Your company's know-how lives in one veteran's head. What happens when they leave?

The veteran glances at a drawing and knows whether to take the job and what quotation still makes money. The young engineers spend two days on it and still won't send it. That judgment was never written down, so when he leaves, the company starts over. This post breaks tacit knowledge into three layers, shows which one is worth recording, explains how to ask “why did you decide that?” while he's doing the work, and says when you shouldn't bother at all. No system pitch.

Johnny Yang12 min read
Your company's know-how lives in one veteran's head. What happens when they leave?

“He's retiring. What scares me isn't losing one person on the line. It's that nobody will dare send a quotation.”

That day I was standing on the floor of a metal fabrication shop. The owner pointed at a master craftsman by the window who was studying a drawing. He's in his sixties. A customer sends a drawing, and in about three minutes he can tell you which machine the part goes on, whether it needs a jig, where in the process the yield will drop, and what quotation the company needs to actually make money. Next to him, two young engineers spend two days on the same drawing. When they're done, the owner still won't send it. It has to go past the old man first. When the owner said that line, he didn't sound like he was talking about a staffing issue. He sounded like he was talking about a critical machine that was about to be discontinued.

I'm Johnny Yang, Brightstream Technology, Taiwan. I've led implementation teams for seven years, across 252+ projects in 32+ industries. “The company's lifeline lives in one person's head” shows up in most of the projects I've seen. The only difference is whether the owner has noticed. Here's the conclusion up front: For a small business, knowledge management isn't buying a system to store files. It's pulling out the part where a veteran employee makes the call, and turning it into something another person can follow and reach the same judgment about 70% of the time. In practice it takes three steps. First, find the few judgment calls that can actually get the company into trouble. Don't get greedy and try to record everything. Second, ask “why did you decide that?” while he's doing the work, and write it down on the spot. Not in a meeting afterward, from memory. Third, have the successor actually do the job with that document, while the veteran stands by, allowed to watch but not to touch. Wherever the successor gets it wrong is a gap in the document. Go back and fill it. The tools you already have are enough: a shared folder, an online doc, that notes field behind each item in your ERP that nobody uses. And if the judgment call only comes up two or three times a year, or the person isn't actually leaving, or you're leaking money faster somewhere else, don't do it. You won't get the time back.

Where the owner's instinct is right, and where it's wrong

Start on the owner's side. Worrying that know-how walks out the door with people is the right instinct, and it's usually more serious than the owner thinks. What actually gets lost isn't “how to do it.” It's “when not to do it that way.” A new hire can learn the procedure in three months. It takes three years to learn when to stop. During that gap, the company pays tuition with real orders and real material.

The instinct is wrong in two places. First, the assumption that you have to “keep all of it.” When owners think about this, they often picture a manual hundreds of pages long, get scared by the picture, and put it off for three years. In reality, 90% of what's in a veteran's head doesn't need recording, and if you record it, nobody reads it. Second, treating it as a “documentation problem.” So admin gets told to organize the folders, and the master gets told to “write something down when you have time.” I've seen both paths too many times, and they end the same way: the files get produced, they sit on the shared drive, and six months later nobody opens them, because what's written is steps, not reasons.

Three layers in a veteran's head, and only the middle one is worth money

Break tacit knowledge apart and it gets much clearer. I usually split it into three layers.

The outer layer is operating steps: how to start the machine, which button to press in the system, how to submit an order. This layer is the easiest to record. Film it. You don't even need to write anything. But it's also the least valuable, because it rarely leaves with the person anyway. A new hire figures it out in two weeks, and it's obsolete the moment the equipment changes.

The middle layer is exception handling: “when this happens, how do you judge it.” The tolerance on the customer's drawing is too tight. Do you send it back with questions or absorb it yourself? This batch of material is a slightly different color from last time. Run it as is, or make a sample first? The customer wants a rush job. Under what conditions does saying yes make money, and under what conditions is it free labor? This layer is where the veteran is actually expensive, and where the company bleeds after he leaves. This layer is worth the effort.

