Tribal Knowledge Is a Single Point of Failure

There's a person on your team who "just knows how it works." They know how a refund actually gets approved, which client won't take a call before noon, the exact order you restart the system when it dies at 2am. Ask where any of that is written down and the honest answer is: it isn't. It lives in one head, and the day that head walks out the door, so does the knowledge.

The pattern hiding in plain sight

Every company runs on knowledge that was never written down. Some of it is small. The trick to getting the old printer to feed envelopes. Some of it is the whole business. How the founder prices a custom job, what a good lead actually sounds like on the phone, why the Tuesday deploy never breaks but the Friday one does.

We tend to read this as loyalty. As someone being indispensable, which sounds like a compliment. It is worth seeing it for what it really is. When one person is the only path to a piece of how the company works, that person is a single point of failure. Not because they're unreliable. Because they're human. They get sick. They take a holiday. They get a better offer. They burn out from being the only one who knows.

The tell is easy to spot once you look. Watch what happens when that person is out for a week. Decisions stall. Someone says "we'll have to wait until she's back." A customer waits longer than they should. The work doesn't stop entirely, but it limps, and everyone quietly agrees not to mention how fragile the whole thing felt. That fragility was always there. The absence just made it visible.

What a surgical checklist actually proved

I keep returning to a study most founders never hear about. A team led by the surgeon Atul Gawande tested a one-page checklist read aloud before operations. Nineteen plain checks. Confirm the patient's name. Confirm the procedure. Is the antibiotic given. Almost insulting in its simplicity for people who had trained for a decade.

The results, published in 2009 across eight hospitals on four continents, were hard to argue with. Major surgical complications fell by roughly a third. Deaths dropped too. These were experienced surgeons in real operating rooms, not students.

The surgeons did not get better overnight. Their skill was the same on the day after the checklist as the day before. What changed is where the knowledge lived. It moved out of memory, where a step can be skipped when the room gets tense and the clock is running, and onto paper, where it can't be quietly dropped.

That's the mechanism, and it's the whole point. Documentation is not bureaucracy. It's a way of doing the thinking once and then having that thinking work for you every time, under pressure, whether or not you're in the room and whether or not you remember. Memory fails exactly when the stakes are highest. Paper doesn't.

Why written knowledge compounds and head knowledge doesn't

Here is the part that matters for a business, and it's bigger than reliability.

Written knowledge can be handed off. It can be improved by the next person who reads it and notices a better way. It can be searched, copied, translated, and eventually given to a tool or a process to run without anyone watching. It gets better over time because each person who touches it can leave it sharper than they found it.

Knowledge stuck in one head can do none of that. It can only be re-explained. Again and again, forever, by the one person who holds it. Every new hire is a fresh download from the same overworked source. Every improvement dies when that person leaves, because nobody could see the thing well enough to build on it.

So the two kinds of knowledge move in opposite directions. The written kind compounds. It's an asset that grows. The unwritten kind decays. It's a liability that you pay interest on, in the form of bottlenecks and re-explaining, until the day it's gone entirely. Most companies have far more of the second kind than they think, and they mistake the cost of carrying it for just how business feels.

Why everyone skips this, and why it feels right to

Most founders skip documentation for a reason that's completely rational in the moment. Writing it down is slower than just doing the task.

And today, it is slower. If you already know how the refund gets approved, doing it takes ninety seconds. Writing the steps so someone else could do it takes fifteen minutes, and at the end you've produced nothing a customer can see. The task itself would have been done. The document feels like overhead.

The cost of skipping it is invisible right up until it isn't. It shows up later, all at once, in the worst form: the week that person is out, the resignation email, the new hire who takes three months to reach competence because there was nothing to read. By then the fifteen minutes you saved has cost you weeks.

This is why it almost never gets done voluntarily. The pain is deferred and the saving is immediate, and humans are wired to take that trade every time. The fix isn't discipline or willpower. It's lowering the cost of writing it down to the point where doing it is barely more effort than doing the task. Which is now genuinely possible in a way it wasn't five years ago.

How to start, concretely, this week

Don't try to document the whole company. That project dies on day two. Start where the failure would hurt most.

Make a short list of the things that would break if a specific person disappeared tomorrow. Not their whole job. The two or three things only they can do. The refund approval. The 2am restart sequence. The pricing logic. Then capture just those, one at a time, in the dumbest format that works.

The Gawande version is the model. One page. Plain steps in order. Written for someone competent but new, not for an expert who already knows. No theory, no polish. "First do X. If Y happens, do Z. Never do W." If a smart person who'd never done the task could follow it and get the right result, it's done.

The cheapest way to produce these now is to talk, not type. Have the person who holds the knowledge do the task once while narrating every decision out loud, and record it. The transcript is your rough draft. A tool can turn it into a clean checklist in minutes, which removes the exact friction that's stopped you before. The person who knows it the best spends the least time, and the knowledge finally leaves their head. That's the whole job. We built MARSA around this kind of work, and you can see more at marsa.ai/business, but the method costs nothing and you can run it yourself starting today.

The honest objection, and where it breaks

The strongest pushback is real, so let's take it seriously. Some knowledge genuinely can't be written down. Judgment. Taste. The feel for which deal is worth chasing. Try to reduce that to a checklist and you get a bad checklist that makes people worse, because they follow the steps instead of thinking.

That's true, and it's the line you have to hold. The goal is not to document everything. It's to document the repeatable so your people have their judgment free for the parts that actually need it. The checklist exists to handle the nineteen routine things that get skipped under pressure, precisely so the surgeon's attention is on the hard call, not on remembering the antibiotic.

There's a second worry worth naming. People resist writing down what they know because, somewhere, it feels like making themselves replaceable. That fear is understandable and it's backwards. The person who can hand off their routine work is the one who gets to do bigger work, gets promoted, gets to take a holiday without their phone going off. The one who hoards knowledge to stay needed stays exactly where they are, doing the same task forever, the human equivalent of a printer driver. Indispensable is not the same as valuable. It's often the opposite.