To capture tribal knowledge, find the skills that depend on one person, get what they know into a form others can use, and turn it into training people are assessed on. Most programmes manage the first two steps and stall on the third: recordings and interview notes pile up because nobody has time to build them into courses. The fix is to run capture and training production as one workflow, with a qualified person approving every source before anything gets built.
What is tribal knowledge?
Tribal knowledge is the know-how that lives in experienced employees' heads instead of in your SOPs. It covers the shortcuts, machine quirks and judgement calls that keep a line running. You'll also hear it called institutional knowledge or tacit knowledge. Every site has it, and most sites only notice it when someone leaves.
Tribal knowledge shows up as specific, valuable habits:
- The setter who runs a changeover in a different order from the SOP because it cuts scrap on the first parts.
- The maintenance technician who knows which pump needs a lower torque than the manual says.
- The quality lead who spots a bad batch by its colour before the lab result comes back.
Is tribal knowledge bad?
Tribal knowledge isn't bad. It's often the best method your operation has. The risk is concentration: when one person holds a critical skill, their absence stops the work. Every other shift learns a slightly different version by word of mouth. A holiday becomes an operational risk, and a resignation becomes a crisis.
The cost of losing that knowledge is measurable. Gallup (2024) puts the cost of replacing a technical employee at around 80% of their salary, and a frontline employee at around 40%. Those figures cover recruiting and lost productivity. They don't cover the decades of unwritten judgement that leave with a retiring expert.
The pressure is building. A large share of the most experienced people on manufacturing floors are nearing retirement, and every departure takes years of unwritten judgement with it.
Why does tribal knowledge capture stall after the interview?
Tribal knowledge capture stalls because a recording isn't training. An hour of at-the-machine video still has to become steps, checks, scenarios and an assessment before a new hire can learn from it. Someone has to build that. In most plants, that someone is an L&D team already behind on its backlog.
The build is where the real teaching happens. Josh Bersin, global industry analyst and founder of The Josh Bersin Company, puts it this way: "Teaching somebody how to do something doesn't mean just teaching them how to do it. You've got to teach them about it, so they know why they're doing what they're doing." A video shows the how. Training has to carry the why.
The usual fixes each solve half the problem. A skills matrix shows who holds a skill. It doesn't build the training to spread it. Paper-first capture produces a document. It doesn't produce a course. One-to-one cross-training reaches two successors. It doesn't reach 40 operators across three shifts.
The gap shows up on the floor. Only 24% of frontline workers believe they receive the training they need to succeed (Axonify, 2026). Captured knowledge that never becomes training doesn't change that number.
Which tribal knowledge should you capture first?
Capture the skills where one departure would hurt most and soonest. Rank each critical skill on two things: how many people and lines depend on it, and how few people can do it to standard. A skill held by one person who retires within a year goes to the top. A skill five people share can wait.
Most plants do this ranking by hand, from spreadsheets and supervisor memory. The data already exists, but it sits in separate systems. The LMS holds completions, the competency tool holds ratings, HR holds roles and quality software holds the errors. No single view shows where the risk sits.
Edplay is building that view into Workforce Intelligence. It will pull those sources into one picture, rank gaps by risk and by how many people they affect, then propose what to build first. The aim is a plan that helps you find and close skill gaps before a departure date forces the issue.
Is it a skill gap or a knowledge gap?
A skill gap means people can't do the task yet. A knowledge gap means nobody has approved material to teach it from. The two need different fixes. A skill gap with a good SOP behind it needs training. A knowledge gap needs capture first, because there's nothing reliable to build from.
Treating every gap as a training problem wastes effort. Designers end up building courses from an out-of-date SOP, or from nothing at all. Three shifts running one task three ways is often a knowledge gap in disguise. The standard was never settled, so no course can teach it.
Edplay is building this distinction into Workforce Intelligence. It breaks SOPs down into the skills they actually teach, then compares those skills against your competency records. Each gap will show whether approved evidence exists, needs review or is missing. That tells you whether to build, review or capture.
How to capture tribal knowledge and turn it into training
To capture tribal knowledge and turn it into training, run five steps as one workflow: request, capture, review, build and assess. The first three get the knowledge out safely. The last two make it teachable.
1. Request knowledge from a named expert
Ask a specific person about a specific skill. "Write down what you know" produces nothing useful. "Show us how you remove the air filter on the routine service" produces evidence. A named request gives the expert a small, clear task instead of an open-ended documentation project.
2. Capture it the way the task works
Match the format to the task. Experts explain some tasks best out loud. Others are easier to show than describe, especially hands-busy work at a machine. The capture process Edplay is building accepts voice descriptions, uploaded SOPs and documents, and first-person video recorded through smart glasses. The learner later sees exactly what the expert saw.
