Can Your Business Learn From Its Own History? Start With the Files

You know the feeling. A lead comes in. It looks promising. You start digging, and you think, “We did a project like this a couple of years ago. What did we charge them?”

So you hunt. Dropbox for the signed contract. Google Drive, Notion, or OneDrive for the estimate that went with it. You know the files are there. You just have to find them, for crying out loud. Is v3_revised or final_final the final version? And where’s the fully executed copy?

Twenty minutes later, you’ve guessed or asked somebody who might know. Neither is a record.

Whether a business can pull a real number out of its own history depends on a few habits you either established years ago or didn’t: a code on every file, a hard and fast rule about where won and lost work goes, and signed agreements in a format you can search. Memory won’t cover the gap. It drifts, and not always in the direction you’d expect.

Where I learned this

I ran my former agency for twelve years on hundreds of jobs. I’m an operations and organization lunatic (I’m not even sure I knew it back then), and most of this I learned the hard way: observing which habits stuck when somebody needed a number and which ones fell apart come the first busy month.

I did all of this with file names and folders, because that’s what we had. Today a lot of it lives in a customer relationship management (CRM) tool, a project management platform, or your accounting system, and artificial intelligence can search and read files in ways nothing could back then. The principle doesn’t change. In a CRM tool, the code becomes a job number field, and the won-or-lost folder becomes a deal stage. And search tools, AI included, work far better when every record of a job shares one consistent identifier to anchor on. Whatever you use, the habits are the same: one identifier on everything tied to a job, a clear won or lost status, and signed agreements you can read as text.

A code on every file

The habit that does the most heavy lifting is a project code at the front of every file name. At ours, it had three components: which office, which client, which job. Here’s an example. PHL-ACM-004 would be the Philadelphia office, a client (call it Acme Widgets) given the three-letter code ACM like an airport, and that client’s fourth job.

Every estimate, proposal, agreement, and invoice for that job carries the code. That’s the whole thing. Folders get moved around, files get renamed to fit whatever person’s taste, and tools come and go, but the code connects a proposal to the agreement it became and the invoices that paid for it.

A project code is the cheapest database you’ll ever build. It costs a few characters at the front of a file name, and years later it may be the only thing connecting a proposal to the invoice that paid for it.

A firm rule for won and lost work

The second habit is crude. Every won job goes in one folder. Every lost job (the ones we pursued and didn’t win) goes in another. It sounds dumb, but it means the folder structure itself answers “did we get this one?” without asking anybody to remember.

It only works if somebody checks it. If a won job gets filed with the lost jobs, it doesn’t announce itself. It just stops counting in every tally of what you sold.

Signed agreements you can search

The third habit is the one most firms skip. A scan of a signed page is a picture, not text. Search can’t read it, and neither can anyone skimming file names a year later, especially if it was saved as “scan_0042.pdf.” The fix is boring: keep the final unsigned version right beside the signed scan so the text survives, and put the word “signed” in the scan’s file name. The Oregon State Archives’ file-naming guidance recommends exactly that label.

How these habits break

None of this fails all at once. It erodes. Someone starts naming files “Client | Proposal | date” because it reads fine to a person, and the code quietly drops off. Every file still looks organized. None of them can be grouped by job.

Someone gets clever with the format, adding a digit or a fourth part for one special client, and that client stops matching the pattern everything else follows. Consistent beats clever. The same Oregon guidance makes the point: uniform names and folder structures are what make files retrievable. Its advice on dates ends with three words that cover everything else too: “Keep it consistent.”

And someone files a won job in the lost pile during a busy week. Nobody notices, because nobody looks.

Your memory is not a record

The other reason to keep good records is less comfortable. The stories owners tell about their own businesses drift.

The best kickoff I ever had was with a craft brewery. When our van pulled into their driveway, the client was standing outside clapping. By 9:30 that morning we were drinking beer in a barn loft, talking about their website. For years, though, I told that project as a cautionary tale: the founder swooped in late, tore up our design direction, and the relationship never recovered. Then, a few years after launch, their marketing director, the same person I’d had the hardest exchange with, reached out on LinkedIn. She’d moved to another company, was hiring a firm for a website project, and wanted to know who I’d trust. That didn’t square with my version. When I thought it through, the founder’s pushback had come at our first designs, not the end. We worked through it, the site launched, and they came back asking for more. My memory had kept the worst few weeks and quietly dropped everything after them.

None of this makes anyone unusual. Harvard psychologist Daniel Schacter calls it bias, one of his “seven sins of memory”: what you know and feel now reshapes how you remember what happened then. In one study he describes, people whose feelings about a partner had cooled remembered their earlier opinions as more negative than they had reported at the time. Client relationships work the same way. When a job ends badly, the whole thing gets rewritten as bad, and the parts that went fine get lost in your own noise.

That matters for more than nostalgia. If your pricing instincts, your “we tried that once” rules and your sense of which clients are trouble all come from stories, you’re running the business on folklore. Some of it is right, but you often don’t know which part.

What to set up now

You don’t need to rebuild years of history. You need the next ten years to be searchable. Here’s what I’d do, in order.

1. Put a code on every job, and put it first in every file name, or in the job number field if your tools have one. Pick a format, write it down where everyone can see it, and keep it boring: same length, same parts, same order. No exceptions for “quick” proposals. Those are the ones that go missing.
2. Make one firm rule for where won and lost work lives, whether that’s a folder or a status in your tools. Moving a folder takes ten seconds at the end of a job and an afternoon of detective work years later.
3. Keep the text of every signed agreement. Save the final unsigned version beside the signed scan, label the scan “signed,” and record the contract total somewhere you can read without opening a picture. Your bookkeeping or project tool is fine.
4. Use one date format in file names, year first (2026-10-20), so files sort in the order they happened.

Some honest limits. If you run a handful of small, one-off jobs a year and can hold them all in your head, this is overkill. And don’t spend a month re-filing old work. Start with whatever’s active, then go back only as far as your old prices are still useful to you. The cost of skipping it is real, though. Atlassian’s State of Teams 2025 survey of 12,000 knowledge workers and 200 executives found teams lose about a quarter of their time just searching for answers. That’s already on your payroll.

Where the payoff shows up

The payoff comes the next time someone asks a real question. What did we charge for a job like this, and did we make money on it? Has this client done this to us before? How long did the last three of these take? With the records set up, those are lookups. Without them, they’re arguments between people who remember it differently.

There’s a longer payoff, too. If you ever sell the business, a buyer is paying for something that can run without you. Records that show what you sold, what you charged, and how the work went are part of what makes a firm transferable, and they’re among the first things a serious buyer will ask to see.

It also changes what you can hand to someone new. A new project lead, an operations hire, or an AI agent working your sales pipeline can learn from real precedent instead of the owner’s war stories. That’s the work I do in documentation and process handoff: getting how the business actually runs out of people’s heads and into a form someone else can use.

Your files already know more than you do. Make sure they can tell you.