Knowledge that does not flow is not an asset

An equipment manufacturer in Brazil, 420 people. In March, leadership buys a knowledge-base tool and announces the company will document its processes. Each area gets a target of ten pages by June. By June there are 214 pages. By December, monthly access is 19 views, almost all on the phone-extensions page.

That same December, the most experienced technician in customer support fields 40 calls from colleagues asking for help with an old machine model. None of those 40 answers is in the 214 pages. Knowledge circulates in the company every day — just not through the path built for it.

01The idea: there is a market for knowledge #

Thomas Davenport and Laurence Prusak published Working Knowledge in 1998, after following dozens of corporate knowledge-management initiatives — many of them failures. The book's most useful idea is treating knowledge circulation as a market, with buyers, sellers, and brokers, driven by reciprocity, reputation, and altruism, and subject to price.

Buyers are people who need to solve a problem. Sellers are people who know. Brokers are people who don't hold the knowledge but know who does — and in practice they are the invisible infrastructure of any mid-size company. The price is never money: it's time, expected reciprocity, the recognition of being recognized as someone who knows.

That's how you explain the dead wiki. Writing a page costs real time for whoever is already the most in-demand person in the area; the return is abstract and future; and in many companies, being irreplaceable is a form of security. On the other side, searching the knowledge base costs more than messaging someone who knows — and the message comes with context attached. The market resolves toward the cheapest path, and the cheapest path is a person.

Hansen, Nohria, and Tierney described in 1999, also in Harvard Business Review, two distinct strategies for handling this: codifying knowledge into reusable documents, or connecting people so they can talk. They argued the choice depends on the type of work — and that trying to do both with equal intensity usually produces neither.

02What actually makes knowledge flow #

Three conditions show up whenever circulation works, and none of them is the tool.

The record has to happen inside the work. The answer the technician gave a colleague was already written once — in the chat, in the ticket, in the note on the service order. When that record stays in the flow, attached to the problem, it's reusable. When it requires a second step, in another system, it doesn't happen.

Searching has to be cheaper than asking. If finding something takes three minutes and asking takes thirty seconds, everyone asks. The comparison is made by the buyer, not by whoever designed the system.

Whoever teaches needs to gain something visible. Recognition in performance review, protected time on the calendar, explicit authorship. As long as teaching is pure cost for the person who knows, the seller exits the market — and has every right to.

There's also an honest limit: not all knowledge is codifiable. Nonaka already argued, in 1991, that much of what an organization knows is tacit and transfers through shared experience and practice, not text. We cover this in what only Paulo knows. Documenting everything is the wrong goal; documenting what repeats is an achievable one.

03The math on the question that keeps repeating #

At the scale of the 420-person manufacturer, across technical and administrative areas. This is a mid-size Brazilian company (figures in Brazilian reais). Stated assumptions:

Want to see how this looks inside a real operation? Explore the platform.

The math: 160 people × 4 questions × 44 weeks gives 28,200 occurrences a year; × 17 minutes, about 7,900 hours; at R$ 65 an hour, roughly R$ 520 thousand, plus the concentrated time of the three or four specialists who answer most of it and stop their own work to do so. Cut it in half if the assumption seems high — the number still exceeds any software license.

There's an additional risk that isn't financial in the short run. When 40 answers a month depend on one person, that person leaving isn't an HR problem: it's an operational interruption with an unknown date.

04Where this breaks down in practice #

1. Documenting becomes a volume target. Ten pages per area produces 214 documents nobody wrote to be read.
2. The knowledge base sits far from the work. Another login, another system, another tab. The shortest path is still a direct message.
3. Nobody owns the update. The page is right the day it's written and wrong six months later; one discovered error destroys trust in all the others.
4. Whoever teaches gains nothing. The specialist answers 40 times a month and is evaluated on their own output, not on what they unblocked for others.
5. The exceptional gets documented, not the frequent. The manual covers the rare case; Monday's daily question is nowhere to be found.
6. Tool gets confused with incentive. The platform changes every two years, and the internal market stays exactly the same.

05What data-driven management answers #

Knowledge circulation improves when the record is a byproduct of the work, not an extra task. A ticket with the solution described, a process with the instruction inside its own step, a decision logged along with the reason: none of that asks anyone to stop and document, and all of it stays searchable afterward.

The second effect is seeing the market itself. Which questions repeat, who is carrying how many answers, where the operation depends on a single person — these are measures that exist once the exchange happens inside the system. On knowledge that doesn't fit in text, see what only Paulo knows; on how that shortens a new hire's ramp-up, a new manager's ramp.

At Relevanti, the record lives inside the processes, operations, and tasks modules — see the platform, the breakdown by area in solutions, or bring your ten most repeated questions to a conversation.

Before buying the next documentation tool, answer this: what does the person who knows gain by writing, and what does the person who needs it lose by searching?