ERGATHA RESEARCH ANTHOLOGY, DESIGN & METHODS

Borrowed Judgment

What moves between an experienced operator and a first-time buyer, and what form it has to be kept in.

A note published before the work it describes. Anthology is Ergatha's research and convening programme. This note sets out two decisions behind its design: how we intend to get at the part of an operator's knowledge that does not usually make it into words, and what form that material has to take to remain useful in a market that keeps changing. It reports no findings, because we have not yet run the sessions. We are publishing the method first so that the people we are asking to take part can see what they would be walking into, and tell us where it is naive.
01 · The problem

Judgment we do not have

A small firm buying and operating established businesses in India faces an uncomfortable arithmetic. The decisions are consequential and mostly irreversible, the number of people inside the firm who have made them before is small, and the gap will not close through experience in time, because the relevant experience accumulates at roughly one deal a year.

The conventional answers are to hire the judgment, which is slow and mostly available to firms that already have a record, or to read it. We have done a good deal of the reading. It is useful and it is not what we are short of.

The third answer is to borrow it: convene people who have made these calls, put a real problem in front of them, and try to get something back that survives the drive home. That is Anthology, and it rests on a question with a substantial literature behind it. When an experienced operator knows something, how much of it can actually move to someone else, and by what route?

02 · What does not transmit

The reports already have the easy part

Michael Polanyi opened The Tacit Dimension in 1966 with the observation that we can know more than we can tell, and argued that skilled performance rests on knowledge the performer cannot fully articulate. Gilbert Ryle had drawn a related line in 1945 between knowing-that and knowing-how. David Autor gave the idea its modern name in 2014: Polanyi's Paradox, the gap between what people can do and what they can specify.

Anyone who has sat with a good operator recognises it. Ask why they walked away from a business and you get a reason. Ask what the reason really was and, if they are honest, you often get a pause. The account is a reconstruction. What actually fired was a pattern they had seen before and could not name at the time.

The consequence for a research programme is unglamorous. A format designed to extract statements will reliably retrieve the layer that was easiest to state, which is the layer already sitting in the reports. The transcript looks like knowledge. Much of it is the residue left after the valuable part failed to make it into words.

Stated plainly

If Anthology's output is a set of well-organised quotations about what makes Indian small businesses succeed, the programme will have been an expensive way to reproduce a literature review. That is the failure mode we design against, and it is the one easiest to mistake for success, because the transcript is long and everybody enjoyed the session.

03 · What is recoverable

Most of it is not as tacit as it looks

The pessimistic reading of Polanyi would stop the programme before it starts. It does not survive contact with the work that came after him.

The sociologist Harry Collins spent three decades on this problem and concluded that Polanyi's single term was concealing three different things with very different prospects for transmission. The weakest he calls relational: tacit for ordinary social reasons rather than deep ones. People keep things to themselves, tell half the story, or most often simply fail to mention something because it never occurred to them that anyone would not already know it. Collins is explicit that there is no insurmountable obstacle here. With enough effort, any piece of relational tacit knowledge can be made explicit.

The second kind, somatic, is tied to the body. It matters enormously in craft and almost not at all in capital allocation. The third, collective, is the hard residue: held not by the individual but by the community they move in, and acquired by being embedded in it rather than by being told anything.

Figure 1 · Three kinds of tacit knowledge, and what reaches them
WEAK · RELATIONAL Tacit by circumstance Nobody asked. Nobody thought to say it. It is a trade secret. Explaining it takes an hour. FULLY RECOVERABLE Reachable by disciplined interrogation. This is the largest share of what we want. MEDIUM · SOMATIC Tacit by embodiment Held in the body. Describable in principle, unusable as description in real time. NOT OUR PROBLEM Capital allocation is not a physical skill. STRONG · COLLECTIVE Tacit by location Held by the community, not the person. What counts as normal, what a silence means, who is owed what in a family business. REQUIRES EMBEDDING No interview reaches it. This is the genuine constraint. Collins, Tacit and Explicit Knowledge, University of Chicago Press, 2010. Categories his; application ours.
The load-bearing reframe. The phrase "you just learn it by doing" is treated as a statement about the second and third columns. Collins' argument is that it usually describes the first. Something goes unarticulated because articulating it is effortful and nobody has ever required it, not because language cannot hold it. If that is right, the binding constraint on a programme like ours is not access to people. It is the quality of the interrogation, which is a thing we control.
The design claim, and how it fails

