
It cannot be reused as is.

Over time, these files accumulate.
They hold valuable knowledge.
But fragile knowledge.

It lives in the files.
And especially in the people.

The real cost: rebuilding the analysis

The real cost is not immediately visible.
It’s the time spent rebuilding.
The mental fatigue.
The analyses that get postponed.
The questions that are no longer asked.
Gradually, autonomy declines.

I become more dependent on people,
less on the system.
And the longer this situation lasts,
the harder it becomes to question it.

What is really at stake

The problem is not aggregation.
It’s not the team.
It’s not the tool.
It’s the chain of events.

The data is aggregated to be readable.
The analysis becomes difficult to dismantle.
It moves outside the system.
Then into the people.
The numbers keep coming out.
But the capacity for exploration decreases.

The question is no longer only
how to produce the numbers,
but what the system allows — or prevents — us from understanding.