How to prevent a crisis of knowledge loss in your industrial plant

In 2006, researchers Salvatore Parise, Rob Cross and Thomas Davenport published a study in MIT Sloan Management Review that remains the main reference on this topic: "Strategies for Preventing a Knowledge-Loss Crisis". Their conclusion was resounding: when a key employee leaves an organization, not only is their technical knowledge lost — their network of relationships, their ability to connect people and problems, and their role as a bridge between departments is also lost.

The example that opened the article was devastating: when Delta Air Lines lost its veteran mechanics in the 90s, compensation costs fell in the short term. But the less experienced employees who remained took much longer to diagnose and repair airplanes. The result was delays, cancellations, dissatisfied customers, and increased cost per seat mile.

That was on an airline. In your workshop or factory, the effect is the same but more concentrated: when the operator who has been working with the folder for 20 years retires, there is no team of 50 people to absorb his knowledge. There are one or two substitutes who have never touched that machine alone.

The three roles you can't lose

Parise, Cross and Davenport's research, based on analysis of organizational networks in more than 80 companies, identified three types of employees whose departure generates the most serious crises:

The central connectors are the people everyone turns to when they have a problem. In an industrial plant, he is the operator that everyone asks when the machine does something strange. It's not necessarily the boss — it's the person who knows the most and shares the most. When it leaves, an information vacuum is created: people no longer know who to ask.

The brokers are people who connect groups that would not otherwise communicate. In an industrial SME, it may be the shift leader who knows both production and maintenance people, or the senior operator who acts as a bridge between the workshop and the technical office. If a few key brokers leave, the company's social network fragments into isolated groups.

Peripherals with niche knowledge are people who are not in the center of the network but have very specific knowledge that is difficult to replace. The technician who knows how to calibrate the old machine that no one else touches. The operator who knows the trick so that the line does not get clogged with certain material. They are easy to ignore on superficial analysis, but their loss can paralyze entire operations.

The specific problem of Spanish industrial SMEs

The original research focuses on large corporations. But the problem is even more acute in industrial SMEs for several reasons:

The concentration of knowledge is extreme. In a company with 30 employees, it is common for 2-3 people to accumulate 80% of the critical operational knowledge. There is no natural redundancy — if one of those people is missing, there is no one to replace them.

Knowledge is predominantly tacit. Industrial workers learn by doing, not by reading. Their knowledge is in their hands, in their ears (the noise the machine makes when something goes wrong), in the intuition accumulated over years. That tacit knowledge is the most difficult to capture and the one that has the most value.

There is no tradition of documentation. Unlike sectors such as pharmaceuticals or aeronautics, where process documentation is mandatory, in the general manufacturing industry documentation is scarce or non-existent. Only 23% of Spanish industrial SMEs have documented and updated standard operating procedures.

Demography puts pressure. The Spanish industrial sector has an inverted age pyramid: many veteran workers close to retirement and few young people entering. Each retirement that is not managed is knowledge that is much more difficult to recover.

What the research says works (and what doesn't)

The MIT Sloan study and subsequent research have distilled which strategies work and which are a waste of time:

What does NOT work

Passive shadowing — putting the new guy next to the veteran for a few weeks — captures only explicit knowledge (what is done). It does not capture tacit knowledge (why it is done, what signals to look for, what alternatives to consider). The new one sees the movements but doesn't understand the logic behind them.

Timely documentation — sitting down one day to “write everything you know” — generates long, incomplete and immediately obsolete documents. The expert cannot verbalize 80% of what he knows because it is unconscious to him. And the document it produces becomes outdated within weeks.

Massive classroom training — taking production workers for in-person training — is expensive, consumes productive hours, and has a very low retention rate. After 48 hours, the operator has forgotten 70% of what they were taught.

What DOES work

Capture integrated into daily work. Knowledge is best captured when the expert is not asked to “document.” It is captured when the expert uses tools that are useful to him and that, as a side effect, record his knowledge. If the operator follows a step-by-step process on his mobile phone with video, he is consuming documentation. If you report a problem and how you resolved it, you are creating new documentation. No one has “documented” — they have simply worked.

Distribution at the point of use. Documented knowledge is only useful if it reaches the operator who needs it, at the time and place where they need it. Immediate desktop access to procedures is infinitely more useful than a PDF in a server folder.

Continuous validation by experts. Procedures cannot be static. They need a mechanism for operators to report problems, propose improvements, and receive supervisor validation. Thus knowledge improves with each turn instead of degrading.

Analysis of knowledge networks. Knowing who knows what — and who depends on whom — allows you to identify the points of greatest risk before the loss occurs. A skill matrix that shows which processes each operator masters immediately reveals where there is dependence on a single person.

From theory to practice: the 5 phases of knowledge retention

Phase 1: Map the risk

Identify which people in your plant accumulate critical knowledge and when there is a risk of losing it: planned retirements, temporary contracts, profiles with high turnover, or simply people who are "the only one who knows" about something.

A practical tool is the process risk matrix: for each critical process, document who controls it, if there is a backup, and when it could go away. Processes where only one person knows how to do it are your greatest vulnerability.

Phase 2: Capture without friction

The most efficient format for capturing operational knowledge is the short video — 30 seconds to 2 minutes, recorded with the cell phone at the workplace. It is faster than writing, richer than text, and easier to consume by the substitute.

Supplemented with step-by-step instructions and troubleshooting (what to do when something goes wrong), an expert operator can document an entire process in 15 minutes. And if the system allows problems and solutions to be captured during daily work, the documentation enriches itself over time.

Phase 3: Distribute where it is needed

Documented knowledge must be accessible in the workplace, without barriers. No apps to install, no complex users, no searching in folders. A simple way to open the right process from each position completely changes the actual adoption.

The substitute operator scans, enters a simple code, and in 3 seconds has the steps of the process with video. This eliminates the dependency on the expert being available to explain.

Phase 4: Validate and update

Static knowledge expires. You need automatic signals that a process needs review: if the same step repeatedly generates problems, if an operator solves something differently than it is documented, if a long time has passed without review.

The ideal validation chain is: the operator detects a problem or improvement → the supervisor reviews it → if it is valid, the company manager incorporates it into the documented process. Thus, each turn improves the knowledge base instead of wearing it down.

Phase 5: Measure the health of knowledge

What is not measured is not managed. A knowledge health index that combines coverage (what percentage of critical processes is documented), versatility (how many operators master each process), freshness (how long it has not been updated), and autonomous resolution (what percentage of problems the operator solves without help) allows proactive decisions to be made.

If the index shows that a critical process is only mastered by one person and has not been reviewed for 6 months, you have a red flag before the person leaves — not after.

Sources

• Parise, S., Cross, R. and Davenport, T.H. (2006). "Strategies for Preventing a Knowledge-Loss Crisis." MIT Sloan Management Review, Vol. 47, No. 4, pp. 31–38.
• DeLong, D.W. (2004). Lost Knowledge: Confronting the Threat of an Aging Workforce. Oxford University Press.
• Korra AI (2025). "The Economic Impact of Knowledge Loss Due to an Aging Workforce in Industrial Companies."
• National Association of Manufacturers (2024). Industrial employment projections 2024–2030.

This article is part of the series of practical guides on industrial operational documentation by GMV Solutions. If you want to know how much the lack of documented procedures in your plant is costing you, take our operational risk diagnosis — result in 2 minutes.