The data room mistake to avoid above all is giving an investor more than one answer to the same question. Before you open the room, pick one number for each question and mark every older version as old.
The company was raising its Series B. It's an energy-finance company, and it had been growing for several years. It asked us to rebuild its financial model and its pitch deck. It also sent its data room: 436 documents.
We read all of it before we touched the model. Then we wrote 280 questions an investor asks at Series B and answered each one from the documents, with the page the answer came from.
The problem: the room argued with itself
The answers didn't agree. The size of the raise showed up as five different figures across eight months of papers. The last funding round was described with two numbers much bigger than the real one. One was the company's value after the round. The other was all the money it had taken in since it began, grants and loans included. Both were true of something. Neither was the round.
There were three revenue forecasts in circulation, and the board hadn't approved any of them as the one investors should see.
Some errors were harder to spot. On the group balance sheet, an amount that should have been subtracted had been added. Assets and liabilities went up by the same amount, so the sheet still balanced. That's why nobody had caught it. A grant schedule multiplied each balance by its number of payments. The year-end workbook gave a different net result on each of its main statements.
None of this meant the business was bad. It meant the company had grown for years and kept every version of every file. But an investor's analyst reads the room the same way we did, and once they find one contradiction, they doubt the next document they open.
The room disagreed with the founder too
A board pack in the room gave one figure for the raise. The founder told us some of those files were earlier versions that were never updated. So the biggest number in the room was out of date, in the company's own files. We moved the model to the founder's figure and marked the older papers as replaced.
What we did: one number per question
We wrote a contradiction log. Each line names the two sources that disagree, what the conflict is, which one is most likely right and whether a founder has to confirm it. The conflicts that had to be settled before any investor saw the deck went to the founders as direct questions, each with its page references.
Then we wrote a context file the team could load into its own AI assistant. Every figure in it carries a label. Canonical means use it. Superseded means it's old, and the assistant has to say so if it quotes it. Open means nobody knows yet, and the assistant should say that instead of guessing.
The deck ran on one rule. A number goes on a slide only with a source a reader can open, and that source has to be about the same thing as the slide. The deck grew to 159 slides. Every old version moved to the back and nothing was deleted, so anyone could see what had changed.
In the old model, sales volumes had been typed in as fixed numbers. We built supporting schedules beside the model, so each total comes from its own reasoning and an investor can trace it back.
What we built
- A 280-question review of a 436-document data room
- A contradiction log with a recommended source for each conflict
- A context file that tells an AI assistant which number to use
- A rebuilt model and a 159-slide deck where every number has a source
Why it matters for your raise
Investors don't expect a growing company to have tidy files. They do expect the founder to know which number is right. If your room gives five answers to "how much are you raising?", the investor picks one, and it may not be yours.
Before you open your data room, pick one number for each question an investor will ask, and mark every older version as old.
Questions founders ask
What are the most common data room mistakes?
Conflicting numbers are the big one. The raise size, the last round or the revenue forecast shows up as different figures in different files. Others are old versions left in with no label, errors that hide because the sheet still balances, and files that disagree with what the founder says. Once an analyst finds one contradiction, they doubt the next document they open.
Does it matter which documents are in a data room?
Less than whether they agree. A typical room holds company and legal documents, the cap table and past rounds, financial statements, the financial model, the pitch deck and key contracts. For the full list, see our investor data room checklist. What trips companies up is consistency. Every figure that appears in more than one place should match, and anything that's been replaced should be marked as old so nobody quotes it.
How do you keep a data room consistent with your pitch deck?
Put a number on a slide only if it has a source the reader can open, and that source is about the same thing as the slide. Keep a log of conflicts between files, with the one to trust. Label every figure as current, replaced or still open. Move old versions to the back instead of deleting them, so anyone can see what changed.