Intervention selection

A remedy has a reach. Most are bought without asking what it is.

Every remedy has a reach

This is the shortest argument in the series and the one that decides the most money. A remedy acts on one condition. If the condition it acts on is not the one that is binding, the remedy cannot work, not because it is a bad remedy, but because it is pointed somewhere else.

Number 03 gives the synthesized evidence behind that: across domains, what moves behavior most is changing what somebody can reach, and what moves it least is adding to what they know (Albarracín et al., 2024). Read as a claim about frequency it is a strong finding. Read as a claim about reach it is a physical fact, and it is the second reading that governs this decision.

Say it as a physical fact and it stops sounding like a slogan. A training program changes what a person knows and can do. It cannot change how long a file waits between two departments, what the incentive scheme pays for, or whether the parts catalogue takes ninety seconds a line. A new system changes what people can reach. It cannot make anyone want to use it, and it cannot supply the standard nobody ever stated. Each remedy has a reach, and outside that reach it does nothing at all: however well it is designed, however good the supplier, however satisfied the participants.

Which produces the arithmetic that governs this decision. Two conditions have to be true for a remedy to move a number: it has to reach the binding condition, and it has to be executed competently. The organization spends nearly all its attention on the second and almost none on the first, and the second cannot rescue the first. A superbly delivered program aimed at a condition that was not binding produces excellent evaluations and an unchanged result, which is, in our practice, the most common ending in this field.

The rest of this paper is about the second test, in three parts: what the full menu of remedies actually contains, which of them reach each condition, and how to choose between the ones that qualify.

The menu is far longer than the habit

Ask an organization what it can do about a performance problem and the answer will be a short list: train, communicate, reorganize, buy a system, replace someone. Five options, arrived at by habit rather than by choice.

Now write down what could actually be done about a case where the standard was never stated. Put the number on the report the performer already receives. Put a worked example of good output where the work happens. Change the moment the check occurs. Give the decision to the person doing it rather than to somebody two levels up. Remove a step. Move a queue. Change what gets discussed in the weekly meeting. Alter what the system will not let somebody do. None of those is exotic, all of them are cheaper than a program, and none of them appears on the list of five.

Which produces a rule about who is in the room, and it does more work than any amount of analytical rigor upstream. If the only people present are the ones who supply training, the remedy will be training. A choice between remedies requires somebody who can commit a process change, somebody who can change a report, and somebody who can change a scheme, which is a meeting most organizations have never held about a performance problem.

What reaches each condition

Six conditions, and for each one a set of remedies that can physically touch it. Read the table as a filter rather than as a menu: it is far better at ruling options out than at choosing between the ones that remain.

And the three carried by the performer

These three arrive with the person or are built into them, which is why they are slower and dearer to change, and why the first table is where most of the money is saved.

The reach is one question. The level is the other.

A remedy can reach the binding condition and still land at the wrong level. That is not a refinement. It is the difference between buying a substrate and buying a result.

Instruction is the right answer when the binding condition is knowledge and skill, and the table says so without hedging. What instruction produces is competence in individual heads. Where the result the organization wants is also produced by an individual (an adviser who prices a case correctly, a technician who completes a repair) that is the end of the matter, and the reach question was the only question.

Most results are not like that. Safer handovers, fewer defects, faster adaptation, reliable service: these are produced by teams and held by the organization. There, individual competence is the substrate. Three further conversions have to happen and each can fail on its own: individual competence has to become team capability, capability has to become reliable collective performance, and performance experience has to be retained so the organization still has it next year. Number 14 carries the part of that work a remedy can plan for: implementation by the teams that run the process, and maintenance so the gain is still there next year.

So the selection question has two halves. Which condition does this reach, and at which level does it land? A remedy that reaches the right condition at the individual level, bought against a result produced collectively, will deliver exactly what it promised and evidence nothing the organization asked for. It is the most defensible failure in the profession, because every number in the report is true.

How many causes did you establish?

The question that gets asked is whether to buy one thing or several. It is the wrong question, and the evidence that is usually quoted in it does not answer it. The question that matters is how many causes you established.

Start with what needs no study, because the strength of this paper sits there. A remedy that cannot reach the binding condition cannot close the gap, not slowly, not partly, not with better execution. That is section 01, and it follows from what the remedy physically touches.

Now the distinction the evidence turns on, because two moves look identical in a proposal and are opposite in kind. Bundling is adding components to one target because each of them might help: a course, a communication campaign and a new dashboard, all aimed at the same complaint, none aimed at an established cause. Combining is one remedy per established condition when more than one is binding: the authority moved and the standard stated, because the analysis found both. The first is a hedge. The second is the analysis being acted on.

What the syntheses actually counted

On bundling, then. Twenty-five reviews of multifaceted against single-component interventions were examined together, and the researchers looked three separate ways for a link between the number of components and the size of the effect. None of the three found one (Squires et al., 2014). That was 2014, and the picture has since been redrawn rather than overturned: pooling 204 studies, multifaceted strategies came out slightly ahead of single ones and the difference did not reach significance, while the variation between studies ran from three-quarters to nine-tenths of the total, and the authors note that the studies which could not be pooled favored multifaceted throughout (Fontaine et al., 2024).

