Tribal knowledge

Tribal knowledge - is the undocumented knowledge held by one person or a small group inside an organisation about how work is really done: the tricks, shortcuts, unwritten rules, customer quirks and machine habits that never made it into a manual. It lives in people's heads and passes along by word of mouth, observation and practice. Most of it is a special case of what Michael Polanyi called tacit knowledge, summed up in his famous line that "we can know more than we can tell"^[1].

The word "tribal" is deliberate. The knowledge belongs to a tribe - a shift, a department, the old hands in the maintenance shop, the three developers who built the billing system - and outsiders can't get at it without being accepted into that group. Every organisation has it. A new employee discovers it on day two, when a colleague says "ignore the procedure, here's what we actually do". Sometimes that's a small mercy. Sometimes it's a warning sign that the official process no longer works, and nobody has told management.

Tribal knowledge and tacit knowledge

The two terms overlap but aren't identical. Tacit knowledge is knowledge that is hard or impossible to put into words: the feel of a correctly tightened bolt, the sound of a bearing about to fail, the instinct that tells an experienced buyer a supplier is bluffing. Polanyi built his argument on such examples. We recognise a face among thousands, he observed, yet usually can't say how we do it^[1].

Tribal knowledge is broader and more practical. Part of it is genuinely tacit. A good chunk, though, could be written down perfectly well - the password to the old server, the fact that one customer always orders double before Christmas, the setting that stops a printer from jamming - and simply hasn't been. It's explicable knowledge that stayed implicit through habit, lack of time or nobody asking. The distinction matters for managers, because the two halves need different treatment. The second half can be captured with a notebook and an afternoon. The first half can't; it has to be passed from person to person (see also explicit and implicit knowledge).

A third ingredient is social. Tribal knowledge includes knowing who - who can approve an exception, who in the supplier's office actually answers the phone, whose word carries weight on the shop floor. That kind of know-how is tied to relationships and doesn't survive a reorganisation well.

The SECI model

The most influential account of how such knowledge moves through a company comes from Ikujiro Nonaka. In a 1991 Harvard Business Review article he argued that Western managers saw the firm as an information-processing machine and treated only "hard", quantifiable data as useful knowledge, while successful Japanese companies such as Honda, Canon and Sharp relied on tapping the tacit, subjective insights of their employees^[2]. One of his examples involved the Matsushita bread-making machine: a software developer apprenticed herself to a hotel's head baker to learn, by watching and copying, how he kneaded dough, and the team then translated that skill into a product specification.

Nonaka and Hirotaka Takeuchi later developed the idea into the knowledge-creating spiral now known as the SECI model^[3]. It describes four modes of conversion:

Mode	From → to	Typical tribal-knowledge practice
Socialisation	tacit → tacit	Apprenticeship, working side by side, job shadowing
Externalisation	tacit → explicit	Interviews with experts, storytelling, metaphors, drafting a checklist with the old hand
Combination	explicit → explicit	Merging notes into manuals, databases, a knowledge management system
Internalisation	explicit → tacit	Learning by doing from the documented procedure until it becomes second nature

Seen through this lens, tribal knowledge is knowledge stuck in the first quadrant. It circulates by socialisation inside the tribe and never gets externalised. Managing it means pushing at least the most valuable part around the rest of the spiral.

Why tribal knowledge is a risk

Knowledge that lives in a few heads walks out of the door with them. The classic triggers:

Retirement. Pew Research Center noted that from 1 January 2011 the oldest American baby boomers began turning 65, and that about 10,000 more would cross that threshold every day for the next 19 years^[4]. Utilities, aerospace, defence and the public sector, where long careers were common, felt it early. David DeLong's Lost Knowledge describes the consequences in detail: repeated mistakes, slower decisions, lost customers and engineers rediscovering what their predecessors already knew^[5].
Turnover. A high employee turnover rate drains tribal knowledge continuously rather than in a single wave. Each departure takes a little; nobody notices until a process breaks.
Illness, holidays, sudden absence. The person isn't gone for good, but the plant still stops when they're away for three weeks.
Reorganisation and outsourcing. Moving a function elsewhere often transfers the documented tasks and loses the undocumented ones.

Software teams have a blunt name for the problem: the bus factor or truck factor, the minimum number of people who would have to be hit by a bus (or simply quit) before a project stalls. Avelino and colleagues estimated it for 133 popular GitHub projects and found that 65% had a truck factor of two or less^[6]. Open-source code is public, documented and reviewed. Imagine the figures for a twenty-year-old in-house spreadsheet that runs the pricing department.

