Ninety-four percent completion. Feedback scores averaging 4.6 out of 5. Post-program assessment scores averaging eighty-one percent on the knowledge check. The L&D team had executed well. The managers who had gone through the cohort genuinely said they had found it valuable. Several had called it the best development experience they had received since joining the organization.
The program had been designed to address a specific behavioral gap: managers in this particular business unit were not having the developmental conversations their direct reports needed. The skip-level data showed it. The engagement scores showed it. Several high-potential exits in the previous year cited it explicitly. The CLO had made it a priority. The design work had been thorough.
Six months after the cohort closed, nothing had changed. The same managers. The same conversations not happening. The same direct reports noting in pulse surveys that they were not receiving the developmental guidance they needed to grow. The L&D team was baffled and, privately, defensive. They had delivered well. The numbers said so.
The numbers were correct. That was almost exactly the problem.
What the Program Was Measuring.
The program had been designed to be excellent at the things it measured. Completion was high because the program had been designed to be engaging and the organization had made attendance mandatory. Feedback scores were high because the program was genuinely good — the content was relevant, the facilitation was skilled, the peer connections formed in the cohort were real. Knowledge assessment scores were high because the learning design was rigorous and the assessments measured what had been taught.
None of those metrics measured behavioral change in the managers' work environment. They measured whether the program had transferred knowledge and generated positive experience. The question they did not ask — could not ask, because it was outside the scope they were designed for — was whether the environment those managers returned to would support or suppress the behaviors the program had taught.
Learning measurement in most organizations stops at the edge of the learning experience. What happens inside the program is measured carefully. What happens when participants return to their roles is measured rarely, with long delays, with instruments that capture self-reported intention rather than observed behavior, and with little accountability for what the numbers show.
The result is a measurement system that tells the L&D function whether it delivered a good program. It does not tell the organization whether the program produced the change it was commissioned to produce. The gap between those two questions is where most of the L&D budget disappears.
The Three Structural Suppressors.
In the case of this cohort, the environment those managers returned to was suppressing the target behavior through three distinct structural conditions. Any one of them, operating alone, would have been sufficient to prevent sustained behavioral change. All three operating simultaneously made the outcome close to predictable from the start of the design process.
The first suppressor was the absence of the behavior in the management layer above the cohort. The managers who attended the program returned to line managers who had not attended the program and who, largely, had not been having developmental conversations themselves throughout their careers. In organizational settings, behavior at a given level is strongly shaped by the behavior that is modeled at the level above. The program asked these managers to develop a behavior that was not being modeled by their own managers. The behavior was therefore signaled as optional — something the L&D function was asking for, not something the leadership of the organization practiced.
The second suppressor was measurement invisibility. The developmental conversation behavior the program was trying to install was not captured in the performance review framework. There was no measure of whether managers were having these conversations. There was no measure of their quality. There was no measure of the developmental progress of direct reports as a function of manager coaching. When a behavior is not measured, the organization is communicating that the behavior is not important. The program asked for the behavior. The performance system said nothing about it. People are very good at reading what an organization actually values, and they read it primarily through what the organization measures and what it rewards.
The third suppressor was structural time scarcity. These managers were operating in a business unit that had been understaffed by approximately twenty percent for the previous two quarters. Their calendars reflected this reality. The program had taught developmental conversation techniques that, executed properly, required between thirty and sixty minutes of focused, private conversation per direct report per month. The structural math was not favorable. The managers' existing workload did not contain that time. The program had provided skills. It had not provided capacity. And the organization had not created the capacity by addressing the staffing gap, by restructuring expectations, or by removing lower-priority demands on the managers' schedules.
Skill without capacity is not a skill problem. It is a design problem that the skill training cannot solve.
Why L&D Takes Responsibility for Architecture Failures.
The uncomfortable pattern in this situation is that the L&D team took responsibility for a failure that was not theirs. After the six-month review showed no behavioral movement, the team commissioned a needs analysis, reviewed the content design, considered whether the cohort composition had been wrong, and discussed redesigning the follow-up reinforcement activities. All of this was earnest and professional. None of it was the correct diagnosis.
The failure was upstream of the program. It was located in the three structural conditions described above: the absence of modeling at the management level, the invisibility of the target behavior in performance measurement, and the structural time scarcity created by understaffing. None of those conditions could be changed by a better program design, a different content sequence, or a more rigorous follow-up protocol. They could only be changed by decisions that were outside the L&D team's authority and scope.
This is the structural trap that many L&D functions sit in. They are held accountable for behavioral outcomes that depend on conditions they do not control. When those outcomes fail to materialise, the diagnosis defaults to the program — which is visible, designed by the L&D team, and available for scrutiny — rather than to the structural conditions, which are dispersed across management practice, measurement architecture, and resourcing decisions made at levels above the L&D function.
The CLO who wants to break this pattern has to be willing to say something that most L&D functions are not structurally positioned to say: this program cannot work until these three conditions are changed. Not "we will design around the conditions." Not "we will make the program stronger." The program is not the lever. The conditions are the lever, and they are not ours to pull. They belong to line leadership, HR architecture, and operations planning. Until they change, the program outcome is predictable and the prediction is not positive.
What the Design Conversation Should Have Contained.
A design process that read the structural conditions correctly would have produced a different scope of work before any program content was discussed.
It would have included a conversation with senior leadership about what modeling at the management level needed to look like, and a commitment — with names and accountability — about who would change that behavior and by when. Not a conversation about the importance of developmental conversations. A specific conversation about which managers at which levels would demonstrably adopt the behavior before the cohort launched, so that participants returned to an environment where the behavior was already being modeled above them.
It would have included a redesign of the performance review framework to include a measure of developmental conversation quality and outcome. Not a soft question on a survey. A measure with weight, visible in the performance cycle, attached to the outcomes of direct reports. A measure that made the target behavior legible to the system.
It would have included a direct question to operations about the capacity constraint: are we designing a program that asks for behaviors the current schedule cannot accommodate? And if the answer was yes — which it clearly was — a conversation about whether to address the staffing gap, reduce other demands on manager time, or explicitly include schedule restructuring as a deliverable of the program rather than an assumption beneath it.
That scope of work is significantly larger than a learning program. It implicates senior leadership behavior, HR measurement architecture, and operational resource allocation. Most L&D teams do not have the mandate, the authority, or the political capital to commission that scope of work. The ones that do are the ones whose organizations are actually changing.
What It Costs to Outsource the Architecture Question.
The cohort described at the start of this article cost approximately 1.4 million rupees to design and deliver. That figure does not include the management time committed to the program over the cohort period. It does not include the opportunity cost of the L&D team's attention. It does not include the follow-up analysis and redesign work that happened after the six-month review showed no movement.
The three structural conditions that suppressed the behavioral change — modeling absence, measurement invisibility, structural time scarcity — could have been identified in a one-week diagnostic. The diagnostic would have produced a clear picture of what needed to change before a program was designed. The conversation with senior leadership about those conditions would have been uncomfortable. The outcome — either structural change or an honest acknowledgment that the program would not produce the target behavior without it — would have been more useful than the outcome that actually occurred.
The classroom was full. The learning was real. The organization got a report showing 94% completion and 4.6 satisfaction. Six months later, the behavior the program was designed to produce was absent, and the structural conditions that had made it absent remained entirely in place.
That is not a learning design failure. It is a system that successfully prevented learning from becoming behavior. Understanding the difference is what changes what L&D functions ask for before they agree to design anything.