Diligence in a data room
A data room is a filing system, not an answer. The work is turning several hundred documents about a handful of companies into a comparison you can defend line by line.
What a row is
A target, not a document. This is the case where bundles stop being convenient and become the whole design: each target’s information memorandum, management accounts, and the trading update that supersedes them are one subject, and they should answer as one.
Put the most authoritative document first in the bundle. Order carries precedence, so a trading update at the top is how “most recently reported” ends up meaning what you intend.
A bundle holds up to 15 documents, and a tight bundle answers better than a large one. Assemble the documents that actually speak to your columns rather than everything the target filed.
The starting columns
This is the Private Equity / M&A preset:
| Column | Type | Prompt |
|---|---|---|
| Key Valuation Multiples | Text | Extract the key valuation multiples referenced for this company (e.g. EV/EBITDA, EV/Revenue, P/E). |
| Revenue Growth | Number | What is the company's most recently reported year-over-year revenue growth rate, as a percentage? |
| Top 5 Customer Risks | List | List the top customer concentration or customer-related risks disclosed for this company. |
| ESG Compliance | Yes/No | Does this company report being compliant with applicable ESG (Environmental, Social, Governance) standards or frameworks? |
What it looks like filled in
| Row | Documents | Revenue growth | Customer risks | ESG compliance |
|---|---|---|---|---|
| Target A — IM.pdfTrading update.pdfManagement accounts.xlsx | 18.4% Trading update, p.2 | Top customer 31% of revenue; 2-year contract expiring IM, p.44 | Yes IM, p.61 | |
| Target B — IM.pdfManagement accounts.xlsx | 4.1% Accounts, p.7 | Top 5 customers 68% of revenue; no long-term contracts IM, p.29 | Not stated IM, p.55 | |
| Target C — IM.pdfManagement accounts.xlsxESG report.pdf | −2.6% Accounts, p.4 | Single public-sector framework, renewal 2027 IM, p.18 | Yes ESG report, p.3 |
Setting it up
- One workspace per process. Access is granted per workspace, so this is also the boundary you will use to keep a deal team’s reading separate from everyone else’s — see sharing.
- Upload the data room, then bundle by target with the most current document first.
- Create the table and take the Private Equity / M&A preset.
- Run one target — the one you already understand. The purpose of the first run is to test your columns against a known answer, which is the only cheap way to find out that “revenue growth” is being read off a forecast.
- Fix the prompts, then run the rest.
Two things specific to this table
Numbers need a stated basis. “Most recently reported year-over-year revenue growth” is ambiguous in a document set containing audited accounts, management accounts and a forecast. Say which, in the prompt: “from the audited accounts, not the forecast”. The column type being number makes the answer sortable, not correct.
Consider a compositional column for anything that spans documents. A question like customer concentration is answered from the whole row rather than from one document, and on a large bundle a compositional column is also the cheaper option — gathering across the documents once costs less than reading each of them separately.
Columns worth adding
- Reporting basis as a category — audited, management, unaudited forecast, Not stated. It is the column that tells you how much to trust the number beside it.
- Period end as a date, so you can see at a glance that one target’s figures are nine months older than the rest.
- Change of control consents required as yes/no, with a text column beside it. This is the diligence finding most likely to matter to the timetable.
What to check before relying on it
Every number in a screening table like this one is a claim taken from a document written to sell something. The citation is what makes that visible: it tells you the growth figure came from a trading update rather than from the audited accounts, which is a judgement you want to make yourself.
Work the low-confidence list first, and spot-check every number column on at least two targets before circulating anything.
Roughly what it costs
Eight targets, twelve documents each averaging 40 pages, is 3,840 pages — 3,840 credits, about $38, once. The columns are where a bundled table differs from a flat one: three typed columns and one list column across eight rows is roughly 48 credits at base rates, plus the bundle surcharge for the documents past the third in each bundle. The cost preview totals it before you run, which on a bundled table is worth reading rather than skipping. See how credits work.
The preset’s prompts are illustrative starting points, not financial definitions, and nothing here is investment advice.