What Investors Actually Check in a Data Room
Founders spend real time making a data room look complete — every folder populated, every checkbox ticked against a template they found online. Almost none of that effort goes into the handful of documents that actually decide whether diligence moves fast or stalls. Here is what an investor's associate opens first, and what makes them close the laptop and start asking harder questions.
I have sat on both sides of a data room — building them for founders raising, and combing through them for companies I was evaluating as an angel. The pattern is consistent enough that I can predict, within the first ten minutes of opening a room, roughly how long diligence is going to take. It has almost nothing to do with the number of files.
A data room with 40 well-organized, internally consistent documents moves faster than one with 400 that do not agree with each other. Volume is not the signal investors are reading. Coherence is.
The order an associate actually works in
Most founders build their data room in the order a checklist suggests — corporate documents first, then financials, then everything else. That is not the order it gets read in. An associate doing first-pass diligence on a term sheet they are trying to get approved internally works backward from the two things that can kill the deal fastest: the cap table and the financial model. Everything else is opened to corroborate or contradict what those two documents already implied.
That means your cap table and your model are not just deliverables. They are the lens the rest of the room gets read through. If they are clean, ambiguities elsewhere read as minor. If they are messy, the same ambiguities elsewhere read as pattern.
The cap table: where the trust is won or lost first
The specific thing an associate is checking is not "does a cap table exist." It is whether the fully diluted ownership in the cap table matches the fully diluted ownership implied by every other document in the room — the SAFE agreements, the option grants, the board consents authorizing them. When those three do not reconcile, and they frequently do not, that discrepancy becomes the first question in the diligence call, and it is rarely a comfortable one to answer live.
The most common version of this I see: a company raised two SAFE rounds at different valuation caps, granted an advisor two percent verbally before formalizing it, and the spreadsheet reflects what the founder remembers rather than what the signed documents say. None of this is fraud. It is just entropy — normal business happening faster than the paperwork could keep up. But an investor cannot tell the difference between honest entropy and something being hidden until they ask, and every minute spent asking is a minute the deal is not moving forward.
Take your cap table spreadsheet and your signed SAFE and option documents into a room separately. Reconcile them line by line, out loud, before an investor does it for you. If you cannot get to the same fully diluted number both ways, you have found the thing that will slow your raise down — and you found it on your own schedule instead of theirs.
The financial model: not the projections, the assumptions underneath them
Nobody serious believes a three-year startup revenue projection. Investors know this, and they are not diligencing the number in year three. What they are actually testing is whether the assumptions feeding the model are stated, sourced, and internally consistent — because that tells them how the founder thinks, which is a much better predictor of outcome than any specific forecast.
A model that shows its work — where a CAC assumption is footnoted to the channel it came from, where a churn assumption ties to the actual cohort data sitting two folders over in the room — reads as a founder who understands their own business quantitatively. A model with round numbers and no visible logic reads as one built the week before the raise to hit a target, because that is usually exactly what it is.
The projection is not the point. The projection is a proxy for whether you can be trusted to notice when reality diverges from plan, and to say so before an investor has to ask.
What actually stalls a deal
In order of how often I have seen each one actually delay or kill a term sheet, not how scary they sound in the abstract:
1. IP that was never assigned to the company
A co-founder or early contractor built something core to the product before signing an IP assignment agreement, or built it on a personal GitHub account under a personal license. This is the single most common finding that stops a deal outright rather than just slowing it down, because it is not a paperwork gap — it is a genuine question of who owns what the company is selling. Fixing it after the fact requires the original contributor's cooperation, and that cooperation is not guaranteed if the relationship has since soured.
2. A cap table that does not reconcile
Covered above, and worth repeating because it is the most common finding, even if it is rarely fatal on its own. It mostly costs time — days to weeks of a lawyer reconstructing the actual ownership picture from underlying documents before the deal can be papered.
3. Customer contracts with change-of-control clauses nobody flagged
Relevant mostly at Series A and beyond, when a handful of enterprise contracts represent a meaningful share of revenue. A clause letting a customer terminate on acquisition or majority ownership change is standard boilerplate that nobody reads until an investor's lawyer does, at which point it becomes a real risk on a revenue line the investor was underwriting.
4. Financials that do not reconcile to the bank
The model says one MRR number, the accounting system says another, and the bank balance implies a third. Individually each has an explanation — timing differences, a refund not yet reflected, deferred revenue treated inconsistently. Collectively they read as a company where nobody owns the numbers end to end, which is precisely the impression a raise cannot afford to give.
Building the room in the right order
Given that reading order, the building order should run the same way, not the checklist order most templates suggest.
Start with the cap table and reconcile it against every underlying instrument before you open a room to anyone. Then build the financial model with visible, footnoted assumptions rather than a clean set of numbers with no sourcing. Only after those two are solid should you turn to corporate formation documents, IP assignments, material contracts, and the rest of the checklist — because at that point you are populating a room that will already survive the first ten minutes of real scrutiny, and everything else is confirmation rather than discovery.
This is also, not coincidentally, the order in which the underlying work actually needs to happen. A cap table you build correctly from day one costs an hour a quarter to maintain. A cap table you reconstruct during diligence costs a lawyer's time, a delay, and a harder conversation with your new investor before the round has even closed.
The room is a proxy for the operation
The uncomfortable truth for founders is that a data room is not really being evaluated on its own terms. It is being read as evidence for a claim investors are trying to verify before they commit capital: is this a company run with the same rigor the pitch deck implies, or is the pitch deck the only rigorous document in the building?
A clean, coherent room does not just speed up diligence. It is itself the strongest signal you can send that the operation behind the pitch is real. Founders who understand this stop treating the data room as a compliance exercise to complete once, right before a raise, and start treating it as a byproduct of how the company keeps its own books all year — which is the only way it is ever actually clean when someone opens it.