What is a transactive memory system? A field guide for professional services firms

Ask a managing partner what their firm knows and you will get a number. Two hundred people, thirty years, four hundred clients. That is a description of a payroll, not of knowledge. The useful question is narrower and much harder: when a specific person needs a specific thing the firm has already learned, can they get to it?

Organisational behaviour has had a name for the machinery that answers that question since the mid nineteen eighties, and it is not knowledge management.

Where the idea came from

Daniel Wegner introduced transactive memory in 1987, and the first evidence came from an unlikely place: intimate couples. Wegner noticed that long term partners divide remembering between them without ever agreeing to. One holds the diary and the birthdays, the other holds the bank details and the car. Neither has memorised the other's half. What each has memorised is where that half lives.

The pair therefore remembers more together than either could alone, and the gain has nothing to do with anyone becoming cleverer. It comes entirely from the directory. Break the pair up and both halves become unreachable, which is why separation is so often described as a kind of forgetting.

Wegner's insight was that the unit of memory in a group is not what any individual knows. It is the distributed system of who knows what, and how the group gets at it when it needs to.

A firm does not lose knowledge when a partner leaves. It loses the directory entry that made the knowledge reachable.

The idea moved out of psychology and into organisations quickly. Kyle Lewis built a measurement scale for it in the field. Linda Argote and Yuqing Ren mapped twenty five years of the research into a single framework in the Academy of Management Annals, and later argued in the Journal of Management Studies that a transactive memory system is one of the microfoundations of a firm's dynamic capabilities. That is a strong claim, and it is worth reading plainly: the ability to know who knows what is not a nice property of a well run firm. It is part of what makes a firm capable of anything at all.

The three moving parts

Field research settled on three dimensions. They are worth learning as separate things, because they fail separately and they need different fixes.

A group high on all three performs measurably better than a group of equally capable individuals who lack the system. The classic demonstration is a training study by Liang, Moreland and Argote in the nineteen nineties: groups trained together assembled radios better than groups whose members were trained individually, and the advantage was explained by the transactive memory the group training built, not by any difference in individual skill.

Nobody in the first group knew more. They knew who knew.

What the three parts look like in a firm

Firms are unusually good examples, because a professional services firm is a transactive memory system that happens to also have a billing department. Its entire product is expertise that no one person holds.

Specialisation in a firm is real and deep. A tax partner, a transfer pricing specialist, a public sector bid lead and an actuary hold genuinely different knowledge, and that division is the firm's whole commercial reason to exist. This dimension is almost never the problem.

Credibility is where firms are quietly strong and quietly fragile. Partners trust the colleagues they have worked with, and only those. Trust is earned in engagements, so it travels along the same lines the work travelled, which means credibility is dense inside a practice group and thin between them. A partner will happily rely on a colleague they shared a project with in 2019 and will independently rebuild the same analysis rather than rely on a colleague two floors up they have never met.

Coordination is where firms fail. Retrieval in most firms is a broadcast email, a corridor conversation, or a partner remembering something at the right moment by luck. There is no index. There is a strong, accurate, richly detailed directory of who knows what, and it exists exclusively inside the heads of the people who have been there longest.

Why it degrades as the firm grows

Wegner's couples worked because of constant contact. Two people who talk every day maintain their directory for free, as a side effect of living together. Every mechanism that keeps a transactive memory system current is a mechanism of proximity, and every one of them weakens with scale.

At twenty people, everybody has worked with everybody. The directory is complete and self maintaining. At two hundred, most pairs of people have never worked together, so most directory entries were never created. At a thousand, across offices and time zones, the firm is not one transactive memory system at all. It is thirty small ones that share a logo, a finance function and a set of clients that each of them can only see one face of.

Three things accelerate the decay:

• Turnover. Every departure deletes directory entries wholesale. Not just what the leaver knew, but everything they knew about who else knew things.
• Hybrid and remote work. The corridor was doing real work. Incidental contact is how a directory entry gets created between people who have no reason to be introduced.
• Growth by acquisition. Two firms merging do not merge their directories. They operate two, and the seam between them is invisible to everyone except the clients who notice they are being sold to twice.

The part firms get wrong

The instinctive response to all of this is to build a repository. Capture the knowledge. Write it down. Every firm has tried this, and the graveyard of dead wikis, unused precedent libraries and abandoned capability databases is the result.

Repositories fail because they attack the wrong dimension. They try to fix specialisation, by moving specialist knowledge into a place where anyone can read it. But specialisation was never broken. The tax partner still knows tax. What is broken is coordination: the firm cannot find its way to the tax partner for the specific question in front of it right now.

You do not need to move the knowledge. You need to know where it is.

This is the difference between knowledge management and transactive memory, and it is not a semantic one. A repository asks every expert to do the expensive thing, which is to externalise what they know. A transactive memory system asks nobody to do anything except be findable. The first requires effort proportional to the knowledge. The second requires effort proportional to the index, which is very much smaller.

What a written down version would have to do

If you were going to build the firm's directory rather than hope for it, four properties fall straight out of the theory.

It has to build itself from work that is already happening. The couples in Wegner's research did not fill in forms. Their directory was a by product of living. Any system that requires fee earners to maintain it will decay at exactly the rate that billable pressure rises, which is to say immediately.

It has to record credibility, not just claims. A directory that says a partner "has experience in energy regulation" is worth very little, because the assertion has no weight behind it. A directory that says they led three specific engagements, and shows them, is worth acting on. The credibility dimension is not a nicety. It is one third of the theory.

It has to be honest about what is not there. A transactive memory system that reports coverage it does not have is worse than none, because the group stops asking. If the firm has never done a piece of work, the directory must say so plainly rather than produce something plausible.

It has to cover both halves. Who knows whom, and who can do what. Most commercial tooling in this space maps contacts and stops. Contacts are the credibility half at best. Without proven capability alongside it you have a social graph, not a memory system.

Next: the three failure modes, one per dimension, and what each one looks like from the inside

Why this matters commercially, not just intellectually

It is easy to file all of this under interesting theory. The commercial translation is short.

A firm whose transactive memory works walks into stakeholder meetings warm, prices work it has done before correctly, staffs engagements with the people who have actually done the thing, and sees the gap between what a client needs and what the firm has sold them. A firm whose transactive memory has decayed does none of those, and experiences the failure as four unrelated problems: business development is hard, margins are unpredictable, resourcing is guesswork, and clients keep buying adjacent services elsewhere.

They are one problem. The firm cannot find its way to what it already knows.

OrgAtlas exists because that problem has a shape, and the shape has been described in the literature for forty years.