Read that heterogeneity rather than the average, because it is the more useful finding. When what a combination does varies that much between settings, the average tells you almost nothing about yours. What decides the outcome is not how many components were bought. It is whether each one was aimed at something that had been established.

Which is why we combine, and why that is not bundling

Our own practice runs against the headline and with the mechanism, and it is worth saying plainly rather than leaving a reader to infer it.

Most performance problems we are called to have more than one binding condition, and the pair recurs. Something in the work obstructs (a step, an authority, a queue, an incentive pointing the other way) and the conditions that would let a team build and hold the capability are absent. Acting on one of those and not the other reliably disappoints. Clearing the obstruction leaves a team that never converts what it can now do into how it works; strengthening the team learning leaves people who have learned something the work still prevents. Two conditions, two remedies, each measurable on its own. [our practice, from client work; not a measured comparison]

There is a reason that pair recurs. In most organizations the substrate is already in order: people have been trained, they know the rule, they can state it correctly in a test. What is missing is the conversion: from individual competence into team capability, from capability into reliable performance under real conditions, and from performance into something the organization still has next year. A course added to that is a component. A remedy aimed at the conversion that is failing is not.

Where the first rule has been put to a controlled test it also holds. The testing happened in clinical guideline work, because that is where somebody funded the trials, and pooling them shows that a remedy chosen after establishing what was in the way beat one chosen without (McHugh et al., 2026). Take the direction from that and not the size: a figure produced in clinical practice has no claim to hold at the same size in a claims department.

The syntheses, and what each one counted

Four selection errors that survive a good analysis

Every one of these produces a proposal that looks defensible, cites the analysis, and cannot reach the cause. They are worth naming because each has a specific counter-question.

There is a fifth pattern that is not an error but is worth recording: the remedy that reaches the cause and that the organization will not authorize. That happens, and the professional response is to write it down as the finding, record what was chosen instead and why, and say what share of the gap the substitute can reach. An analysis that quietly rewrites itself to match what was affordable has destroyed the only thing it had.

The laboratory that bought a portal

A regional diagnostic laboratory serving 340 general practices. The measure was the share of test requests arriving complete enough to process without a call back. Desired 95 percent; running at 71. The analysis was good. The remedy was not.

Questions four and five had done their work. The shortfall was uniform across practices, so not a small group of offenders. The requesting doctors could complete a form correctly when asked to, so not knowledge. And the request form itself presented eleven fields, of which four were required only for certain test types and the form did not say which: the standard existed and had never been stated where the work happened. First condition, binding, at the process level.

Completeness eighteen months later: 74 percent. Three points, on a project whose cost is not this paper’s object and which number 13 would have made visible before the decision rather than after.

What we should have done differently

This one is ours to answer for, because the analysis was ours and the recommendation was accepted in the room without the one sentence that would have held it.

The finding was written as “the requirements are not visible at the point of request.” That is true, and it is a description of a condition rather than a specification of a remedy. It was read by a project team, entirely reasonably, as a case for a better requesting system, and a better requesting system is something that can be procured, budgeted and led, which the two-week form change was not.

What the report should have carried, in the same paragraph as the finding, is the reach test in writing: any remedy for this cause must make the four situational requirements appear, named, at the moment the requester chooses the test type. A remedy that does not do this cannot close this gap, whatever else it improves. One sentence, and it would have traveled into the specification, because specifications are written from documents and not from meetings.

What this does not claim

This paper filters remedies. It does not choose between the ones that survive the filter, and five limits follow from that.

• The matching table is a practitioner instrument, not a validated mapping. No study establishes which solution type works for which condition; what the table encodes is the reach argument in section 01, applied case by case. It is strong at ruling out and weak at ranking what remains. [our formulation, drawn from client work; not a published mapping]
• The tailoring evidence is modest and from one setting. It supports the mechanism, there is nothing clinical about “a remedy chosen against an established cause beats one chosen without”, and it establishes no magnitude for commercial work. What would settle that is a trial in ordinary commercial work: units matched on a measure, half receiving a remedy chosen after a structured cause analysis and half receiving the organization’s default for that complaint, measured on a business metric over four quarters. A phased rollout supplies the design almost for free, and as far as we can establish nobody has done it.
• The option set in section 02 is ours, not a published taxonomy. It is what twenty years of analyses have turned up as options that reach each condition. It is useful for opening a choice that had closed, and it is not evidence that any particular option works.
• The level argument is a framework, not a trial result. That individual competence must convert into team capability, collective performance and organizational memory is a conceptual architecture, not a finding. It is testable and has not been tested as a whole.
• Nothing here prices anything. The choice between two remedies that both reach the cause is usually decided on cost against the value of the gap, and that arithmetic belongs entirely to number 13.

Six lines, before anybody signs anything

Twenty minutes, on the proposal currently in front of you. Line two is the one that ends most discussions, and line six is the one that survives you.

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.