The innermost layer I call “appetite for risk”: trade-offs shaped by personal style. Some masters lean conservative and would rather earn less than make a mistake. Some take bets, and two wins a year carry the whole year's numbers. This layer is tied to the person's character. You can't record it, and you shouldn't. Force one person's risk appetite into company policy and you usually get a document nobody dares follow.

LayerWhat it looks likeWhat to do with it
Operating stepsThe order of startup, input, submitting ordersVideo or screenshots, do it once and you're done
Exception judgmentTrade-offs and reasons when something looks offThis is the layer to dig into, one page per judgment
Personal risk appetiteWhether to take it, whether to betDon't record it, let the next person grow their own

How to turn the middle layer into something you can hand over

There are only three actions, and the order can't change.

Find the judgment calls: trace back from where things went wrong

Don't call a meeting and ask the master to “share your experience.” Ask that way and you'll get a pile of correct but useless lines, like “be careful” and “communicate more with the customer.” He's not hiding anything. People genuinely can't articulate what they've already made automatic. Ask someone who's driven for thirty years how he takes a corner and he'll say “you just take it.”

The question that works is tracing back incidents. Pull the orders from the past year or two that actually went wrong: the ones that lost money, needed rework, drew complaints, blew the delivery date. Ask the same set of questions for each one. Where in the process did it go off? Who was it, at which step, making which decision? If the most junior person in the shop handled it today, would they make the same decision?

Once you've traced them, you'll find the truly fatal judgment calls are usually few. In the projects I've seen, they mostly cluster in a few places: whether to take the order, whether the quotation should include some invisible cost, whether to block incoming material, whether to accept a customer's last-minute spec change, and “is this batch ready to ship?” Whoever holds these few points holds the company.

Lock onto the two or three most expensive ones first. Trying to inventory the whole company's knowledge at once is like trying to digitize every process at once. It's the most common way these efforts die. I made the same point in “So many processes. Which one do you fix first?”.

Catch him while he's doing it, not afterward from memory

The best time to dig out the reasons behind a judgment is while he's doing it. An order comes in, he opens the drawing, and for those three minutes you stand next to him. Just don't film it. Video only captures his hands, not his head.

Three questions are enough. First: “What did you just see that made you decide this?” That digs for the signal, meaning what he's actually watching. The answer is often surprisingly specific. It might be a note next to one dimension on the drawing, a sentence in the customer's email, or the customer's name itself. Second: “Did you consider doing it another way? Why didn't you?” The path he rejected says more about the boundaries of his judgment than the path he chose. Third: “If a new hire did this today, where would they most likely go wrong?” This one almost always surfaces things no document would contain, because he's answering from an error-proofing angle.

When you write it down, fix the format to four parts: in what situation, what signal I saw, what I did, why I didn't do the alternative. One judgment per page. If it doesn't fill the page, you're doing it right. Don't make it a flowchart, don't make it a manual, don't add a cover page. This document has exactly one reader: the person taking over three months from now.

A common source of resistance on the floor is that the master doesn't want to talk. That's normal. Don't treat it as an attitude problem. What he's thinking is “if I tell you this, does the company still need me?” Only the owner can have that conversation, and it has to be direct: change his role from “the one who can do it” to “the one who teaches others to do it,” and make that second role a step up in the company. Tricking him won't work. He'll feed you correct but empty answers.

Writing it down is only half; the next person has to run it

A finished document is not a finished handover. The real acceptance test: the successor does the job with this document in hand, and the master stands by, allowed to watch, not allowed to step in. Wherever the successor gets stuck is a gap in the document. Fill it on the spot. Don't leave it for next time.

Run that two or three rounds and the content settles into something very usable. There's also a good side effect: the master starts adding things on his own, because watching someone follow what he wrote makes him itch to fix it.

