We have no shortage of ways to measure learning. Unfortunately, almost none tell us what learning’s value is.
We’ve spent sixty years building models to evaluate training. We’ve all heard of Kirkpatrick, Phillips, LTEM, and then a dozen more with names that are less familiar. You’d think that would be enough to answer a simple question such as “was the learning program worth it?”
In our last blog, we wrote about why it usually isn’t — you can show people learned something, but not what the learning was worth to the business. That’s because existing models were mostly built to answer a different question.
The usual suspects
Kirkpatrick is the one almost everyone knows, and often the only one they know. It sorts evaluation into four levels: whether people liked the training, whether they learned anything, whether they changed how they work, and whether the business saw a result. It’s a reasonable way to think, and it’s been the default for decades.
Phillips built on it. The Phillips ROI Methodology takes that fourth level — business results — and carries it into money: isolate the effect of the training, convert the result to a dollar figure, set it against what the program cost. Of the well-known models, it’s the only one made to answer the question a finance leader would ask. It’s also slow, costly, and demanding to run, which is why hardly anyone does. Fewer than one in ten organisations report using it regularly — and even fewer run it all the way to a dollar figure.
LTEM, Will Thalheimer’s Learning-Transfer Evaluation Model, was built to replace those four levels. It looks harder at what Kirkpatrick skates over — whether people can make the decisions and do the work the training was for, and whether any of it lasts once they start to forget. For measuring whether learning worked, it’s a real step up.
All of these are careful, sometimes brilliant answers to one question: how do you measure learning well? Thalheimer has spent years arguing, rightly, that most of us measure it badly — still counting heads, still handing out end-of-course surveys that tell us next to nothing. But measuring learning and measuring its value to the business are different jobs. Phillips is the only one that tries to put a value on it, and most teams have decided it’s more trouble than it’s worth.
Measuring learning and measuring its value to the business are different jobs.
Every one of these models has the same gap we started with: each was built to tell you whether the learning worked, not what it was worth to the business in dollars. And the frustration is old — Kirkpatrick himself complained in 1960 that measuring training by its results was moving slowly. Sixty years on, the research says most of us still want to measure better, and few of us do.
Every so often someone proposes a different yardstick — capability, or how fast people pick up new skills. There’s something to those, but they’ve never really caught on, and they measure something other than what the business is asking about.
Whose result is it anyway?
And even when you measure exactly what the business is asking about, you hit a wall: that result rarely has a single cause. Sales climb, turnover drops, errors fall, and any of a dozen things could be the reason behind it. The market moved, prices changed, a new product shipped, hiring got better — and the training may have helped too … somewhere in there.
Working out how much of the result belongs to learning is hard. Done properly, it calls for control groups and comparisons — the kind of clean before-and-after most workplaces can’t manage. Thalheimer’s research says the same: credible attribution takes methods most teams have no time or budget for.
So L&D gets caught between two bad options. Take credit for a result you can’t fully prove, and a finance leader will pick the claim apart. Say nothing, and the budget goes to the teams that made their case. We think there’s a better way.
A different way to show value
What we’d change is the thing you’re trying to prove in the first place.
Rather than claiming the training caused the result, you lay two things side by side: what the program cost — design, delivery, people’s time — and the business outcome you set out to move, with whatever happened to it. You don’t draw a line between them. Say a manager program cost $200k to run, and over that same year, first-year turnover — the thing it was built to reduce — fell. You set the cost beside the drop in turnover and let the people who own the budget decide what it was worth.
We call this investment efficiency.
It makes no causal claim, so there’s nothing to dispute — just cost against outcome, in terms finance already uses.
It only works if you start before the program does. You pick the business outcome you want to move, find the person who owns that number, and agree with them up front that you’ll track it together. Get that in place and measuring later is straightforward, because you already know what you were aiming at. It also means the evidence won’t sit in your LMS — it sits with the stakeholder who owns the result.
Working this way takes more thought at the start than most teams are used to, and far less scrambling at the end. We’ll get into how that framework works — and what it looks like on a real program — in the next few posts.
