Commercial due diligence, every number with its receipts

A commercial due diligence asks a small number of hard questions. How big is the market, and where is it growing? Who wins in it, and why? Is the target's plan credible against both? The answers end up in a report and an investment committee deck, and every figure in them will be challenged in the room.

This playbook walks through how a team runs that work in Aqmen, from the first day of the data room to the handover. The aim is the one the best firms hold themselves to: a working model on day one, the same frameworks every time, any figure walked back to its document in seconds, and a next deal that starts where this one ended.

The worked example throughout is a European HR-software market. The figures are illustrative; the method is the point.

Set up the project

A workspace is the client: its members, its data, private to the firm. Inside it, the diligence is one project, with the brief written in its notes: the questions the work must answer and the decision it informs. Inside the project, collections are folders by topic, one per workstream, so the market model, the landscape and the report each have a place before anything is built.

Workspace · the client

Northwind Group

• Commercial due diligenceProject

  Brief: is the target's growth plan credible?

  • Scope
  • Market model
  • Landscape
  • Report

• Pricing reviewProject

  Reads the market data the diligence loaded

  • Scope
  • Price tests

Members with roles, private to the firm

Two things follow from this structure. First, everyone on the team, and every assistant they connect, starts from the same brief. Second, the project is not a silo: a later pricing review in the same workspace reads the market data this diligence loaded, without copying it.

Workstream 1: research and the data room

Public statistics, analyst figures, filings and expert calls all arrive in different shapes. In Aqmen each one becomes a source, stored and cited:

• a file is kept in the workspace (the analyst report, the management presentation, the data room export);
• a link opens the page it came from (a statistics portal, a pricing page);
• a reference records what has no document: who was asked, when, and what they said, as with an expert call.

Sources · stored and cited

What the market model rests on

Source	Kind	Cited at
Eurostat, enterprise statistics 2024	Link	Table 2
Vendor price benchmarks	File	pp. 12–14
Interview, former VP Sales	Reference	45 min

A file is kept in the workspace, a link opens the page, and a reference records what has no document: who was asked, when, and what they said.

Every dataset cites where in its source

The numbers inside those sources become datasets: tables that cite the document and the place in it they came from, with units, scale and a definition on every column. Datasets arrive clean. Aqmen is not where a messy export gets fixed; it is where clean tables are kept with their receipts, so the whole team reads "employed" or "price per seat" the same way.

Data · with its sources

Firms and employment by size band

	Size band	Firms	Employed
Germany	10–49	415,261	8,130,428
Germany	50–249	65,181	6,354,318
Germany	250+	14,862	17,482,691
France	10–49	166,848	3,124,398

Cited¹Units and definitions on every column

1. 1Eurostat, enterprise statistics by size class, 2021–2024. Official statistics.

The result is that the data room stops being a folder and becomes something you can query. The same tables serve the market model, the landscape and the plan review, and they are still there for the next deal in the sector.

Workstream 2: market sizing

The market is built bottom-up, as a value driver tree, in a spreadsheet model with formulas you can read and defend. In the example the market is split by firm size, then by country, and each branch is the product of its drivers: seats, which are people employed times adoption, times a price per seat.

Exhibit · market sizing

A $7.5B market, built bottom-up

Firms of 10–49 people buy a flat plan, so their branch is Customers × Plan price. Every leaf is a dataset with its source.

1. 1Eurostat enterprise statistics, Germany, 2024.
2. 2Eurostat ICT survey, average of 2023 and 2025. A proxy: read the split, not the level.
3. 3Vendr purchase benchmark for 500+ employees, adjusted to Germany's price level.

Three practices keep the model defensible:

1. Drivers at their native grain. A statistic published every two years, or by a coarser size band than the model uses, is adapted to the model's grain with the method stated once and applied everywhere: the years between two surveys are the average of their neighbours, say, for every country alike.
2. One expression per segment where the economics differ. Small firms buy a flat plan, so their branch is customers times plan price, not seats times price per seat. Forcing one formula on every segment is how market models go quietly wrong.
3. A check that the total holds at the finest grain. The market at the top must equal the sum of its finest cells. The check sits in the model and fails loudly when it does not.

Scenarios are named sets of assumptions, held side by side rather than forked into copies of the file. A bridge between two years shows what moved the market (volume, price, region) so the growth story is drawn from the model, not written next to it.

Workstream 3: the competitive landscape

Every player goes on one grid: the same criteria for every player, scored on what matters in this market (archetype, pricing model, published price, segment focus), and a source behind every cell. Where a vendor publishes no price, the cell says so, and the figure used instead cites the benchmark it came from.

