Checklist · PE due diligenceCheck the data before it underwrites the thesis.
35 questions in eight sections, in the order a deal runs: from the AI plays in the management case to whether the target’s systems agree with each other. Use it in the data room, in management meetings and before the final investment committee.
Read it in full on this page, tick as you go, or print it. Nothing to sign up for.
The checklist
35 questions. 7 red flags.
Tick as you go: your ticks are kept in this browser and nothing is sent anywhere. Red flags are situations to follow up before the investment committee, not automatic deal-breakers.
The AI in the thesis
Before the data room
Start from what the plan assumes. Every AI play in the CIM, management presentation or value creation plan is a claim about data.
The systems estate
Data room
Map where the business actually records customers, revenue and people, including the spreadsheets.
Cross-system reconciliation
Data room, with exports
Every system can look clean on its own. Ask for the comparisons that show whether they agree.
Red flags
- Management can’t say how many customers the business has without choosing a system.
- Closed-won deals with no matching invoice and no explanation. This is a traceability gap, not proof of lost revenue.
- Board-pack figures that only one person can reproduce.
Record quality
With exports
Look inside the systems the AI plays depend on, field by field rather than on average.
Red flags
- Key fields for a planned AI play are mostly blank.
- A recent migration with no history carried across.
Ownership and governance
Management meetings
Clean data decays without owners. Check who keeps it right.
Red flags
- Nobody owns master data, or everyone does.
- Two teams report the same metric with different definitions.
Lineage and reporting
Management meetings
If the numbers the deal rests on can’t be traced, neither can anything AI produces from them.
Play by play
Before final IC
Bring the first six sections together, one AI play at a time.
Handling the data
Throughout
Diligence on data shouldn’t create a data problem of its own.
How to use it
Three passes through the same list.
- Before the data room
Sections 01 and 02
List the AI plays and the targets that lean on them. Map the systems. This tells you which exports to ask for.
- With exports
Sections 03 and 04
Compare the systems with each other and look inside the ones the AI plays depend on. This is where the surprises are, and the slow part by hand.
- Before final IC
Sections 05 to 08
Ownership, lineage and the play-by-play view: what is ready, what needs fixing first, and what the 100-day plan should do before the AI spend.
Where we fit. A Truth Score Read runs sections 03, 04 and 07 on the target’s exports, under NDA. Our engine matches the customers, people, deals and invoices across them, every red flag is checked by hand against sample records, and you get a signed memo within 5 working days of complete exports. One company first; then every company in the portfolio on the same method.
Data diligence, before the deal See a real sample memo How exports are handled
Sections 03 and 04
Running this on a live deal? Tell us the shape of it.
We’ll reply with which of the target’s exports sections 03 and 04 need, and how a data diligence read would sit alongside financial and commercial diligence. The checklist is yours to use either way.
- A reply from hello@getdatatruth.com, not a sales sequence
- The export list for the target’s systems
- Timing against your IC date
Questions
About the checklist.
Who it is for, why it has no thresholds, and where DataTruth fits in a live process.
Book the callWho is this checklist for?
Deal teams and operating partners underwriting a company where AI is part of the value creation plan, and the management teams preparing for that diligence. It sits alongside financial and commercial diligence; it doesn’t replace either.
Why are there no numeric thresholds, like a maximum duplicate rate?
Because a threshold without a use case is guesswork. A level of duplicates that doesn’t matter for one AI play can block another. The checklist asks what the data needs to support and whether it does.
How long does it take to work through?
Sections 01, 02, 05 and 06 are questions for management and the data room. Sections 03 and 04 need exports and someone to compare them; that is the slow part if you do it by hand.
Can DataTruth run sections 03, 04 and 07 for us?
Yes. A Truth Score Read compares exports from two or more of the target’s systems and returns a signed memo within 5 working days of complete exports. It works under NDA alongside the rest of diligence. We sell no AI, no software and no fixes, so nothing we could sell the target afterwards depends on the answer.
Can you take several targets or portfolio companies at once?
We take a limited number of reads at a time, so every red flag is checked by hand. For a portfolio we agree a sequence, one or two companies in parallel, each memo within 5 working days of complete exports.
Sections 03 and 04, done in 5 working days.
The target’s exports, compared record by record, with a signed memo for the investment committee. From someone with nothing to sell the target afterwards.
Starts with a 30-minute call. Commits you to nothing.