The trees, not the wood
We've all been there at the start of a project or in the midst of a client workshop. You start thinking about how you'll know whether a learning intervention has worked once it’s out there in the wild. You talk about success measures, and metrics, and surveys, and data capture… and then before you know it, your measurement criteria are longer than a kid's Argos catalogue Christmas list.
Those of us with years of learning design experience might naturally start to lean heavily into the Kirkpatrick model.
And in an ideal world, what we really want to measure are the things the organisation cares deeply about: fewer incidents, better wellbeing, increased sales, lower risk. All the important stuff.
How do we know what worked?
But then come the inevitable questions
How can we be sure it was our intervention that caused that result? If sales rose, perhaps it helped. Or, perhaps pricing changed, marketing ran a campaign or demand increased.
And if your goal is prevention, it's even harder. Serious incidents, breaches and failures are all, hopefully, rare, meaning sometimes success in fact looks like nothing is happening at all.
So how do we keep our success measures focused, meaningful and close enough to our intervention to understand its contribution?
To try and answer that conundrum, here are four tips that draw on my own work and conversations with specialists in behavioural risk and economic evaluation.
1. Go back a step
I had a wonderfully clarifying conversation about ‘taking a step back’ with behavioural risk specialist Dr Wieke Scholten at BR Insights. I was particularly curious to explore Dr Scholten's world where the aim is often to prevent serious incidents. These thankfully tend to be rare events, so it’s really important to look at the conditions which make them more, or less, likely. Her advice was reassuringly practical.
“Don't go too big. Look at earlier indicators of change.”
For example, if the final outcome is difficult to attribute, look at what should have changed on the way there. Take a look at the behaviours that need to change, and the drivers of those behaviours.
Imagine for a moment that your business goal is increased sales. Now go back a step: what behaviour contributes to that? Perhaps salespeople asking better diagnostic questions.
Go back another step: what's preventing that behaviour? Perhaps they feel pressure to rush calls.
Go back one final step: what can your intervention realistically change?
Perhaps you want to see if you can reduce the feeling of pressure via a comms campaign that highlights how the very top salespeople manage to do it successfully. And then after that, take a look at how that comms campaign was received and whether it changed perceptions.
Dr Scholten's behavioural risk approach offers a useful pathway from driving factors to behavioural patterns to outcomes. Those driving factors might, for example, include knowledge, confidence, social norms or incentives.
The trick is to measure at different points along that pathway, particularly the factors and behaviours your intervention was designed to change. These can act as earlier signals that are far easier to pick up than the eventual final outcome.
2. Pick a model
Once you know what you're trying to change, the second tip is to not try and invent some brand new measures from scratch. It’s far better, and a more productive use of your time, to find out what good research can already tell you.
On a recent project with a global pharmaceutical company, we wanted to create a learning experience that helped employees and managers better understand what psychological safety can mean at work. To do this, we embedded survey questions into the actual experience itself, creating an opportunity for us to listen while people were already finding out about the topic. But we didn't simply invent the survey questions. We drew on Amy Edmondson's research on psychological safety to shape what we asked.
Furthermore, using established research helps make sure you're measuring the right things, rather than measuring whatever is easiest or most obvious. Dr. Scholten introduced me to a lovely example called HROX (Haslam et al., 2026): a recently validated measure of high reliability in organisations. Rather than waiting for a catastrophic failure to judge reliability, it measures practices such as learning from failures, anticipating problems and allowing expertise to influence decisions. These offer earlier signals of whether an organisation is becoming more reliable.
So in short, don't start from scratch. Start with "what do we already know about how this should be measured?".
3. Measure and communicate what people really care about
Sometimes, we can spend so much time designing a technically impressive measurement framework that we forget something important:
Who will actually use this information?
Insightful conversations that I’ve been lucky enough to have with Dr. Sima Rafiei, Health and Public Policy Researcher at the University of Warwick, have encouraged me to think beyond traditional return on investment (ROI). For example, cost-consequence analysis presents the costs alongside different outcomes, rather than squeezing everything into one financial figure. This can give decision makers a more useful picture of value.
But what matters more is who is looking actually looking at this picture.
A board might want to understand overall financial savings and risk reduction. A construction site manager might ask something much more immediate, like "if I get people raising concerns earlier, what does that save on my project?".
The underlying evidence might be the same, but the way we communicate its value should change. So before choosing a measure, it’s important to ask yourself: who needs this information? And what decision will it help them make?
4. Measure behaviour change as carefully as you can
Let's zoom in on one part of the pathway we explored in Tip 1: behaviour. As a Behavioural Strategist, I'm obsessed with asking: "What would somebody actually do differently if this worked?". Answering this means getting specific information about the behaviours you'd expect to see in the real world.
Not only does this give you a clear target for your intervention - it also gives you something more tangible to measure. So, at the end of the intervention, ask people what they intend to do differently, and then follow up with… did they actually do it?
You might need a mix of evidence. Surveys, observations, manager interviews and operational data all might tell us something different. And if intention increases but behaviour doesn't, that's useful information too: e.g. perhaps knowledge wasn't the barrier - perhaps it was workload, incentives, social norms or manager behaviour.
The goal isn't to measure everything.
Please don't measure everything!
Your goal is to gather enough evidence to test the story behind your intervention: did you change the factors driving behaviour, did the behaviour change, and did that contribute to the outcome?
In short, start with the outcome… but measure the pathway.