Competitive landscape

Every player on one grid

Player	Archetype	Pricing	Published price
Personio	HR suite	Per seat	Not published
HiBob	HR suite	Per seat	$6–15 a seat²
Factorial	SMB core	Per seat	From €8 a user¹
Workday	Enterprise	Contract	Median $51K a year²

A source behind every cellHeld against the market sizing

1. 1Vendor pricing pages, September 2026.
2. 2Vendr purchase benchmarks; the vendor publishes no price.

The grid is held against the market sizing. If the players' revenues add up to more than the market the model says exists, or a leader's implied share is implausible, two numbers disagree, and they disagree early, in the workspace, instead of in the committee.

Workstream 4: the business plan

The target's P&L is modelled from drivers to EBITDA: revenue as volume times price times mix, at the grain the business is actually run; history from the filings; the management plan as one scenario among others, not as the model itself.

Business plan analysis

EBITDA, 2024 to 2027

The plan assumes 14% a year. The market grows 5.7%: the plan outruns it.

1. 1Illustrative target. Base case against the management plan.

Held against the market's growth, the plan shows where it outruns it. In the example the plan assumes 14% a year in a market growing at 5.7%, so the plan depends on share gains, and the landscape grid is where those gains have to be found. A bridge between two years splits the change into volume, price, mix and costs, which is usually where the conversation with management starts.

Where did that number come from?

Every assumption in the model has an origin: the dataset it was derived from, the method that derived it, and the source that dataset cites. All of it is on the record. When someone in the room points at a figure, the answer is a path, not a memory.

Lineage · the path home

Where does $1.67B come from?

1. Market · Germany · 250+ · 2024$1.67B

   Seats × Price per seat

2. Seats15.6M

   Employed × Adoption

3. Adoption89.1%

   Average of two surveys¹

4. Price per seat$107

   Vendor band price² × price level³

5. Employed17.5M

   As published⁴

1. 1Eurostat ICT survey, 2023 and 2025.
2. 2Vendr purchase benchmarks, HiBob, 500+ employees.
3. 3Eurostat price level index, Germany, 2024.
4. 4Eurostat enterprise statistics, Germany, 2024.

In practice that means:

• the method that derived a figure is stored with the figure, not in an analyst's notes;
• a confidence is stated on every input, and flagged where it is a proxy or an estimate (adoption in the example is a proxy: read the split, not the level);
• before you change a source, you can see everything it touches.

The same question asked a year later, by someone who was never on the deal, has the same answer.

Hand over the workspace, not a file

The report and the investment committee deck still come out, in your house style. Both are built in Aqmen: a document in which every figure is a sourced block read from the model, and a deck whose charts are added by source rather than pasted. If the model moves after a figure was written, the figure is flagged as stale before it reaches the client.

But the client can also join the workspace as a guest: read the pages, open the models, see every source, and ask the next question through their own assistant, with the data and the reasoning already there.

Shared with the client deal team

How big is the market, and where is it?

$7.5B¹

HR-software market, 2024

55%²

in firms of 250+ people

+5.7%³

a year since 2021

By region, 2024 · USD billions

DACH2.9

Western2.1

Southern1.8

CEE0.4

Nordics0.3

1. 1Eurostat business statistics and ICT survey; vendor list prices and purchase benchmarks.
2. 2Bottom-up by size band. Adoption is a proxy: read the split, not the level.
3. 3Prices held at today's list, so growth is all volume.

MRClient · Partner

What does that look like for mid-sized firms in Spain?

$0.24B in 2024: 23,000 firms of 50–249 people, 1.8 million seats at $134 a seat.

From the same model · Eurostat · Vendr

The deck and the report link back to the workspace, so the follow-up question after the meeting is answered from the same model that produced the answer in it.

Chat, workbook or workspace

Most teams already use AI somewhere in this work. The difference is where the work lives once the session ends.

What it runs on

• MCP first. Claude, ChatGPT or any MCP-capable assistant works in Aqmen. Bring the one you already pay for; Aqmen gives it a workspace that keeps the work.
• Open skills. The step-by-step practice, from scoping to delivery, is an open repository you can read, fork and improve. Run the work in the app, or hand it to your assistant with the skills.
• Your formats. Spreadsheet models, documents and decks export editable, in the formats your clients use.
• Confidential. Private to your organization, never used to train models.

One honest limit: Aqmen has no live market-data feed. Statistics, prices and benchmarks come from files you load or from your assistant's research, and either way they arrive as cited sources.

If you run diligence and want to see this on one of your own markets, talk to our team.