Human performance technology

A field that started by being wrong.

Sixty years of one question

The discipline did not set out to be what it became. It was founded to promote a teaching method, and the whole of its history is the record of that method turning out to be the answer far less often than anybody expected.

In 1962 a group of military and civilian training people in Texas founded a society for programmed instruction and began publishing its journal the same year. Programmed instruction was then the most serious idea in training: sequence the content correctly, test at every step, and learning becomes engineering rather than art.

It worked, within its limits, and the limits arrived quickly. Practitioners kept meeting problems that better instruction did not touch: people who already knew what to do and did not do it, work that failed between two competent departments, incentives that quietly paid for the wrong behavior. By 1973 the society was publishing under a different name, the National Society for Performance and Instruction, conceding, in one added word, that instruction was one intervention among several. In the mid-1990s “instruction” disappeared from the name entirely: the International Society for Performance Improvement.

A field that is honest about having been wrong

That sequence is worth dwelling on, because it is rare. A professional body founded to advance a technique publicly abandoned the technique as its organizing idea, twice, and kept the membership.

It is also the reason the discipline can be trusted on this particular question. A field that sells training and concludes that training is rarely the answer is not arguing in its own interest. It arrived there by watching what happened when it kept prescribing the same remedy, and it wrote the disappointment down.

Harless said it most bluntly in 1985, in an article whose whole subject is the myths the profession told itself: “Selling training, or its dumber brother, education, is hucksterism when there is no front-end present” (Harless, 1985, p. 4). He published it in the society’s own journal, about the society’s own practice.

The question the field has been asking for sixty years is therefore not “how do we teach this better?” It is: what is actually stopping the work from happening, and is it anything we know how to sell?

Four people and what each of them broke

The discipline was assembled by people working on different problems in different industries. What makes their agreement interesting is that none of them was trying to agree with the others, and each arrived by breaking something the others had left intact.

What none of them could see alone

Taken separately each of the four is incomplete, and the incompleteness is instructive because it is still visible in how organizations argue today.

Gilbert’s six conditions are a superb instrument for examining a job and a poor one for examining a handoff, because they take the performer as the unit and a handoff has no performer. Rummler’s levels find the handoff immediately and are less help once you are standing in front of a single person who is struggling. Deming establishes that the system dominates but does not tell you which part of it; his arithmetic is a prior, not a diagnosis. And Harless insists on analysis before design without specifying, in any detail, what the analysis should ask.

Put together, each covers the others’ blind spot. That is what the five questions are: Harless’s insistence on a front end, Gilbert’s conditions, Rummler’s levels and Deming’s prior, arranged in the only order in which they answer each other. Mager and Pipe turned the same logic into a decision sequence for practitioners, and their formulation, describe the discrepancy before accepting the remedy, is still the shortest statement of the method (Mager & Pipe, 1997).

And two who settled what the four had left open

The four broke the training reflex. Two others closed a question none of the four had answered: what counts as a need in the first place, and whose job it is to ask.

Nobody else has been added, and that is a decision. The discipline has a long roll of translators and packagers beyond these six, and a page of names is what a field writes when it wants to look established rather than useful.

Why it is called a technology

The word is not decoration and it is not borrowed from computing. It is a claim about what the method produces, and it is a demanding one.

A technology, in the sense the founders meant, is a body of practice in which the same inputs produce the same output regardless of who applies it. Not a craft, where the result depends on the practitioner’s judgement, and not a school of thought, where it depends on the practitioner’s convictions. Two competent analysts, given the same organization and the same problem, should reach the same causes.

The unit of that technology is the accomplishment: not what people do, but what they leave behind (a distinction number 05 works operationally, because it is where most analyses go wrong in the first ten minutes).

That is the standard the field set itself, and it is the standard by which its own practice most often fails. Anyone who has watched two consultants examine the same situation knows that they frequently do not converge. The discipline’s answer is that convergence depends on following the order, and that most non-convergence is the result of somebody skipping a question.

Ten standards, and the one that gets squeezed

The clearest statement of what this discipline requires is not a definition but a set of standards. The International Society for Performance Improvement certifies practitioners against ten of them and publishes them openly; they are reproduced below in the society’s own wording, with a line on each of what it asks for (ISPI, 2026).

Read as a flat list of ten they look like ten items of equal weight, which is misleading. They group into four obligations, and the order of those four is the method.

What each obligation actually asks you to produce

A standard is only useful if it names something you can be shown to have failed to do. Here is what each of the four requires somebody to be able to put on a table.

Notice what the ten do not contain. No solution type is named anywhere: not training, not coaching, not a system, not a reorganization. The standards describe an obligation to find out and an obligation to prove, and they are silent on what the answer will turn out to be, which is the only honest position for a discipline that claims not to know in advance.

What an organization gets for working this way

The case for the discipline is usually made as savings, and savings are the least interesting part of it. Five changes follow, and four of them are structural.

What that is worth in money is a separate question, and number 13 does it. The point here is that the arithmetic is not the argument.

What the evidence actually supports

The discipline as a whole has never been tested as a whole. Two of its components have been, and the gap between them is where an informed reader will push.