3. Review and approve before anything gets built
Nothing should become company knowledge until a qualified person verifies it. An expert's shortcut might be the best method on the floor. It might also break a safety rule. A review gate catches the difference and keeps one approved version, instead of three shift-by-shift variations.
4. Build training from the approved source
Build directly from what the reviewer approved. Edplay already builds courses from SOPs, documents and SME conversations: a structured first draft in four to five minutes, and roughly 80% of a course from a verbal conversation with an SME, with no documents required (see how to turn SOPs into interactive training). Edplay only references the content you give it and doesn't train any language model on it. Your designers review the structure, objectives and assessments before anything rolls out.
5. Assess against the expert's standard
Test learners against the standard the expert set, not against completion. Edplay courses include knowledge checks and assessments. Scenarios can rebuild a real incident, so learners practise the judgement as well as the steps.
Edplay co-founder Thomas Bellamy shows how Edplay is building this loop in a Workforce Intelligence walkthrough. A routine equipment service flags a gap in one skill: air filter removal. Edplay requests evidence from an operator, who records the task through smart glasses. A reviewer approves the recording, and Edplay builds a course around it, covering the steps and what to look out for. Anyone else who needs the skill trains from the expert's point of view.
Data Breakdown:- Dimension: Where capture starts | Old Way (Articulate): Blank SOP template or sit-down interview | Bolt-On Way (ChatGPT): Expert's notes pasted into a chat window | Edplay Way: Named request for a specific skill (in development)
- Dimension: Capture formats | Old Way (Articulate): Written notes, occasional video | Bolt-On Way (ChatGPT): Typed text | Edplay Way: Voice, documents and first-person smart-glasses video (in development)
- Dimension: Verification | Old Way (Articulate): Depends on reviewer availability | Bolt-On Way (ChatGPT): None; internal SOPs sit in a public AI account | Edplay Way: Builds only from content you provide; approval gate before build (in development)
- Dimension: Time to first draft | Old Way (Articulate): Days of storyboarding and build | Bolt-On Way (ChatGPT): Minutes, but unstructured text, not a course | Edplay Way: Four to five minutes to a structured course draft
- Dimension: Proof of transfer | Old Way (Articulate): Completion records | Bolt-On Way (ChatGPT): None inside your LMS | Edplay Way: Knowledge checks and assessments built into every course
How do you know the knowledge actually transferred?
Knowledge has transferred when someone other than the expert passes an assessment against the expert's standard. Completion records don't prove it. A finished course shows someone clicked through. It doesn't show they can remove that air filter correctly on a night shift without help.
Track two numbers for every critical skill: whether approved knowledge sits behind it, and how many people have shown the skill in an assessment. A skill with approved evidence and three assessed operators is safe. A skill with one expert and no source material is your next capture request.
Edplay is building this feedback loop into Workforce Intelligence. Pre- and post-assessment results will flow back as skill proficiency, and you'll be able to send them to your LMS. You see what changed after the training, not only who completed it.
Key Takeaways
- Tribal knowledge is the undocumented know-how in experienced employees' heads, and most of it is tacit judgement a blank SOP template won't capture.
- Tribal knowledge isn't bad in itself; the risk comes from concentration, when one person's absence stops the work.
- Capture stalls because recordings aren't training, and someone still has to build steps, scenarios and assessments from them.
- Capture the skills first where one departure would hurt most, ranked by how many people depend on them and how few can do them.
- A skill gap needs training, while a knowledge gap needs capture first, because there's no approved material to teach from.
- To capture tribal knowledge and turn it into training, run request, capture, review, build and assess as one workflow.
- Knowledge has transferred when someone other than the expert passes an assessment against the expert's standard.
Frequently asked questions
What is tribal knowledge?
Tribal knowledge is the know-how that lives in experienced employees' heads instead of in your SOPs. It covers the shortcuts, machine quirks and judgement calls that keep a line running.
Is tribal knowledge bad?
Tribal knowledge isn't bad. It's often the best method your operation has. The risk is concentration: when one person holds a critical skill, their absence stops the work.
Why does tribal knowledge capture stall after the interview?
Tribal knowledge capture stalls because a recording isn't training. An hour of at-the-machine video still has to become steps, checks, scenarios and an assessment before a new hire can learn from it.
Which tribal knowledge should you capture first?
Capture the skills where one departure would hurt most and soonest. Rank each critical skill on how many people and lines depend on it, and how few people can do it to standard.
Is it a skill gap or a knowledge gap?
A skill gap means people can't do the task yet. A knowledge gap means nobody has approved material to teach it from. A skill gap needs training; a knowledge gap needs capture first.
How do you capture tribal knowledge and turn it into training?
To capture tribal knowledge and turn it into training, run five steps as one workflow: request, capture, review, build and assess.
How do you know the knowledge actually transferred?
Knowledge has transferred when someone other than the expert passes an assessment against the expert's standard. Completion records don't prove it.