We have grown oddly comfortable managing learning by its least meaningful number. A 97 percent completion figure reads beautifully in a board deck and reveals almost nothing: whether the work got faster, whether mistakes became rarer, or whether the person can perform the next time the situation appears. No finance leader would manage capital by counting the meetings held about it. It is time for the skill equivalent of closing the books — attaching a real number to what happens after the course ends, rather than before it begins.
The first shift is from volume to transfer. Instead of asking 'what did the workforce consume?', ask 'what did they reuse within thirty days?' In our partner programmes we measure three things: how much of the trained capability shows up in the first weeks of the job; how long a new hire takes to reach a defined performance target; and how often the behaviour survives a change of task. Calibrating that transfer against manager observation closes the gap far better than a single test at the end of a module.
The third metric is adoption over the medium term. A completion metric says everyone was taught; an adoption metric asks how many actually moved. We score short, real-role scenarios — the daily judgement call, the mid-course pivot, the governance edge — and adoption almost always trails completion by 15 to 20 points. That gap is precisely where the value of learning actually lives, and it is eminently teachable once it is made visible.
None of this means coverage no longer matters; of course the population must be reached. It means adding reach to questions of quality. Reach times transfer times judgement — a scorecard built monthly, on the business, and not on the portal. When a chief executive asks what the learning spend returns, the wrong answer is '97 percent completion.' The right answer is a measurable improvement in time-to-value, a drop in escalation volume, and judgement calibration holding. Only when the answer reaches the P&L can learning be treated as an investment it demands.