One line of evidence deserves care rather than a footnote, because it is the one most often quoted at this discipline. At the organizational level the association between training and performance is real and small. Kim, Chang and Bell synthesized 159 studies covering just over 75,000 organizations and found a positive but modest relationship, ρ = .13, varying substantially with what was trained, who was trained, how and when the outcome was measured, and the industry it happened in (Kim et al., 2025).

Why this series argues from effect sizes in neither direction

The more important finding is methodological, and it has a consequence for how this series is allowed to argue: one that cuts against us as much as for us.

Reviewing 207 quantitative studies across forty years, Garavan and colleagues conclude that the field’s central difficulty is that training cannot be separated from the system it sits in: the effects are interconnected, time-dependent and bidirectional, and a design that treats training as an isolated cause of firm performance is measuring something the world does not contain (Garavan et al., 2021). Their proposal is to put the system back into the research, which is the same move this discipline makes in the analysis.

Why this field has few studies, and why that is not a defect

A reader who checks the references will notice something. The evidence cited across these sixteen numbers is almost entirely meta-analytic, and almost none of it is about performance analysis itself. That is deliberate, and the explanation is part of the argument rather than an apology for it.

Medicine can run trials because it has standard diagnoses and standard interventions. A trial compares the same named treatment for the same named condition across many patients, and the whole apparatus of evidence-based practice rests on that being possible. Performance analysis has neither. Every case is one organization, with its own measure, its own process, its own history and its own binding condition, and the intervention that follows is designed for that case and for no other. There is no standard diagnosis to sample and no standard intervention to randomize.

That is not a weakness of the discipline. It is what the discipline is. A method whose entire purpose is to establish that this problem, in this organization, has this cause cannot also produce a population of identical cases for someone to compare.

The nearest thing to an exception proves the point. Pritchard’s Productivity Measurement and Enhancement System, a structured method in which a unit builds its own measures and reviews regular feedback against them, has been measured at scale: 83 field studies across three decades, large improvements in productivity, effects that persist for years, and results that vary most with how faithfully the method was followed (Pritchard et al., 2008). What could be standardized, the measurement-and-feedback loop, proved measurable, and it works. What cannot be standardized is the diagnosis this series is about: ProMES begins after that question, inside a unit that already knows its own outputs.

The strongest published case against doing this at all

Cause analysis has a serious critic and it should be read rather than avoided. In BMJ Quality & Safety, Peerally, Carr, Waring and Dixon-Woods set out why root cause analysis in healthcare has so often failed to produce improvement (Peerally et al., 2017). Several of their objections apply to what this series proposes. Four of the five below are answerable. The fifth is not: every analysis here produces a finding for one organization and one decision, and nothing accumulates across them. Anybody adopting this method is buying a local instrument rather than a contribution to a body of knowledge, and that is a real limitation of the discipline as practiced.

The performance consultant

The posture is not new and it is not a job title. The five sections before this one are its history. It is also one of the roles a learning function has to be able to play if it is going to be of use to a business: the destination piece at the end of this series sets out the four models a function can run, and this posture is what carries it across the divide.

What matters is the posture rather than the competence, and the posture is the hard part. It is the one a good investigator takes: do not begin from a suspect, begin from the thing that happened, establish what is actually known, and treat the most obvious explanation as the one most in need of checking (precisely because it is the one everybody has already accepted).

The awkward implication is that the posture is often unwelcome. Establishing a cause means telling somebody that the thing they asked for will not reach it, and doing that on every request rather than on the two a year where there is time. That is a standing capability, not an occasional project, and it is the reason the last paper in this series exists.

What this is not

A discipline that claims sixty years and a certifying body should be plain about its boundaries. Four of them.

• It is not a change methodology. It establishes what is wrong and what would reach it. Getting two departments to alter a handoff is political work with its own literature and its own failure rate, and nothing in the standards makes it easier.
• It does not guarantee convergence. The claim in section 03 is a standard, not a finding. Two competent analysts following the order will agree far more often than two consultants following their instincts, but “far more often” is not “always”, and anybody who promises otherwise is selling.
• It is not a substitute for management. Deming’s point cuts both ways: if most of what goes wrong belongs to the system, then most of what fixes it is a management decision. An analysis can make that decision informed. It cannot make it.
• The history here is compressed. Four people stand in for several dozen. The chronology is deliberately coarse: we give only what independent library records carry (the founding year, the journal, and the names the society published under) and leave out the founding-meeting detail that circulates in secondary accounts. The standards are quoted from the society’s current wording (ISPI, 2026).

Causes, or what you can see?

Ten minutes, on your own organization rather than on a problem. Six questions, and the pattern of the answers tells you more than any of them separately.

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 books behind the series are listed here in full, in the number that carries the lineage, so that they do not recur in every reference list. Kostbaar misverstand, which much of the thinking here comes from, was never translated.

One date, held to what the records support

The society’s founding year and its changes of name are taken from library catalogue records of its own journal and newsletter, which show the society publishing from 1962 and under the name National Society for Performance and Instruction by 1973. Its history page could not be retrieved at the time of writing, so no more precise date is given here than those records support.