If this person is retiring or leaving, formally schedule this “watch, don't touch” shadow period into the handover timeline. Not in the last week. The value of a handover isn't the final document. It's the overlap period when there's someone to ask. That's also the core of succession, which I covered in “Digitizing for succession: what's the first step?”. What the second generation lacks is never data. It's “why the previous generation decided that way.”

When this isn't worth doing

This section may matter more than everything above, because knowledge management is very easy to overdo.

The first case is when the frequency is too low. Some judgment calls come up two or three times a year. Turning them into something you can hand over costs far more time than just calling someone each time it happens. Don't record those. Keeping a phone number beats keeping a document. A retired master is more likely to take your call than a new hire is to read your document.

The second case is when the judgment is about to expire. The equipment is being replaced next year, the product line is winding down, a major customer's specs are being revised. Carefully recording the current logic now is like hardbinding something that's about to go out of date. Check the shelf life of the judgment first, then decide whether to start.

The third case is when the person isn't actually leaving, and you have other holes leaking more money. Owners often get spooked by “what if he leaves” and pour resources in, while quotations are still being miscalculated, receivables are still dragging, and the inventory numbers are still fake. Those bleed every month. A veteran leaving is only a “maybe.” Get the order backward and you'll do both badly.

The fourth case is awkward: the person with the know-how in their head is the owner, and the owner doesn't actually plan to let go. In that case, all the organizing is self-comfort. As long as the owner is around, everyone will keep coming back to ask him, and the documents will never actually get used. Either decide which piece to hand off first, or admit it's not time yet and save the effort.

Why you shouldn't buy a knowledge management system first

I know that at this point, many people's first thought is “so should I get a knowledge management system?” Not yet.

I've seen plenty of companies buy a tool, move their folders into it, and watch it die quietly within six months. The reason isn't mysterious. Nobody is responsible for maintaining it. Search never finds the page you want. Worst of all, it's separate from the workflow: an employee first has to remember “there's a document for this,” then switch to another system to look for it. If either of those two steps doesn't happen, the whole thing is useless.

What works better is the reverse: put the knowledge where the work happens. Write the quotation judgment into the field notes on the quotation template, so people can't miss it. Laminate the incoming-material checkpoints and stick them next to the inspection bench. Type an item's special warnings into that item's notes field in the ERP, so whoever places the order sees it the moment they pull up the part number. It's crude. Its key advantage is that “nobody has to remember to look.”

In other words, documents should live in the process, not in a folder. Once you've built up dozens of judgment cards that people actually use, too many to manage, then talk about a tool to hold them. It won't be too late. Do it in the reverse order and you've bought a very expensive empty cabinet.

The first thing you can do tomorrow

Take a sheet of paper and write one question: if a certain person didn't show up tomorrow, which three things in the company would stall immediately? Write down the names. Write down the things.

Pick the most expensive one. This week, when that thing happens again, stand next to him and ask the three questions: what did you see, what did you rule out, where would a new hire go wrong. Write one page on the spot, in the four-part format, and have him read it once to confirm. One page is enough.

Same time next month, you'll have three to five pages. Those pages are worth more than any system, because they grew out of the places where your company actually gets into trouble. When you find one day that they're not enough, and you need a better way to store and connect them, then talk about tools. If you're not sure whether this is what you should be doing right now, or whether to plug another hole first, we offer a free 30-minute diagnostic to help you sort out the order.

Found this useful? Pass it on
30 minutes · free

Bring the one thing you are stuck on — a free 30-minute diagnostic

You do not need to know what system you want. Tell Johnny Yang about the one thing that keeps blocking you, and in 30 minutes you will know whether to skip it, build it with coaching, or pay for a custom build. No pitch — if we cannot help, you will hear that straight.

Johnny Yang replies himself · limited spots each month · or just book a free 30-minute diagnostic

Johnny Yang (熊董)
Johnny Yang (熊董)
Founder, Brightstream Technology

Serial founder leading a small team that builds systems into an owner's everyday work.

As a founder
7 yrs
Projects delivered
252+
Industries served
32+
About Johnny