Environment before capability

Telling people more changes almost nothing.

The reflex, and why it is reasonable

When a result moves the wrong way, the first instinct is to inform, to persuade, or to train. That instinct is not lazy. It is the one option that is visible, purchasable and available this week, and every organization is built to supply it.

Look at what the alternative would require. Changing what someone can reach at the moment of the decision means touching a process, a system, an authorization, a rota or an incentive: each owned by somebody else, each with its own committee. Changing what someone knows requires a supplier, a date and a room. One of those can be arranged this quarter. The other cannot.

So the request writes itself, and it is sincere. The manager has watched the problem for six weeks, has a theory, and puts the theory in the only form the organization reliably accepts. Nobody in this chain is being careless. They are choosing the intervention that the system makes available.

The reflex has a long record of being named and surviving anyway. Harless told the profession in 1985 what selling training without a front-end analysis amounted to, in the society’s own journal and in language nobody could mistake: number 02 quotes the sentence. Four decades later the request still arrives with the remedy attached.

The uncomfortable part is not that this is wrong in principle. It is that the size of the effect is now measured, across hundreds of studies and dozens of domains, and it is very small.

What the evidence actually says

In 2024 Dolores Albarracín and colleagues did something the field had not done before. Rather than run another study, they gathered the available meta-analyses of behavior prediction and intervention efficacy across domains, and ranked what actually moves behavior and what does not.

Read that figure from left to right and it inverts the way most performance budgets are built. Handing people information has a negligible effect. Changing what people think in general has a negligible effect. Teaching a broad, non-specific skill has a negligible effect. So do rules and sanctions, which is the other reflex, if information will not do it, forbid the behavior.

Effects only start to appear once an intervention stops addressing the mind and starts addressing the situation. Money helps a little. A concrete, behavior-specific skill helps a little more. A changed routine helps more again. Arranged help from colleagues helps more still. And the largest effects belong to interventions that remove whatever was standing between the intention and the act.

Two distinctions that keep it honest

A figure this convenient invites two objections, and both are worth answering before anyone builds a decision on it.

A determinant is not an intervention target

The strongest single predictor of how well someone performs is not in that figure at all. In the training-transfer research it is cognitive ability, which tracks how much of a program survives contact with the job more closely than any feature of the workplace does (Blume et al., 2010). That does not weaken the argument. You cannot buy an intervention that raises cognitive ability. Something that explains why people differ, but cannot be moved, is useful for selecting people and useless for improving performance.

The same result, found inside organizations

Albarracín’s evidence comes from health and policy behavior. The parallel research inside organizations reaches the same place, and it has been reaching it for fifteen years. Pooling eighty-nine studies of training transfer: whether the manager backs it, and what the climate is like in the place people return to, both track how much of a program survives, and the work environment weighs six times more heavily on open-ended work than on tightly scripted work, because the more discretion the job allows, the more the surroundings decide what happens (Blume et al., 2010). Support from peers, manager and organization together accounts for roughly a third of the difference in whether training transfers at all (Hughes et al., 2020), which is to say that the environment decides most of what a program achieves, and it keeps deciding long after the room has emptied.

And none of it is new

None of this is new. Gilbert set the priority in 1978: “only when we have made a proper analysis of accomplishments and their measures will we have any sensible reason to concern ourselves with behavior” (Gilbert, 1978, p. 73). Rummler established where those accomplishments leak: the seams between departments (Rummler & Brache, 2013). Three disciplines, three routes, one conclusion.

Five resources, one word

“Access” sounds abstract until you watch a working day. Then it resolves into five concrete provisions, each of which someone can arrange, and each of which is quietly missing somewhere in every organization. The five are our own grouping of what analyses keep turning up, not a published taxonomy; what they are worth is in the middle column.

Four of the five behave alike. They are visible once you look, they belong to someone who can be named, and they can be changed within a quarter by a manager with a modest mandate. Unclear information, missing tools, absent time and withheld authority are the ordinary furniture of performance analysis, and the fixes for them are ordinary too.

The fifth behaves differently, and it is the one this paper is really about. Expertise cannot be bought in a quarter, cannot be moved to where it is needed, and does not scale by hiring, because the people who have it are the same people whose time you are trying to protect. Every organization rations it, and almost none of them decided to.

That rationing is not a policy. It is geography and seniority. Whoever sits near the person who has seen the problem before gets an answer in ten minutes. Whoever sits elsewhere gets a manual. The difference between those two people is not competence, and it is not attitude. It is what they can reach.

How to test each one without running a project

A list of five is a checklist, which is worth little. What makes it an instrument is knowing how to test each one, what a positive test looks like, and roughly what it costs to fix. None of these five tests takes more than a day.

The cheap fixes are at the top, which is the opposite of where attention usually goes: an organization will run a development program before it will put a field on a screen.

Which of the five is actually binding

Usually more than one is missing and only one is binding: the one that, if you arranged it tomorrow, would move the measure on its own. Finding which is a matter of sequence rather than judgement.