We are betting that a substantial fraction of what experienced operators know about Indian small-business acquisition is relational rather than collective: unstated because the occasion has never arisen, not because it is inexpressible. The first sessions are the test. If we run four properly structured sessions and recover nothing that was not already in the public literature, the bet was wrong and we will say so here.

04 · The target

We are not trying to become operators

A second correction changes what success looks like. Collins' other contribution is interactional expertise: the ability to converse expertly about a practice without being able to perform it, acquired through sustained immersion in the language of the people who do. He tested it on himself. After decades among gravitational wave physicists without ever doing physics, he submitted to an imitation game in which a panel of working physicists tried to pick his written answers out from a real member of the field. They could not do so reliably.

Between formal propositional knowledge and embodied skill lies interactional expertise: the ability to converse expertly about a practical skill, but without being able to practise it, learned through linguistic socialisation among the practitioners. Harry Collins, Cardiff University

Collins notes where this shows up in ordinary working life: it is much what the managers of technical projects possess. Enough command of the practice to argue, to press, to spot an evasion, while remaining unable to do the work.

That describes the job. We will not run the businesses we buy; the operators we hire will. What we must be able to do is judge those operators, argue with them productively, recognise when a plan is being dressed up, and allocate between them. That is an interactional standard, not a contributory one, and it is reached primarily through sustained argument with practitioners in their own language. Which is a description of a convening programme rather than a curriculum.

05 · What carries it

Formats ranked by what they move

If the recoverable layer is largely relational and the standard is interactional, the session design question becomes concrete: which formats actually move the material, and which only feel as though they do.

Figure 2 · Channel bandwidth
WHAT THE OPERATOR HOLDS already codified comps, structures, facts unarticulated which number is lying when to stop diligence what a seller means what good enough is the part worth convening for Read what they have published codified only Ask them general questions codified, and a little more Put a real problem between you the unarticulated moves a live memo, a real set of numbers, an argument Watch them decide something real highest, and rarest Line thickness is bandwidth, not certainty. Polanyi 1966 · Ryle 1945 · Lave and Wenger 1991 · Collins 2010 · Autor 2014
The asymmetry we design around. An expert responding to a specific wrong thing produces far more than the same expert answering a good abstract question. The error gives them something to push against, and the pushing surfaces the cue structure that an abstract question leaves buried. Every Anthology session is therefore built around an object rather than an agenda.

Where an individual conversation is the right instrument, the research is unusually specific about how to run one. Gary Klein's Critical Decision Method, developed in naturalistic decision-making research and used on fireground commanders, paramedics and structural engineers, works by walking a single non-routine incident several times: once uninterrupted, once to rebuild the timeline, once with probes aimed at what the expert saw and expected, and once changing a condition to find what breaks. One incident walked four times beats twelve good general questions, because generality invites philosophy and philosophy is a reconstruction.

06 · The form

Predictions decay. Conditions do not.

Suppose the elicitation works. A second question decides whether any of it reaches a decision: in what form should the captured thing be kept?

The intuitive answer is to record what the expert concluded. We think conclusions are close to worthless, for two independent reasons.

The first is structural. The environment is not stationary. The relationship between a business's observable features and its eventual cash generation is shaped by credit conditions, labour markets, succession norms, sector cycles and regulation, all of which move. A conclusion is a compressed prediction, and it inherits every assumption in the environment that produced it. It decays quietly and without notice. Necessary conditions are far more stable than sufficient ones. A statement about what must be true for a business to work survives a change in regime. A statement about what will happen does not.

