Learn / Process graveyard

The tribal knowledge audit: measure what walks out the door

Roughly 70 percent of how your operation actually runs is never written down. It lives in a handful of people, and when one of them leaves, recovering what they knew can take a year or two. This is a short, practical audit to quantify that exposure before a resignation forces the issue, and a note on why the way you capture the knowledge decides whether it survives at all.

Tribal knowledge is the part of the job that never made it into a document: the exception you handle a certain way because of something that went wrong in 2019, the customer whose orders always need a second check, the informal call you make before a handoff so the next team is ready. It is not written down because it is obvious to the person who holds it, and things that are obvious to an expert are invisible to them. That is precisely why it is dangerous. The most load-bearing knowledge in your operation is the knowledge nobody thought to record.

The numbers are stark. By common estimate around 70 percent of operational knowledge is undocumented, and when a key expert leaves, rebuilding what they held often takes a year or two, sometimes never fully. You do not feel any of this day to day, because the people who know are still there holding it together. You feel it all at once, on the day one of them hands in their notice, and by then measuring the exposure is too late to help.

The audit: quantify the exposure in an afternoon

You do not need a consultant or a project to size this. Take your list of critical processes, the ones that would cost you a customer, a shipment, or an audit finding if they ran wrong, and score each one on three axes.

  • Concentration. How many people could actually run this if the usual person was unavailable? If the answer is one, that is a single point of failure wearing a lanyard. If it is "one, plus someone who could probably work it out," that is still one.
  • Documentation coverage. What fraction of how this really runs is written anywhere accurate? Not "is there a document," but "would the document get a newcomer through the exceptions." Be honest: most documents cover the happy path and go quiet exactly where the hard-won knowledge lives.
  • Recovery time. If the person who holds this left tomorrow, how long before a competent replacement could run it at the same quality? Weeks is a manageable exposure. A year or two is an existential one.

Score each axis high, medium, or low, and multiply the risk in your head: a process that only one person can run, that is thinly documented, and that would take a year to recover is your number-one exposure. Rank the list. The processes that sit at the top are what would walk out the door with a single resignation, and they are the ones to capture first. This is a focused cousin of the wider process graveyard audit: that one measures how stale your documents are, this one measures how much of the truth was never written down in the first place.

The number that should worry you is not how many documents you have. It is how many of your critical processes score "one person, thinly documented, a year to recover." Every one of those is a countdown timer you cannot see.

Why "just write it down" fails

Once the audit names the exposure, the instinct is to send the expert away to document what they know. This is the most common approach and one of the least effective, for a specific reason: free recall misses exactly the knowledge you are trying to save. Ask an expert to write down what they do and they will produce the happy path, because the exceptions and judgment calls have become so automatic they no longer register as steps worth mentioning. The very thing that makes them valuable, the instinct for the edge case, is the thing they forget to write down. You get a clean document that is missing the 70 percent that mattered.

There is a second failure. Even when the knowledge is captured, if it lands in a static document it starts drifting immediately, and you are back to a page nobody trusts. Capturing tribal knowledge into a dead file is not preservation, it is delay. The knowledge that took a year to extract will be stale again inside two.

Capture by walking the real process

The method that works is not free recall, it is guided walkthrough. Sit with the person who runs the process and walk the actual work, step by step, as it really happens. At every step, probe the branches: what do you do when this is a dangerous goods shipment, when the payment fails, when the customer is on the watch list, when the sensor reads out of range? The exceptions that free recall would have skipped surface naturally when you are standing in the middle of the process asking "and what if." You are not asking the expert to remember their knowledge; you are drawing it out against the reality of the work.

And you capture it into a record that stays alive, not a document that dies. A living process with owners, versions, sign-off, and an audit trail keeps the captured knowledge true after the expert is gone, because it is maintained as the process changes rather than frozen on the day of the interview. The capture method and the storage method are the same decision: walk the real process, into a real record. Anything less and you have documented the easy part into a file that will be wrong again by the time you need it. This is also why where the process lives matters as much as who holds it.

What to do next

Run the audit this week. Rank your critical processes by concentration, coverage, and recovery time, and pick the single worst one. That is the knowledge most likely to walk out the door and hurt the most when it does. Then capture it properly: walk it with the person who holds it, probe every exception, and put it into a record that will stay true, not a document that will not. If you also want to stop the newly captured knowledge from going stale, keeping SOPs up to date without the quarterly review theater covers the structural side.

That is the job FLOW is built for. Bring the expert and one high-risk process to a pilot, walk it into a governed process system of record, and leave with the knowledge captured as a living process the next person can actually run, exceptions and all.

Common questions

What is tribal knowledge?

Tribal knowledge is the operational know-how that lives in people rather than in any record: the exceptions, the judgment calls, the workarounds, the informal handoffs, the reasons behind a step. It is how the work actually gets done, as opposed to how the official document says it gets done. By most estimates around 70 percent of operational knowledge is never written down at all, which means the majority of how your company runs exists only in memory.

How do you measure tribal knowledge risk?

Score exposure per critical process on three axes. Concentration: how few people could run it if someone was out. Documentation: what fraction of it is written anywhere accurate. Recovery time: how long it would take a competent newcomer to relearn it from scratch. A process that only one person can run, that is mostly undocumented, and that would take a year to recover is your highest-risk item. Rank your processes this way and the list of what to capture first writes itself.

How long does it take to recover lost operational knowledge?

When a key expert leaves, rebuilding the operational knowledge they held commonly takes a year or two, and sometimes it is never fully recovered. The replacement relearns exceptions the hard way, by hitting them, and the organization absorbs the errors in the meantime. That recovery window, not the salary of the role, is the real cost of concentrated tribal knowledge, and it is why the exposure is worth measuring before a resignation forces the issue.

What is the best way to capture tribal knowledge?

Do not just ask people to write down what they do. Free-recall documentation misses exactly the exceptions and judgment that make the knowledge valuable, because experts forget to mention the things that have become automatic to them. Instead, walk the real process with the person who runs it, capture it into a living record with owners and versions, and probe every branch and exception as you go. Capturing into a governed process record, rather than a document, is what keeps the knowledge true after the person is gone.

Capture the knowledge before the notice period does.

Bring one high-risk process and the person who holds it to a 30-minute pilot session. Walk it into FLOW and leave with it living: owned, versioned, exceptions and all.

Book a pilot →