Due Diligence

Investment decisions used to run on conversation and impression. A founder said revenue was at a certain level, the patent was filed, the lead engineer was staying, and the other side decided whether to believe it. The trouble wasn't dishonesty so much as the fact that founders frequently don't know their own company precisely.

The gaps showed up after the money moved. Monthly revenue described as $60,000 turned out to be half one-time implementation fees, with only $30,000 actually recurring. A co-founder who left in the first year still held 15 percent of the company, and that surfaced after the wire cleared. At that point the investor owns a stake in a different company than the one they priced, and the only remaining instrument is a lawsuit.

So a step got inserted into the sequence: after the terms are agreed and before the money moves, everything said out loud gets checked against documents. Due diligence isn't a trial of the company. It's a comparison between a decision already leaning one way and the evidence it was supposedly built on. Strip that comparison out and investing goes back to being a contest of impressions.

It starts once the terms are agreed and usually runs three to eight weeks, in five stages.

1. Sign a confidentiality agreement and open a data room — a single folder holding every file the company hands over, visible only to named people on the investor's side. Trading documents one at a time over email leaves nobody sure what's already been sent.
2. Receive the request list. It typically arrives as a table of 60 to 150 line items, each naming a document and a date. Plenty of companies can't fill half of it in the first week, not because the records don't exist but because they're scattered across five tools.
3. Work through the items by area. Finance ties revenue in the books to money actually received and recalculates annual recurring revenue (ARR) — revenue expressed on a yearly basis — against the definition the company used. Legal reads the cap table, the record of who owns what share of the company, along with major contracts. People checks employment agreements and stock option grants for key staff; engineering checks who owns the code and what outside licenses it depends on. The most frequent mismatch found here is between the numbers in the investor deck, the presentation used to explain the company, and the ledger.
4. Questions come back on whatever doesn't line up. Most get resolved with an explanation. Some become conditions to fix before signing, like papering a customer relationship that had been running without a contract.
5. Fold the findings into the agreement. Clean results mean signing at the agreed terms. Findings mean a lower price or added representations and warranties — the company's formal promise that what it stated is true, with liability attached if it later isn't.

Institutional investors run this harder because the money isn't theirs. A venture capital (VC) firm invests funds raised from pension plans and corporations, and it eventually has to explain each decision to those backers on paper.

It belongs in front of decisions that are large and hard to reverse. A Series A round worth several million dollars, a merger or acquisition that buys a whole company, and the security questionnaire a large enterprise sends before adopting a vendor are all versions of the same check. At the other end, an angel writing a check worth tens of thousands of dollars usually skips almost all of it, because the legal fees and weeks involved grow out of proportion to the amount at stake.

The limit is that only what exists on paper can be read. Contracts and ledgers are verifiable; whether the lead engineer plans to leave in six months, or whether the largest customer is quietly weighing a switch, isn't written anywhere. That's why a separate step of calling a handful of customers directly usually gets added. Even then intent stays unverifiable, so investors tend to manufacture reasons for key people to stay through the contract instead.

Time is a cost of its own. A founder and a finance lead spending six weeks assembling documents means six weeks of hiring and selling that didn't happen. Companies that keep the data room current compress the stretch to three weeks. Companies that don't spend eight and still get the price trimmed over documents they couldn't produce.

"We have the term sheet, so diligence starts this week. Get them data room access first."

Terms are settled and the document-checking phase is opening.

"Diligence flagged the cap table — the departing employee's shares were never cleaned up."

A gap between paperwork and description that has to be fixed before signing.

"Their request list has 120 items and we have one person on it, so realistically six weeks."

A warning that the volume of requests exceeds what the company can absorb on the proposed timeline.

When Pagedeck raised its Series A, this was what took most of Grace's time. Outside accountants and lawyers handled the finance and legal items on a 110-line list, but the product questions came to Mia: how user counts are defined, at what moment the free-to-paid conversion rate is measured, which cohorts are excluded and why.

The item that ate the most calendar wasn't a number. It was an early contract with the agency that built the first version, which never assigned ownership of the resulting code. Getting that one clause signed retroactively took two weeks.

Many people treat diligence as a formality — terms are agreed, so what's left is stamping paper. In practice this is exactly where deals change shape or die. Valuations get revised downward after findings. An unresolved ownership issue can push a close out by months while the company goes and fixes it. Until signatures exist, nothing is settled.

The other misreading is that diligence exists to find the company's faults. With that volume of questions and document requests, feeling investigated is a natural reaction. But what the other side measures isn't wrongdoing, it's the distance between statements and records. Problems disclosed up front with a cleanup plan attached usually pass. A small problem that was hidden and then discovered sends every other answer back through verification. Concealment is what makes the process long, not the problems themselves.