Three months before the veteran engineer retires, the knowledge-management project begins. A junior colleague is assigned to interview her. Templates are filled. A wiki grows. Everyone involved can feel the futility, and everyone continues, because the ritual lets the organization say it captured her knowledge. A year later, the wiki is a mausoleum: accurate, complete, and unread, while her former team calls her mobile about the things that were never in it.
The project failed before it started, because it misidentified what needed to move.
Facts move easily. Judgment doesn't.
What the departing expert knows divides roughly into three layers. Facts: the vendor's real lead time, where the bodies are buried in the codebase. These document well, and mostly they are already documented or quickly rediscoverable. Patterns: what this class of failure smells like at week two rather than month six. These document poorly; written down, they read as platitudes until you have seen the failure once. And judgment: when to break the pattern, which rule is currently wrong. This does not document at all. It is exercised, not stated, and the expert often cannot articulate it, which is why the interviews yield modest returns.
The formal transfer mechanisms organizations buy reflect this gradient. The most comprehensive evidence on mentoring, Eby and colleagues' 2008 meta-analysis of 112 studies across youth, academic and workplace settings, found consistently favorable outcomes but generally small effects (workplace effect sizes ranged roughly from .03 to .19), and the authors explicitly warned against overselling mentoring as a fix for serious organizational problems. Mentoring helps; a mentoring program is not, by itself, a transfer pipeline. Reverse mentoring, the fashionable variant where juniors teach seniors, rests on an even thinner base: a 2021 review by Chaudhuri, Park and Johnson found the reported benefits real but the evidence predominantly qualitative and conceptual. Calibrated expectations are part of honest design.
The conditions under which transfer happens
If judgment moves at all, it moves the way it was built: through repetitions of real work, done together, with commentary. The apprenticeship, not the archive. That has an organizational precondition that knowledge-management budgets ignore: people share what they know when it is safe and normal to do so, and hoard it when it is not.
A 2022 multilevel study by Hao and colleagues, covering 282 individuals in 37 virtual teams at three Chinese internet companies, found that trust in coworkers increased knowledge sharing, and that the relationship was fully mediated by team psychological safety. Trust did not produce sharing directly; it produced safety, and safety produced sharing. One study in one context, but it names the mechanism every veteran recognizes: no one volunteers the real reasons, the mistakes, the rule-of-thumb that contradicts official policy, into a team where candor is expensive. A transfer program layered onto an unsafe team transfers the sanitized version, which is the version nobody needed.
There is also a seniority-specific poison: an expert whose remaining value is her monopoly on knowing has a rational incentive to keep the monopoly. Organizations that treat seniors as walking risk registers, valuable only until de-risked, teach exactly that lesson. The ones that get transfer treat teaching as the senior's next act of mastery, visible and rewarded, not as the extraction that precedes disposal.
Transfer as a staffing decision
All of which converts knowledge transfer from a late-stage project into an early staffing pattern. The working rules are unglamorous.
Pair on live problems, years out. The successor works the same incidents, negotiations, and design calls alongside the veteran while stakes are real and the veteran still enjoys the work. Frequency beats intensity; a hundred small repetitions outperform a five-day handover.
Have the veteran narrate decisions, not write summaries. Commentary attached to a live case ("I'm ignoring the metric this time, and here's the smell that tells me to") is the only format in which pattern knowledge survives contact with paper.
Keep the documents, but demote them. Write down the facts layer, indexed, thin, maintained. Stop pretending the other layers fit in it.
And accept the residue: some judgment will leave anyway. The satellite question, whether the departing expert remains reachable, contracted, welcome, is the honest backstop, and it is a relationship decision, made long before the farewell cake.
What transfers is what was shared while it still mattered. Everything else is a wiki.
Sources
- Eby, Allen, Evans, Ng & DuBois (2008). Does Mentoring Matter? A Multidisciplinary Meta-Analysis (Journal of Vocational Behavior 72, DOI 10.1016/j.jvb.2007.04.005). 112 studies, 116 samples: outcomes favorable across domains but effects generally small (workplace rc ≈ .03-.19); authors warn against overselling mentoring. https://pmc.ncbi.nlm.nih.gov/articles/PMC2352144/
- Hao, Zhang, Shi & Yang (2022). How Trust in Coworkers Fosters Knowledge Sharing in Virtual Teams (Frontiers in Psychology). 282 individuals, 37 virtual teams: trust raises knowledge sharing, fully mediated by team psychological safety. https://pmc.ncbi.nlm.nih.gov/articles/PMC9479451/
- Chaudhuri, Park & Johnson (2021). Is reverse mentoring a panacea to all? Findings from literature review (European Journal of Training and Development, DOI 10.1108/ejtd-01-2021-0005). Reverse-mentoring benefits reported for engagement, inclusion and knowledge sharing, but the evidence base is predominantly qualitative/conceptual. https://doi.org/10.1108/ejtd-01-2021-0005