The Wrong Instrument
When someone asks whether a learning program "worked," what they usually mean is: did it change something that matters? That's a fair question. Organizations invest real money in learning and development, and the people who approve those investments have a right to know whether the money was well spent.
Where things go wrong is not in the asking. It's in the instrument we reach for to answer.
For decades, the dominant answer to learning measurement has been ROI — return on investment. Calculate what you spent, isolate the business results attributable to training, express the difference as a percentage. In the most formal versions of this approach, a positive ROI becomes the gold standard for program success.
The problem isn't that this is unreasonable. The problem is that it was never designed to do what we're asking it to do.
ROI is a capital allocation tool. It was built to help organizations compare investment options — whether to buy a piece of equipment, open a new market, acquire a company. In those contexts, it works well. The inputs are defined, the outputs are measurable, and the causal chain between investment and return is relatively short and traceable.
Learning is different on every one of those dimensions.
The inputs — design quality, facilitation, manager reinforcement, learner readiness, organizational context — are variable and hard to control. The outputs — behavior change, application on the job, sustained performance improvement — unfold over time and are influenced by dozens of factors beyond the program itself. The causal chain is long, branching, and fragile.
When you force that kind of complexity through an ROI calculation, you don't get a clear answer. You get a number that someone constructed, often under pressure, from assumptions that are difficult to defend. The calculation looks rigorous. It isn't.
This is a design problem, not a measurement problem.
L&D didn't adopt ROI because it was the best available instrument for evaluating learning. It adopted ROI because it was the language finance spoke, and speaking finance seemed like the path to credibility. The intent was right — demonstrate that learning has organizational value. The design was borrowed from a different domain without checking whether it fit.
The consequence is a field that spends enormous energy on measurement theater: isolating training effects through control groups that don't exist in corporate settings, monetizing outcomes that resist monetization, and producing ROI figures that experienced stakeholders don't fully trust but haven't found a compelling reason to challenge.
Meanwhile, the questions that would actually improve learning design — Are people using what they learned? Are they using it correctly? Is it changing outcomes at the team or unit level? — often go unasked, because they don't fit the model.
Will Thalheimer's Learning-Transfer Evaluation Model offers a different way to think about this. LTEM organizes evaluation across eight tiers, from basic attendance and satisfaction through to long-term performance improvement and organizational impact. It's notable for two things in particular.
First, it's explicit about what most current evaluation practices actually measure — and how far that is from what most of us claim to care about. Satisfaction surveys, knowledge checks, and even behavior observation all appear in the model, and each is located at its appropriate level of rigor. Most organizations, if they're honest with themselves, operate in the lower tiers most of the time.
Second, LTEM makes the measurement ladder visible. It doesn't prescribe that every program needs to reach the top tier. It asks practitioners to be clear about what they're measuring and what that measurement does and doesn't tell them.
That's a more honest conversation than ROI allows. ROI implies a definitive verdict. LTEM invites a more accurate one.
None of this is an argument that learning shouldn't connect to organizational performance. It should, and the expectation that it does is legitimate. An L&D function that can't articulate how its work serves the business has a real problem.
But "connect to performance" and "produce a defensible ROI figure" are not the same thing, and conflating them has cost the field more than it has given. The organizations that evaluate learning most rigorously aren't necessarily the ones calculating the highest ROI — they're the ones asking the most precise questions about what changed, for whom, and under what conditions.
That precision is available to anyone. It doesn't require a finance-friendly number. It requires a better instrument.
The Gray Henley Practice Notes are periodic essays on learning, performance, and talent practice. If you're working through a measurement challenge in your own organization, get in touch.