In operational risk terms, a sole knower is a single point of failure, no different from a machine without a spare. Unlike a machine, it rarely shows up on the asset register.

Valuable or harmful?

Tribal knowledge isn't automatically bad. Plenty of it is the very thing that makes a team good at its job.

Where it adds value

It fills gaps that formal procedures can't anticipate. No manual covers every combination of material batch, humidity and machine wear.
It's fast. Asking the person at the next desk beats searching a portal.
It carries context - why a rule exists, when it may be bent - that dry documentation strips away.
It holds teams together. Shared stories and shortcuts are part of organizational culture.

Where it does harm

Workarounds hide broken processes. If everyone knows to re-enter an order twice because the system drops half of them, the system will never get fixed, and the cost stays invisible to management.
Gatekeeping. Some people hoard knowledge on purpose because it makes them indispensable. Others hoard it without meaning to, simply because they're too busy to explain. Either way, the organisation depends on goodwill it can't control.
Inconsistency. Three shifts with three private versions of "how it's really done" produce three quality levels.
Barriers to newcomers. Slow onboarding, tribes that close ranks and a silo mentality between departments often have tribal knowledge at their core.
Compliance exposure. An auditor or regulator wants the procedure on paper to match what happens on the floor.

How to capture tribal knowledge

Capturing all of it is neither possible nor worth the effort. The sensible approach is to find the knowledge that would hurt most to lose, then pick a method that matches its nature: explicit-but-undocumented knowledge can be written down, while deeply tacit skills have to be transferred person to person.

Person-to-person methods

Mentoring and apprenticeship. Pairing an expert with a successor for months, not days. Dorothy Leonard and Walter Swap, who call experience-based expertise "deep smarts", argue that it can't be passed on through slides, best-practice websites or lectures; the expert has to coach the novice individually through guided experience^[7].
Job shadowing. The successor watches the expert at work and asks questions on the spot. Remote variants - virtual shadowing - use video calls and screen sharing.
Cross-training and job rotation. Deliberately increasing the number of people who can perform a critical task, which attacks the bus factor directly.
Communities of practice. Etienne Wenger described them as groups held together by mutual engagement, a joint enterprise and a shared repertoire of routines, stories and tools built up over time^[8]. A community of maintenance technicians across several plants, meeting monthly, spreads tribal knowledge beyond a single tribe without forcing it into a document first.

Codification methods

Standard work and work instructions. In lean environments the current best method is written down as the standard; improvements found by operators become the new standard through kaizen.
Video work instructions. Filming an experienced operator doing a changeover often captures movements and cues that text misses. Cheap, quick and surprisingly effective for manual skills.
Knowledge bases and wikis. Searchable repositories, ideally edited by the people doing the work, not by a separate documentation team.
Expert interviews and storytelling. Structured interviews before a retirement, asking "what do you do when..." questions and recording near-misses.
After-action reviews. Short structured debriefs after a project or incident: what was supposed to happen, what actually happened, why, and what we'll do differently. They turn fresh tribal experience into shared lessons while memories are still sharp.

None of these works if people don't want to share. Incentives, time allowances and a culture that rewards sharing of information rather than hoarding matter as much as any tool. That's also why many firms treat the topic as part of a wider learning organization effort, not an IT project.

Example of a knowledge-risk assessment

The following case is hypothetical. A mid-sized foundry wants to know which undocumented knowledge to tackle first. It lists six critical processes and scores each on three factors:

Criticality (C): impact on safety, customers or revenue if the process fails, 1 (low) to 5 (severe).
Knowledge concentration (K): based on how many people can do the task without help - 1 knower = 5, 2 knowers = 4, 3 = 3, 4-5 = 2, 6 or more = 1.
Retirement horizon (R): time until the nearest expected departure among those knowers - under 1 year = 5, 1 to under 2 years = 4, 2 to under 3 years = 3, 3-5 years = 2, more than 5 years = 1.

The risk score is the product:

\[ Risk = C \times K \times R \]

so it ranges from 1 to 125. The firm sets three bands: 60 or more = act now, 30-59 = plan transfer this year, below 30 = monitor.

Process	C	Knowers	Years to departure	K	R	Score	Band
Furnace restart after shutdown	5	1	1.5	5	4	100	Act now
Casting surface inspection	5	4	0.8	2	5	50	Plan
Tooling set-up for largest customer	4	2	2.5	4	3	48	Plan
Alloy supplier negotiation	3	1	4	5	2	30	Plan
Legacy ERP month-end workaround	4	1	6	5	1	20	Monitor
Packaging line changeover	2	6	1	1	4	8	Monitor

What does the table say? The furnace restart is the clear priority: 5 × 5 × 4 = 100. One person knows it, and he leaves in eighteen months. The response is a combination of methods - film the next restart, pair him with two successors for every restart until he leaves, and turn the recordings into a work instruction.

