Back in 2018 I wrote a blog called Shopkeeper to Engineer. I’d been using the phrase for a few years before that, trying to describe the difference between supplying learning and understanding how performance works.

My Dad was a shopkeeper, so it wasn’t intended as an insult. He ran a local shop, supplied what people needed and changed what he stocked as the market changed. He was good at what he did. But the shopkeeper responds to demand for products, while the engineer needs to understand how the whole system works and how its different parts contribute to the outcome.

Eight years later, I was talking about the same thing at the CIPD Annual Conference, on a panel with Laura Overton and Vicky Keith, discussing L&D’s role in a performance-led skills strategy.

We started with a show of hands. Who’d been asked to organise training to solve a problem at work? Lots of hands went up. Keep your hand up if you’d had the chance to explore the problem with the person asking. Then if you’d been able to investigate what was causing it. Finally, keep your hand up if you’d had a say in what to do about it.

Hands came down along the way.

We often talk about L&D needing to ask better questions, diagnose problems and become more commercially aware. All sensible things, but the exercise showed me you can have the skills to diagnose a problem without having the authority to do anything about it.

I shared an example of an organisation wanting sickness absence training. Managers weren’t completing the necessary processes, so HR assumed they needed training. We investigated and found managers weren’t using the reporting system properly. A six-frame GIF showing people how to use it went out to around 1,000 people. Reported sickness absence went through the roof.

That sounds like a terrible result until you understand that sickness absence hadn’t suddenly increased. Reporting had improved. The organisation could now see the extent of a problem that had previously been hidden by inaccurate data. We hadn’t solved the sickness absence problem, but we’d established what was actually happening. The requested training wouldn’t have done that.

We also had an interesting disagreement about training catalogues. I suggested that having a catalogue made an organisation supply-led rather than demand-led. Laura challenged that, pointing out that existing learning resources can be valuable when they’re used alongside other support to help people apply what they’ve learned.

And she’s right. Having a catalogue doesn’t necessarily make you supply-led. It’s what determines the use of that catalogue that matters.

If the starting point is what we’ve already got available, we’re looking for ways to use our stock. If the starting point is what people need to do differently, we can decide whether any of that stock will help. There’s nothing wrong with using existing resources. There’s plenty wrong with using them simply because they exist.

It’s a clear distinction, and one I didn’t make clearly enough on the day.

We also talked about measurement. I kept asking, ‘so what?’ Someone completes a programme, watches a video or downloads a resource. So what? Ten days later, what are they doing differently?

When I worked on the lateral flow testing training, we had detailed data showing how people used instructional videos and could examine the relationship with void test results. It was possible to investigate whether engagement with the content was associated with the performance we needed, rather than simply count how many people had watched it.

Towards the end, we discussed how L&D gets involved in the important conversations. Laura suggested not waiting to be invited. Vicky talked about persistently getting involved in technology projects before someone suddenly decides training is needed.

My suggestion is, as always, three words. Help me understand.

I’ve used those words for years. They’re a way of opening up a conversation without immediately challenging someone’s judgement. Help me understand what you’re trying to achieve, what you’re seeing and why you’ve decided this is the answer.

But I’ve been thinking about those hands coming down at the beginning of the session.

In my 2018 blog, I suggested that organisations were rarely ready for an engineering approach to L&D. I also talked about the convenience of the existing arrangement. The business asks for learning, L&D supplies it, and everyone knows what to expect. Change that arrangement, and you potentially make life more difficult for both sides.

Eight years later, I’m not convinced that much has changed.

We have more tools, better access to evidence and increasingly sophisticated technology to help with diagnosis and performance support. The CIPD has published practical toolkits, including the skills-first resources I’ve developed, to help practitioners work through these issues.

But having the tools and being able to use them are different things.

At the end of the session I asked how many people were recording it and might use AI afterwards to help them work out how to apply the discussion. Very few hands went up. We’d spent half an hour talking about performance, but listening to us wasn’t going to change anyone’s performance by itself.

And that applies to L&D as much as anyone else.

If someone isn’t performing as expected, we should investigate whether they have the skills, the opportunity, the resources and the authority to do what’s required. We shouldn’t automatically assume they need training.

So when an L&D function keeps behaving like a shopkeeper, why do we keep assuming the solution is to develop its people into engineers?

Perhaps we should be asking what the organisation is actually commissioning them to do, how their work is measured and who gets to decide what happens when they identify a problem that training won’t fix.

I was arguing for a whole-system approach in 2018. Eight years later, I think we need to be much better at applying it to L&D itself.