I worked with a mid-level data engineer who had come from a data science background.
He was technically strong. Smart, curious, good at getting things to work. But he didn't really understand why we modelled data. To him, data was something you queried, not something you designed. The distinction felt academic.
The shift came from an exercise rather than an explanation.
I sat with him and we worked through how an end user would actually try to answer a question with the data in front of them. What joins they'd need. Where they'd get confused. What they'd have to know in advance just to get a sensible number out.
I didn't tell him what good modelling was for. He arrived at it himself.
After that, he started thinking about data from the other end. Not how do I build this, but how does someone use this. It changed how he approached almost everything.
But there was a second problem, and it took longer to see.
He'd grown in understanding but not in confidence. He'd come to me with questions he already knew the answers to. Not because he was uncertain about the technical side, but because he hadn't yet learned to trust his own judgement enough to act on it without a second opinion.
I started pushing back differently. Instead of answering, I'd ask what he thought. When he gave me an answer, I'd tell him to go with it. When he was right, which was most of the time, I'd make sure he registered that.
The knowledge was already there.
What he needed was permission to rely on it.
It's one of the more important things I've learned about developing engineers. There are two different blockers. One is capability. The other is self-belief. They look similar from the outside but they need completely different responses.
Giving someone more context when the problem is confidence just makes them better informed and still hesitant.
He went on to become a senior data engineer at another company.
I don't think I taught him much he couldn't have figured out. What I think I did was create enough space, and enough repetition of the question what do you think, that he eventually stopped looking for someone else to validate the answer.
That's the job, most of the time. Not filling gaps in knowledge. Getting out of the way of someone who already has what they need.