The broadcast email problem: what "does anyone know about X?" really costs

Every firm has an expertise search system. In most of them it is an email that begins "Apologies for the broad distribution".

It is worth costing this properly, because it is treated as free. Nobody budgets for it, no system reports on it, and it never appears as a line in any analysis of why margins are what they are.

The arithmetic

Take a firm of four hundred. Someone sends a query to a practice-group list of a hundred and twenty people asking whether anyone has experience of a particular regulatory regime.

Thirty seconds of attention each is an hour of firm time, most of it from people for whom the answer is no. That is the visible cost, and it is the smaller one.

The larger costs are structural.

Selection bias in the replies. The people who answer a broadcast are the people with capacity to answer a broadcast. That correlates negatively with being the busiest expert on the subject. You reliably hear from the helpful, and unreliably from the relevant.

Silent failure. If the right person is on a client site, is on leave, filters the list, or has left the firm, the query returns nothing and the sender concludes the firm has no experience. That conclusion is usually wrong, and it is acted on: the team scopes the work as new, prices it as new, and rebuilds something the firm already owns.

Recency bias. Broadcasts surface what is top of mind. A colleague who ran the definitive engagement in 2021 may not connect the query to that work at all, because they have described it to themselves under different words ever since.

It does not accumulate. Every query starts from zero. The answer to the same question next quarter costs exactly as much as it did this quarter, and the person who answered last time may be gone.

The broadcast email is not a search. It is a poll of who happens to be free.

The 5.3 hours

The published figure for this is that knowledge workers lose around 5.3 hours a week either waiting on information from colleagues or recreating knowledge that exists but was never documented. Treat it as an order of magnitude rather than an accounting figure, since most of these numbers come from vendors, ourselves included.

Even discounted heavily, the shape holds. That is more than half a day per person per week, in an industry that sells time, spent on the retrieval failure rather than on the work. In a four hundred person firm it is not a productivity annoyance. It is a headcount.

Why the alternatives in place do not work either

Firms that recognise this usually have tried two things.

The capability database. Everyone completes a profile listing their skills and sectors. It is accurate on the day it is created and decays immediately, because nobody returns to it after an engagement to add what they just learned. Worse, it is self reported: in an organisation where staffing follows stated capability, understating yourself is career limiting, so everyone claims broadly and the database loses the ability to discriminate. A search returns forty names, which is the same as returning none.

The document search. Full text search over the shared drive. This finds documents that mention the regime, which is genuinely useful and stops one step short of the answer. The searcher now has eleven documents and still has to work out which of them represents real depth, and who was actually responsible. Authorship metadata typically names whoever last saved the file, which is often an analyst formatting a deck.

The academic work on expert finding, reviewed systematically in the MDPI survey, has been circling this for two decades and lands consistently in the same place: expertise has to be inferred from evidence of work done, not from claims and not from document proximity.

What would end the broadcast

The replacement has to beat the email on three dimensions at once, because the email is genuinely good at two of them and that is why it persists.

It has to be faster than a day. This is the easy one, and it is why any lookup at all wins.

It has to be more accurate than the volunteers. That means ranking by evidence of delivery, not by self assessment and not by how often someone's name appears near a keyword.

It has to be trusted without a follow-up. Which means every name comes with what that person actually did: the engagement, the client, the date, the document. The asker should be able to satisfy themselves in one click, without contacting anyone.

Get all three and the broadcast stops, not because it is banned but because it becomes obviously slower than the alternative. That is the only way behaviour like this ever changes in a firm. Policy does not do it. Being faster does.

What this looks like in practice

In OrgAtlas, capability is not claimed, it is accumulated. When a proposal, report or thread comes in, the people, engagements and capabilities in it are extracted and linked, and each link carries the document behind it. A colleague's experience in a subject is therefore a consequence of what they have delivered rather than of what they said about themselves at induction.

Two consequences follow that matter for this problem specifically.

The first is that the running order is meaningful. Because every capability carries the weight of the evidence behind it, the person who has genuinely led four engagements sits above the person who was copied on one, without anyone having to adjudicate.

The second is that the gaps are honest. Where the firm has no proven experience in something, the record says so plainly rather than producing a plausible name. That matters more than it sounds: the whole value of a fast answer is destroyed if you cannot rely on a negative.

An expertise system is only useful if "we have not done this" is one of the answers it can give.

The broadcast email survives because it is honest by accident. Silence means nobody replied, and everyone knows how to interpret that. Any system replacing it has to be honest on purpose.

Next: expertise location systems, an old category with a new reason to work