The meeting had seventeen people in it.
The CHRO had called it to review the L&D function's quarterly metrics. Training hours: up 23% year-on-year. Completion rates: 91%. Manager satisfaction scores: 4.2 out of 5. The L&D head presented the numbers with the quiet pride of someone who had worked very hard to produce them and had every reason to believe they were good.
Nobody challenged them.
Nobody asked what 91% completion actually meant for the performance of the people who had completed the programs. Nobody mapped the training hours back to any movement in the business metrics they were supposed to influence. Nobody pointed out that manager satisfaction with learning programs is roughly as useful as asking passengers to rate the in-flight entertainment on a plane that has just had engine failure.
The meeting ran for ninety minutes. A follow-up was scheduled. Action items were assigned. Everything was fine.
Nothing was fine.
The fraud is not in the numbers.
The KPI problem in L&D isn't that the numbers are wrong. It's that the numbers are right. They measure exactly what they claim to measure. Completion rates measure completion. Hours measure hours. Satisfaction scores measure the satisfaction of the people who showed up long enough to fill out the survey.
The fraud is in the claim that these numbers mean something about performance.
They don't. And everyone in the room knows it.
The L&D head knows that a 91% completion rate tells her nothing about whether anyone walked out of that training and applied a single thing. The CHRO knows that the manager satisfaction score went up this quarter because the facilitator was warmer and the lunch was better. The business unit leads know that the leadership development program ran, everyone got their certificate, and their direct reports are still making the same decisions they made before.
They all know. They sit in the same room. And they review the same numbers.
Because the alternative is considerably more uncomfortable.
What the real measurement requires.
If you measure whether training actually changed behavior, you have to define what changed behavior looks like. You have to agree on the performance gap that was being addressed. You have to build a measurement framework before the training runs, not after. You have to explain to the CFO why the leadership program produced a 12% improvement in decision velocity in one business unit and no detectable change in three others.
That conversation is genuinely hard. It requires L&D to own outcomes they cannot fully control. It requires business leaders to share accountability for a function they typically treat as a vendor. It requires the CHRO to go to the board with a number that might be embarrassing rather than a number that looks clean.
The completion rate is not a measure of learning. It is an insurance policy. It exists so that when something goes wrong, the L&D function can say: we ran the program. We measured the hours. The scores were good.
The KPI is not a performance indicator. It is a deflection mechanism dressed in a spreadsheet.
Why the intellectual infrastructure isn't the constraint.
The organizations that are genuinely disrupting this are not the ones with better dashboards. They are the ones that decided, usually under significant internal pressure, to define success in terms of business outcomes and trace L&D's contribution backward from those outcomes.
This is not a new idea. The Kirkpatrick model has been around since 1959. Transfer climate research, performance consulting frameworks, learning impact measurement — the intellectual infrastructure for measuring actual outcomes has existed for decades. The tools are not the constraint.
The constraint is that measuring real outcomes is politically costly in ways that measuring completion is not.
When you show that a leadership program produced measurable improvement in escalation rates, three people take the credit. When you show that it produced no measurable improvement despite full attendance, twelve people assign the blame to each other. The incentive structure of most organizations makes the second scenario career-limiting for everyone involved. So everyone agrees, quietly and without explicit discussion, to measure the things that cannot be contested.
Training ran. People attended. They said they were satisfied. Nobody owns what happened next.
The mandate gap.
There is a particular version of this I have encountered more than once. A large L&D function with a significant budget, strong internal capability, and a genuinely talented team. Everything measured. Everything reported. The quarterly review is a model of operational rigor. And yet the business leaders, when asked privately, describe the function as disconnected from what they actually need. Not hostile. Not incompetent. Disconnected.
The L&D team is solving a different problem than the one the business has.
The business has a performance problem: leaders who cannot navigate ambiguity, managers who cannot develop their people, teams that execute well in stable conditions and fail visibly under pressure. The L&D function has a delivery problem: how to design, schedule, facilitate, and measure learning programs at scale. Both problems are real. Both require expertise. They are not the same problem.
When the KPI measures delivery, the function optimizes for delivery. The delivery metrics improve. The performance problem stays exactly where it was.
The gap between the two problems is not a measurement gap. It is a mandate gap. Most L&D functions were never asked to own performance. They were asked to own programs. The KPI reflects the mandate they were actually given, not the mandate that would make a difference.
What rebuilding actually requires.
Rebuilding this is not primarily a measurement exercise. You cannot add a Level 4 column to the existing dashboard and expect the incentive structure to change. The measurement follows the mandate, and the mandate follows a decision by someone — usually the CHRO — to accept accountability for outcomes that are harder to defend and impossible to fake.
That decision is rarer than it should be.
The organizations that make it tend to have one of two things in common: a new CHRO in the first hundred days who hasn't yet accepted the previous function's definition of success, or a business crisis large enough that the comfortable metrics have become impossible to defend. Neither is a comfortable entry point. Both are windows.
SSUNDAR's Performance Systems Architecture engagement begins with this conversation — not with a learning audit, not with a content review. With the question of what success was actually supposed to look like, and why the current definition was chosen instead.
The KPI nobody wanted to own is usually the one that tells the truth.