Casting inspection scores 50 even though four people can do it, because one of them retires within a year and the task is safety-related. The ERP workaround is interesting. Only one person knows it, yet it scores just 20 because she isn't going anywhere soon. Low score, but not zero risk: illness or a resignation would still hurt. And frankly, a workaround known by one person is a hint that the ERP itself needs fixing. The packaging changeover, known by six people, barely registers at 8.

The method is crude, like any risk matrix. Multiplying ordinal scores isn't mathematically elegant. But it forces an honest conversation about who knows what, which is usually the first time anyone has asked.

Advantages of managing tribal knowledge

Lower operational risk: fewer processes that stop when one person leaves.
Faster onboarding and shorter learning curves for new staff.
More consistent quality across shifts, teams and sites.
Hidden workarounds come to light, which exposes process problems worth fixing.
Better preparation for retirements, succession and reorganisations.
Expertise gets recognised. Experts who are asked to teach often feel valued rather than replaced.
Knowledge becomes reusable across the organisation, supporting transfer of knowledge between units and organizational learning.

Limitations

Much tacit knowledge resists codification. A video of an expert doesn't make the viewer an expert, as Polanyi's argument implies^[1].
Capture takes time from the very people who are busiest. Experts close to retirement may have little motivation to help.
Documents go stale. A knowledge base that nobody maintains becomes a new source of error.
Knowledge is power. Employees may fear that sharing makes them replaceable, especially during cost-cutting.
Over-codification can kill useful flexibility and the informal exchange of information that keeps teams sharp.
Risk scoring relies on subjective judgement; the example above is only as good as the honesty of the people filling it in.
Results are hard to measure. The benefit of knowledge that wasn't lost rarely shows up in any report.

Related approaches

Tribal knowledge sits inside the wider field of knowledge management and touches human capital management, succession planning and business continuity. In lean management it is handled through standard work and continuous improvement; in software development, through code review, pair programming and attention to the bus factor. The SECI model, communities of practice and Leonard and Swap's work on deep smarts give the theoretical background, and organizational knowledge is the end goal: knowledge that belongs to the company rather than to whoever happens to remember it.

References

Avelino G., Passos L., Hora A., Valente M.T. (2016), A Novel Approach for Estimating Truck Factors, 2016 IEEE 24th International Conference on Program Comprehension (ICPC), pp. 1-10 (preprint)
Cohn D., Taylor P. (2010), Baby Boomers Approach 65 - Glumly, Pew Research Center, Washington
DeLong D.W. (2004), Lost Knowledge: Confronting the Threat of an Aging Workforce, Oxford University Press, New York
Leonard D., Swap W. (2004), Deep Smarts, Harvard Business Review, September 2004
Nonaka I. (1991), The Knowledge-Creating Company, Harvard Business Review, November-December 1991 (republished July-August 2007)
Nonaka I., Takeuchi H. (1995), The Knowledge-Creating Company: How Japanese Companies Create the Dynamics of Innovation, Oxford University Press, New York
Polanyi M. (1966), The Tacit Dimension, Doubleday, Garden City (reissued 2009 by University of Chicago Press, Chicago)
Wenger E. (1998), Communities of Practice: Learning, Meaning, and Identity, Cambridge University Press, Cambridge

Footnotes

↑ Polanyi M., The Tacit Dimension, Doubleday, 1966
↑ Nonaka I., The Knowledge-Creating Company, Harvard Business Review, 1991
↑ Nonaka I., Takeuchi H., The Knowledge-Creating Company, Oxford University Press, 1995
↑ Cohn D., Taylor P., Baby Boomers Approach 65 - Glumly, Pew Research Center, 2010
↑ DeLong D.W., Lost Knowledge: Confronting the Threat of an Aging Workforce, Oxford University Press, 2004
↑ Avelino G., Passos L., Hora A., Valente M.T., A Novel Approach for Estimating Truck Factors, ICPC 2016
↑ Leonard D., Swap W., Deep Smarts, Harvard Business Review, 2004
↑ Wenger E., Communities of Practice: Learning, Meaning, and Identity, Cambridge University Press, 1998

Tribal knowledge — recommended articles

Tacit and explicit knowledge — Communities of practice — Knowledge management — Transfer of knowledge — Mentoring — Job shadowing — Organizational learning