Implementation and maintenance

The decision was right. The Monday stayed the same.

Where well-chosen solutions go to die

Everything before this number ends in a decision: a cause established, a remedy chosen against it, a measure agreed that will count as proof. Between that decision and a changed result lies a phase that most organizations plan as an email, and it is where more analyses are lost than to any error in the analysis itself.

The pattern is recognizable at one glance. The diagnosis was sound, the fix was right, the budget was approved, and six months later the measure sits where it was. The new form exists and the old one is still in circulation. The view was built and nobody's morning includes opening it. The rota was changed on paper and the evening shift never heard. Nobody decided against the solution. It was simply never installed in the work, because installing it was nobody's job with nobody's deadline.

The method itself names this phase. In the performance-improvement tradition the cycle does not end at intervention selection: after the analysis and the selection comes implementation and maintenance, a phase with its own design work, its own owner and its own failure modes (Van Tiem et al., 2012). Number 10 chooses the remedy; number 11 says what may be claimed afterwards. This number is everything in between, and it connects that stretch backward to the causes the analysis established, sideways to the solution that was chosen, and forward to the measurement that will judge it. None of it starts from a blank page.

What the tradition settled, half a century ago

Implementation is not a modern addition to the method. The discipline settled its two ground rules early, and both run against how organizations naturally behave.

First: implementation is designed with the intervention, not after it. The tradition treats the rollout as part of the solution's design (who will do what differently, what they need in hand on day one, and what happens to the old way of working) specified before anything is bought or built (Van Tiem et al., 2012). A solution whose implementation cannot be described is not yet a solution; it is an intention with a budget.

Second: maintenance is designed up front, because the pull of the old way outlives the launch. Rummler's version of the point is structural: performance is produced by a system, and a system that produced the old behavior for years does not stop doing so because a program ran (whatever was misaligned before the fix will quietly reassert itself unless something keeps the new arrangement in place (Rummler, 2004)). The tradition therefore asks, at design time: who notices drift, on which signal, and with the authority to act on it? A fix without a maintenance answer is a fix on loan.

Both rules have the same consequence for practice. The question how will this be implemented? belongs in the same conversation as which solution do we choose? Not a quarter later, in a project kickoff, when the design decisions that determine the answer have already been made by default. What the tradition could not yet supply was a size: how much does this phase actually matter? That answer arrived from a different research field, and it is the next section.

The size of the phase, measured

How much does implementation matter? The one field that measures it for a living has put a number on it, and the number is larger than almost anything a solution designer can buy.

The landmark synthesis reviewed the results of 542 interventions: 483 of them gathered in five meta-analyses, and 59 further studies that measured the influence of implementation on outcomes directly. Its conclusion carries the whole weight of this number: the level of implementation decides the outcomes obtained, and mean effect sizes run at least two to three times higher when programs are carefully implemented and free from serious implementation problems (Durlak & DuPre, 2008).

Read that as a purchasing decision. An organization that agonizes for a quarter over which remedy to buy, and then delegates the rollout to a newsletter, has optimized the smaller variable. The same solution, at the same price, returns double or triple depending on the phase this paper is about. On the evidence, implementation is the largest effect modifier left after the solution is chosen, which is why the tradition was right to make it a phase of the method rather than an afterthought of the project.

Five syntheses, and what each one settles

Note what the five have in common: they route authority back to the analysis. Modern implementation research does not replace the diagnosis with a rollout playbook: it says, from its own data, that no playbook survives contact with a setting, and that the only durable guide is what this organization established about its own causes. The method did not need rescuing. It needed confirming, and it got it.

One chain, four connections

Implementation is not a project that starts when the analysis ends. It is the same chain, entered further along, and every working connection runs on something the analysis already produced.

Four working parts no rollout survives without

Strip every implementation that held from our own practice and four parts remain: our formulation, from twenty years of watching fixes live and die, and each one is checkable in the first week.

First whether it arrived, then whether it worked

The number that will judge the solution is not the number that steers the rollout. A rollout is measured on two ledgers, and the order in which they are read decides whether the verdict means anything.

The team as implementer, and where that authority ends

The most contemporary change in this phase is not a technique. It is who does the installing. The older picture had somebody above the work rolling change downward. Much of the time, that picture is now wrong, and the method is better for it.

Teams rearrange a growing share of their own conditions: the order of a handover, what a shift starts with, which step quietly disappears because it never helped, what the daily review looks at. Where the chosen fix lives inside that span, the fastest implementation is the team’s own: the four working parts still apply, but the owner is the team, the Monday is next Monday, and the feedback loop is theirs by construction. The research on work redesign points the same way: interventions designed with the people who do the work carry the more promising record (Knight & Parker, 2021), and a team that can alter its own conditions acts on a finding in a week rather than waiting a quarter for permission.

The launch fights for attention. Maintenance fights gravity.

Everything up to here gets a fix into the work. Keeping it there is a different problem, and the research that followed programs past their launch says the problem is usually lost: quietly, and without a decision.

The broadest review of what happens after implementation, 125 studies that followed programs into their later years, found partial sustainment more common than continuation of the whole program, even where the initial implementation had fully succeeded; in the studies that assessed fidelity rigorously, fewer than half of the providers still ran the practice at a high level (Wiltsey Stirman et al., 2012). That is a direction, not a fate: the programs studied mostly had no designed maintenance at all, which is precisely the tradition's complaint from section 02. What the finding removes is the comfortable default: the assumption that an installed fix stays installed because nobody decided to remove it. Nobody decides. The old system's pull does not need a decision.

The design answer has four parts, and all four are cheap at design time and expensive after. A drift signal: one named person who notices the old way creeping back, on a defined signal, with the authority to act, not a value in a culture deck. The loop left running: the feedback loop from section 05 drops to a lower rhythm after the first quarter, but it does not stop, because a number nobody sees is a number that drifts. The removal held: a workaround cannot return if its reason is gone, which is why the removal in section 05 is maintenance disguised as implementation. Ownership that survives the project: maintenance transferred by name to the line before the project closes, or it is transferred to nobody.

One fix, two rollouts

A field-service organization, two regions, one finding: engineers were closing jobs without the customer-approval step, and the rework was traced to exactly those jobs. The chosen fix was the same in both regions: the approval step built into the closing screen. The figures are illustrative; the pattern is from our practice.

What this does not claim

Five limits, and the first one protects everything that came before this number.

• Perfect implementation of the wrong solution is still the wrong solution. Nothing in this number substitutes for the analysis. A remedy that cannot reach the established cause fails identically well however cleanly it is rolled out: number 10 owns that argument.
• The figures belong to the sectors they were measured in. The syntheses cited here come largely from healthcare and prevention programs. Their figures are quoted for those sectors; what this paper carries into yours is mechanism and direction, nothing more.
• Modest means modest. The best-evidenced strategy on the list moves practice by single-digit percentage points on average. Implementation multiplies a real effect; it cannot create one: the two or three times of section 03 multiplies whatever the solution is worth, including nothing.
• The four working parts are our formulation. Practice, distilled, not a validated instrument. They are offered as a checklist a stranger can verify, which is what a practice heuristic is for.
• The sustainment evidence is the weakest link, and says so itself. The sustainability research is largely retrospective, with unsettled definitions: its own authors' conclusion (Wiltsey Stirman et al., 2012). The four-part maintenance answer in section 08 is design logic and practice, not a proven protocol.

Six lines, before the rollout starts

Thirty minutes, on the solution you are about to install. A blank on lines one to four is the finding, and cheaper now than in two quarters.

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

Where every claim comes from · continued

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.