The second reason is empirical. Philip Tetlock spent two decades collecting tens of thousands of specific forecasts from nearly three hundred experts and scoring them. The headline result is well known: the experts barely beat chance. The result that matters here is the second one. Accuracy depended not on what forecasters knew but on how they held it. Tetlock borrowed Isaiah Berlin's distinction. The hedgehog knows one big thing and reads the world through a single organising theory; the fox draws on many frameworks, tolerates ambiguity, and updates. Foxes outperformed hedgehogs consistently, and the gap widened over longer horizons.

This maps onto a distinction from the entrepreneurship literature that we had adopted for other reasons. Saras Sarasvathy, studying expert repeat founders, separated two logics. Causation rests on a logic of prediction: fix the goal, forecast the environment, select the means. Effectuation rests on a logic of control: start from the means in hand, work forward, treat surprise as material rather than error, and choose by what you can afford to lose rather than what you expect to gain.

Figure 3 · Two logics, and what each one leaves behind
CAUSATION · LOGIC OF PREDICTION fixed goal Means selected backward from a forecast. WHAT IT LEAVES BEHIND A conclusion Valid only while the environment holds. Decays silently. EFFECTUATION · LOGIC OF CONTROL means in hand Several reachable futures, one set of resources. WHAT IT LEAVES BEHIND A conditional structure Cues, thresholds, what is controllable, what is affordable to lose.
Sarasvathy, Academy of Management Review, 2001; Tetlock, Expert Political Judgment, 2005. Two literatures built for different purposes arriving at the same instruction. Do not ask an expert what will happen. Ask what they would need to see, what would make them stop, and what they would try to control. The first answer is a hedgehog's; the last three are a fox's.
What we keep, and what we thank people for

"I think this business will do well" is discarded. It is a conclusion with the reasoning compressed out of it, it cannot be checked, and it quietly stops being true when conditions move.

"I would walk if receivables stretched past sixty days two quarters running" is kept, as a threshold. "He says he wants the best price; he wants his son kept on and the name kept on the gate" is kept, as a reading of what a counterparty actually optimises for. "I have never made money betting I could fix a sales culture" is kept, as a boundary on what we allow ourselves to treat as controllable.

A judgment about the world is discarded. A judgment about what to watch, what to refuse, and what to shape is kept. The second kind is harder to elicit, which is the whole justification for the session design above, and it is the only kind a person who was not in the room can use two years later.

This is where the argument closes on itself. The reason we want effectual judgment rather than causal conclusions is not a stylistic preference for humility. It is that we are operating in a system that changes while we watch it, and a fixed worldview is the thing Tetlock found performs worst there. What we are trying to take from an experienced operator is not their model of the market. It is the conditional structure underneath the model: the cue, the threshold, the lever, the acceptable loss. Those survive the market changing. The model does not.

07 · Where it lands

Into a decision, and into the operating businesses

Two short notes on what happens to the material, since a method that stops at capture is a filing system.

Elicited thresholds become a set of necessary conditions that a prospective acquisition must pass before anyone scores anything, and failing one ends the process without a vote. What follows is not our invention: we use the mediating assessments protocol set out by Kahneman, Sibony and Sunstein in Noise. Decompose the decision in advance into assessments kept as independent as possible, anchor each in a base rate rather than the story of this deal, score them privately before anyone speaks, discuss, rescore, and only then form a view on the whole. Kahneman's phrase for the last step is the useful one: do not eliminate intuition, delay it.

The second note concerns where any of this can be tested. Investment decisions are a poor teacher: we will make few of them, each resolves over years, and by then conditions have moved. Operating decisions inside the businesses we own are the opposite. Pricing, collections, scheduling, supplier terms and hiring resolve in weeks, repeat often, and produce a reasonably clean signal. The portfolio is our laboratory before it is our proof. Cues and thresholds get applied first where they can be checked within a quarter, and only what survives is promoted into the acquisition gate. It is slower than building the investment layer first. We think it is the only version that is not self-deception.