Three days before signing, the buyer’s counsel asked for the 2022 supplier agreement. It took nine hours to find. The deal slid a week.
That’s the whole story, really. Every deal blow-up I’ve watched up close started the same way: a document that existed somewhere, owned by nobody, filed under a folder name that made sense to exactly one person. If the buyer’s team can’t find what they asked for, they don’t assume it’s filed elsewhere. They assume it doesn’t exist. So they send a fresh request list, which lands on your desk as a second round of work you already did once.
The folder tree is the fix, and it works best when the whole file lives in one controlled environment rather than five shared drives with five naming conventions. Teams that run diligence out of a tool like datarooms.jp tend to spot this fast, because the first thing a seller has to do inside a data room is build the index that questions will be mapped against. Build that index early and you spend the last two weeks answering questions. Build it late and you spend them hunting for files.
Why a bad folder tree costs more than a bad filing cabinet
Here’s the part that surprises people. Physical filing punished you once. You lost one document, in one room, and somebody walked over and pulled it. A folder tree in a shared system punishes you on a delay. You file something in the wrong branch in March, the buyer’s analyst pulls two versions of the same contract in June, and now the buyer’s counsel is asking which one is the operative agreement. That question doesn’t cost you five minutes. It costs you a red flag on the deal summary, and red flags get priced.
I sat through a sale where the seller’s folder for “HR” contained forty-one files across three sub-branches, two of which had nearly identical names. The buyer’s team eventually stopped browsing and just started requesting documents by name, one by one. That’s the failure mode. When your structure stops being navigable, the buyer’s answer is not to try harder. The buyer’s answer is to make you do the work in request form.
Build the tree around questions, not around your business
Most sellers organize the way the company is organized: by department, then by year, then by whatever the department head felt like that quarter. It reads sensibly from the inside. It reads terribly from the outside.
A buyer doesn’t browse your org chart. A buyer works a due diligence request list, which is a list of questions. So the tree should look like the list. I’d argue this is the single highest-leverage change a seller can make, and I’d pick it over any software feature on the market.
The four-branch skeleton
Start with a small number of top-level branches, and resist adding a fifth because something felt orphaned. Four covers nearly every mid-market sale:
- Corporate: formation documents, ownership records, board consents, cap table history.
- Financial: statements, tax filings, debt agreements, receivables aging.
- Commercial: customer contracts, supplier agreements, pricing, pipeline records.
- Legal and Risk: litigation, IP, insurance, employment agreements, regulatory correspondence.
Two levels deep is plenty. Go deeper and you’ve built a maze. If a file doesn’t fit, the problem is usually a missing branch at level two, not a need for level four.
File names are the search engine nobody sets up
Search inside a data room is only as good as the text it can see. Filenames are text. So name files the way a buyer would type the request.
My rule: format plus counterparty plus date plus version, in that order. So a supplier agreement with a parts vendor signed in March 2023 becomes “Agreement – Nordvik Parts – 2023-03 – v2.” Boring, consistent, findable. The alternative is a folder full of “final_final_v3 (1).pdf,” which is how a nine-hour search happens.
Dates go in year-month-day order. Not because it’s elegant, but because it sorts correctly when someone scans a list, and because mixed date formats are a genuine source of confusion when your team and the buyer’s team write dates differently. Record-keeping conventions like this are exactly what the National Institute of Standards and Technology covers in its work on information and data management practice, and that’s a useful sanity check when you’re deciding how disciplined to be.
The permission map you should draw before anyone logs in
Here’s where folder structure and access control collide, and this is the part most sellers skip.
Draw a simple grid. List the top-level branches down the left. List your audiences across the top: buyer’s counsel, buyer’s finance lead, lender, your own internal team. Mark each cell with what that audience can see. This takes twenty minutes. It prevents the worst kind of diligence incident, which is a document that reaches an audience it was never meant for.
Folders and permissions should line up. If you find yourself blocking a single file inside an otherwise open folder, your tree has a category problem. Move the file into a branch whose audience is already correct. One folder, one audience, no exceptions. Defense and government procurement offices have worked this way for decades, and the General Services Administration publishes records and access-control guidance that reflects how much trouble the mismatched version causes.
A retrieval test you can run this week
Before you invite anyone in, run this. Pick five people who didn’t build the tree: an accountant, an ops lead, a junior attorney, and two people from outside the company if you can get them.
- Write down the ten most likely buyer requests. Real ones. “All customer contracts over 100k signed since 2022.”
- Hand the list to each person and time them.
- Anything that takes more than two minutes, or takes two people to find, is a structural failure. Rewrite the folder, don’t just rename the file.
- Repeat with the same list. If the second pass isn’t faster, your problem is naming, not nesting.
Run it twice and you’ll catch most of the traps. The first pass usually fails on naming. The second usually fails on something filed in two places, which is worse than filing it wrong, because two copies means two answers.
What to do when the tree and reality disagree
Every seller eventually hits a document that genuinely belongs in two branches. A shareholder agreement is both corporate and legal. A lease is both commercial and financial.
Pick one home and leave a pointer in the other branch. A one-page summary that names the document and states its location keeps the buyer moving without creating a second copy to keep in sync. Nobody wants to maintain two versions of anything under a deadline.
And if you’re selling a business that’s been running for decades on paper, budget real time for the digitization pass. That work doesn’t compress just because the buyer is impatient.
Start with the index, not the upload
The urge is to start uploading. Resist it for one afternoon and sketch the index first, on a whiteboard or a napkin, mapping your four branches against the requests you expect. Everything after that is mechanical.
Deals don’t die from missing documents nearly as often as they die from documents that exist and can’t be located. Fix the tree, and you fix the clock. So here’s the question worth answering before your next buyer call: if someone asked for your March 2023 supplier agreement right now, how long would it take you to send it?