Step two is the one people skip, and it is free. The exemplary performer has arranged the resource and stopped noticing it; the standard performer is standing in front of its absence every day and can describe it exactly. Ask one what they have and the other what stopped them, and the two answers name the condition between them.

Two engineers and three regions

A manufacturer runs the same production line at its main plant and at three regional sites. Same equipment, same procedures, same certification for every technician. And the same fault takes four hours to clear at the main plant and two and a half days everywhere else.

The reading offered by the situation is the familiar one. Regional technicians must be less experienced, or less thorough, or less committed. The request that follows is a three-day technical refresher for sixty people, and it is entirely defensible: nobody has a better explanation, and training is the explanation the organization knows how to buy.

Now the diagnosis. The technicians are certified and the procedures are identical, so knowledge is unlikely to be the constraint, and a two-minute check confirms it: on the rare occasions a regional team reaches the main plant by phone, the fault clears in about the same four hours. What differs is not what people know. It is that two engineers who have seen this failure two hundred times sit at the main plant, and everyone else has a manual that describes the procedure but not the reasoning.

The missing resource is access to expertise. And note what the refresher would have produced: sixty well-trained technicians returning to sites where the same two engineers are still unreachable, and a manager concluding a year later that the training did not stick.

The last scarce resource

Information became inexpensive, then abundant, then free. Tools became rentable by the month. Authority can be delegated by writing it down. Expertise allowed none of this, and until very recently there was no reason to expect it to.

Expertise is different because it is not a document. What the two engineers have is not the manual; it is the reasoning that decides which of nine plausible causes to test first, built from two hundred encounters with the same machine. That reasoning has always traveled in exactly two ways: by standing next to the person who has it, or by spending years acquiring it yourself. Both are expensive, both are slow, and both are strictly limited by where you happen to sit.

Which is why every organization, without deciding to, ends up with a competence gradient that follows its floor plan. The head office is where the answers are. The distance from head office is, in practice, the resolution time.

What this does not mean

An argument this convenient for one side deserves its boundaries stated first. Five of them, plainly.

• It is not an argument against learning. Where the cause genuinely is a shortfall in knowledge or skill, well-designed learning is exactly right and nothing substitutes for it. The evidence does not say teaching never works. It says teaching is a weak lever when the obstacle is not ignorance, which is most of the time, and is a claim about frequency, not about value.
• Training effects cannot be isolated anyway. Reviewing forty years of organizational-level research, Garavan and colleagues conclude that training cannot be separated from the system it sits in: the effects are interconnected, time-dependent and bidirectional (Garavan et al., 2021). That cuts both ways here. It supports the argument that the environment dominates, and it forbids anybody, including us, from settling the question with an effect size.
• The evidence comes from another domain. Albarracín and colleagues synthesized health, policy and social behavior, not work organizations. The ranking is stable within that body; carrying it across to a production line or a finance department is an inference rather than a finding, and this paper says so rather than dressing it as a result. It is a well-supported inference: Gilbert and Rummler arrived at the same ordering from inside organizations, by different routes and without the psychology.
• Access without competence is noise. Reaching an expert helps only someone who can act on what they hear. The five resources are conditions, not substitutes for one another; a technician who has never been certified is not helped by a faster answer.
• A system that carries reasoning can be wrong. It should therefore say how confident it is and what would raise that confidence, and the decision should stay with the person who owns the consequence. A tool that hides its uncertainty is a worse adviser than a manual.

And what to do about the one that is not settled

The third limit is an open debate rather than a settled question, and it deserves to be treated as one.

Whether a ranking established in health, policy and social behavior carries over to work organizations has not been tested directly. It could be. A replication in an operational setting (same design, outcome measures taken from the work rather than from self-report) would settle in one study what currently rests on inference, and the field would be better for it. Until somebody runs it, the defensible position is that the direction is well supported and the size of the effect in a workplace is not.

Which of the five is actually missing?

Fifteen minutes, on one problem you own. Take the group whose performance is below where it needs to be, and answer for them rather than about them.

Two problems, one method, sixteen papers

Two models carry the method. Gilbert’s six conditions ask what kind of cause; Rummler and Brache’s three levels ask where it sits. Both are sound and this series uses both, and it leads with Gilbert, because his unit is the accomplishment, which is what an organization already counts, and because his six conditions are the categories an organization can actually change.

Where the method comes from

Two further pieces close the set and carry no number, because each points at all the numbered papers. Before the guide sits Arets L&D Business Models, the destination piece for the people who receive the requests: four models a learning function can run, and the route from HR-driven to business-driven. Last sits Solutions that cannot work, the field guide, ordered by the sentence that lands on your desk: seventeen forms a performance problem arrives in, each with an address in one of the numbered papers. Three orders in all: the destination you choose, the question you are on, and the form that landed on your desk.

Sixteen papers, six parts

Number 01 is the map and number 02 is the foundation. From number 05 on, every number is depth on exactly one step of the method. The filled block is the paper you are holding; the two unnumbered pieces close the set.

Where every claim in this paper comes from

The three books named earlier in this paper are listed in full in number 02, so that they do not recur in every reference list.