How to measure the impact of training?

You've probably heard the stat. "Only 10% of training actually transfers to the job." It comes up a lot in L&D circles.

The trouble is, nobody can trace it back to a source.

A researcher called Fitzpatrick went looking. He traced it back to a 1982 article where the author admitted the number came from chatting to a few training directors, not from any study. That estimate got quoted, then quoted again, until it became treated as fact. There's even a name for it now, coined by researchers in 2011: the "10% delusion".

What's the real number? Most actual research puts transfer somewhere between 10% and 30%, better than the myth but still not a figure anyone running a training budget wants to hear.

Why training was never supposed to carry this alone

The 70:20:10 split is worth keeping in mind here. Roughly 70% of how people develop professionally comes from doing the job. 20% comes from other people, a good manager, a peer, someone showing you the ropes. Only 10% comes from formal training itself.

If that's even roughly right, expecting one workshop to change behaviour on its own was always asking too much. Training plants seeds, but what happens in the weeks afterwards decides whether they grows.

The research backs this up: the single biggest factor in whether training "sticks" isn't how good the session was, but what happens back at work afterwards, whether a manager reinforces it, whether there's a chance to use it, whether the session left action points or practical processes behind, whether anyone follows up at all.

How to measure it?

A few frameworks have tried to solve this:

Kirkpatrick's four levels is the one most people have heard of: reaction, learning, behaviour and results, or whether people enjoyed the training, whether they learned it, whether they used it, and whether it changed anything. It's been criticised for assuming each level causes the next, which isn't always true.

CIRO flips the order, looking at context, input, reaction and outcome. It checks what's needed and what resources exist before training happens, not just how it went afterwards, though it tends to be stronger on the planning side than on tracking what people do differently once they're back at their desk.

Phillips ROI bolts a fifth level onto Kirkpatrick, turning the result into a number: what this actually saved or earned against what it cost. It's powerful when you can do it well, but difficult in practice, because isolating training's effect from everything else going on (a new manager, a better year, market conditions) is hard to do with confidence.

The Success Case Method is the cheap and fast alternative. Instead of trying to measure everyone, you find your best example and your worst example of someone applying the training, then dig into why. One good story, properly investigated, often tells you more than a stack of survey scores.

The honest bit nobody likes to admit

Even with the best framework in the world, you can't fully prove training caused a result because there are just too many factors at play. Real certainty would need a control group, and nobody's running one of those for a leadership course.

There's a reason for the old line about lies, damn lies and statistics. Numbers can almost always be arranged to support whatever point someone's trying to make, even with good intentions behind it. A training ROI figure is no exception.

That doesn't mean giving up on measuring, it means being honest about what you're claiming.

The reality right now

Survey data from the CIPD paints a pretty clear picture of where most organisations actually are. Only around half have any process at all for assessing learning impact. The number of senior leaders who think L&D is making a difference to organisational priorities has been falling consistently.

Most teams are still asking "how was the workshop" and calling that evaluation, when a feedback form often only measures reaction and tells you almost nothing about behaviour.

What does work

Strip away the jargon and a few things consistently come up across every serious framework:

• Decide what you're trying to change before you design the training

• Capture where people started, so you have something to compare against

• Check in weeks later as well as on the day

• A specific story about someone doing something differently is real evidence in its own right

• Numbers and stories work best side by side

What we do

At Teamshaper, we built our process around treating measurement as something that starts before the workshop and continues weeks after it, not a single form at the end.

We agree what success looks like in one line, gathered from the commissioning team and from a staff survey we run with participants before the workshop. We display that line at the start of the session itself, so everyone in the room can see their own input shaped the day. That builds ownership and sets the tone and expectations from the first minute.

During the programme, people make specific commitments about what they'll do differently. We follow those up, sometimes with a session a month later, and we always give each participant's line manager the full list of commitments so they know what to look out for, support or reinforce.

A follow-up session, run by us or handled internally, creates a natural check-in point. It's where people raise what they tried, what worked and where they got stuck, giving everyone a chance to troubleshoot together while the work is still fresh.

We also measure a benchmark before and after the programme. The measures are chosen carefully and are subjective by nature, but the more specific they are, the more meaningful the change between them becomes. We typically track three: clarity, effectiveness and confidence, in whatever form fits the programme.

Everything comes together in an Impact and Value Report: the numbers, backed by trainer observations, specific success stories and recommended next steps. It's not a perfect science, but it means we can show our clients more than a happy sheet or a fancy meaningless infographic.

Example of a case-study that demonstrates those measures can be found here.