Reading a data room fast: distilling due diligence into a decision

A data room is the set of files a seller shares so a buyer can check a deal before committing. Financials, contracts, customer lists, cap tables, employment terms, legal correspondence, tax filings. On a small deal that is a few dozen documents. On a real one it is hundreds, often thousands, dumped into folders named by whoever uploaded them.

You have limited time and a decision to make. Invest, acquire, walk away, or proceed on conditions. The failure mode is obvious and common: people read top to bottom, burn the clock on folder one and skim the rest. The risk that kills the deal is rarely in folder one. This is how you read a room fast and come out with a call you can stand behind.

Triage by risk, not by folder order

Before you open a single file, write down the one thing that would make you walk. For a SaaS business it might be that growth is bought, not earned. For a services firm it might be that one client is half the revenue. For a manufacturer it might be that the margins are propped up by a contract that ends next year. That is your first stop.

Order your reading by how badly each area could hurt you, not by how the seller arranged the files. A clean legal folder does not save a deal with hollow revenue. Go where a problem would sink the deal first, confirm or kill that risk, then move to the next. If the deal-breaker is real, you have saved yourself days. If it is clean, you read the rest with a calmer head.

The single most useful question in a data room is not "what is here?" It is "what would have to be true for this deal to be a mistake?" Read to answer that first.

The numbers that actually matter

Most data room financials are a wall of figures. A handful of them carry the decision. Spend your time here.

• Revenue quality and concentration. Is the revenue recurring or one-off? Contracted or hopeful? And how spread is it. If the top five customers are most of the book, the business is only as stable as those five relationships.
• Margins and the trend. A single year of gross and operating margin tells you little. The direction over three years tells you a lot. Improving margins on flat revenue is a different business from falling margins on fast growth.
• Cash and working capital. Profit is an opinion, cash is a fact. Look at how cash moves through the year, how much is tied up in receivables and inventory and whether the business funds its own growth or leans on the next raise.
• Quality of earnings and one-off adjustments. Sellers present an adjusted profit number with add-backs: legal costs, a founder salary, a "one-time" expense that appears three years running. Strip the add-backs that are not genuinely one-off and see what real, repeatable earnings look like.

You do not need to rebuild the seller's model. You need to recompute the four or five figures the decision hangs on, from the underlying files, and check they say what the summary says. The gap between the headline number and the recomputed one is often where the deal lives.

The red flags to scan for

Some patterns recur across bad deals. Scan for them early, because they are cheap to check and expensive to miss.

1. 1.Customer concentration. One or two accounts carrying the revenue, especially if a contract is up for renewal soon.
2. 2.Declining cohorts. New customers spend less, or churn faster, than the ones before them. Growth can hide this for a while. The cohort data does not.
3. 3.Related-party items. Revenue or costs flowing to entities the owners control. These can inflate the picture and rarely survive the change of ownership.
4. 4.Restated figures. Numbers that changed between one document and another with no explanation. Either the earlier version was wrong or the later one is convenient. Both need a reason.
5. 5.Gaps in the data. The missing month, the contract referenced but not uploaded, the segment that is never broken out. What is not in the room is as telling as what is.

None of these is automatically fatal. A concentrated customer base with twenty year relationships and renewed contracts is a different risk from concentration with year to year deals. The flag tells you where to dig, not what to conclude.

Separate what you can verify from what goes on the list

You will not resolve everything from the room. The skill is knowing the difference between what you can verify now and what needs the seller, an expert, or more time.

Verify now: anything the data supports. Recompute the margins, check the concentration, trace the adjusted earnings back to the raw figures, reconcile two documents that should agree. If the files answer it, answer it now while it is in front of you.

Follow-up list: anything the data raises but cannot settle. The unexplained restatement, the missing month, the contract you cannot see, the cohort that needs the seller's own breakdown. Keep this list tight and specific. A vague "tell us more about churn" gets a vague answer. "Why did 2023 revenue restate down by 4 percent between the April and July packs?" gets a real one.

Turn it into a decision, not a summary

The work ends in a decision, not a document. A hundred page summary that restates the data room is a way of avoiding the call. The output should be short and it should commit.

State the recommendation. Proceed, proceed on conditions, or walk. Then the few things that drove it: the two or three numbers that matter, the risks you found and what would change your mind. If you are proceeding on conditions, name the conditions precisely. Confirmed renewal of the top two contracts. A clean cohort breakdown that holds up. An explanation for the restatement. Conditions turn a soft maybe into a firm yes with guardrails.

A good diligence output fits on a page or two: the call, the numbers behind it, the risks and the conditions. The deeper work sits underneath, ready if anyone asks.

How Nexlyr AI reads the room fast

The method above is the job. The slow part is the mechanical work: opening hundreds of files, pulling the figures out, recomputing them and reconciling documents that should agree. That is where Nexlyr AI fits.

You give it the files from the room, the spreadsheets, PDFs, contracts and prior packs and a short brief of what the deal is and what would worry you. It reads them and computes the figures that matter straight from the source. It writes the calculations and runs them against your actual data, so the concentration, the margin trend and the recomputed earnings come from the files, not from a guess. If the data does not support a figure, it does not show one. That is the difference between a summary you have to re-check and numbers you can act on, the operational trust of knowing the figure came from your own files.

It traces each number back to the file it came from, so when a figure looks off you can open the source and check it yourself. It flags the gaps, the missing months and the things the data raises but cannot settle, which is most of your follow-up list built for you. And once the work is assembled it runs a "think further" pass over the finished case, like an analyst reviewing a colleague, raising the questions, risks and what-ifs you would want a second pair of eyes to catch before you commit. The output is an editable deck you can take into the deal discussion, not a hundred page file nobody opens.

It does not make the call for you. Diligence judgement is yours. It does the reading and the arithmetic fast and grounded, so you spend your time on the decision instead of the data entry.

Putting it together

Read by risk. Recompute the few numbers that decide it. Scan for the red flags. Split what you can verify from what needs a follow-up. End on a decision with conditions, kept short. Do that under time pressure on a thousand file room and the bottleneck is no longer your judgement, it is how fast you can get to the figures you can trust. Closing that gap is the whole point.

Keep exploring

• Why AI tools invent numbers, and how source-grounding stops it
• From a raw spreadsheet to a finished analysis
• Due diligence summary