Mentoring Made Me a Better Analyst

For a few years now I have mentored data science students, through the University of Melbourne's peer and industry programmes and a virtual analytics challenge for students at another university. I went in thinking of it as giving something back. I came out having quietly improved at my actual job, which surprised me, and the reason it worked is worth writing down.

Explaining forces you to understand

The first thing mentoring does is expose the gap between knowing something and being able to explain it. You can carry a technique around in your head for years, using it correctly, and still not really understand it, in the sense that you cannot rebuild it from scratch for someone who has never seen it. A student asking "but why does that work" finds the soft spots immediately.

So you go back and fill them in. Not because a deadline forced you to, but because a person in front of you is depending on a clear answer. After enough of those conversations, the things you thought you understood you now actually understand, because you have had to say them out loud, simply, to someone who would notice if you were bluffing.

The simplest version is the hardest one

The harder skill mentoring builds is compression. A student does not need the full landscape. They need the one idea that unlocks the next step, said in the fewest words that still carry it. Finding that one idea, and trusting it enough to leave the rest out, is genuinely difficult. It is much easier to say everything you know than to choose the one thing that matters.

This is the same muscle the rest of my work runs on. A decision-maker reading an intelligence product is in exactly the position of a student: limited time, no appetite for the full derivation, one choice in front of them. The analyst who can find the single sentence that moves that choice is doing the same thing as the mentor who can find the single sentence that unlocks the concept. I got better at the second by practising the first, with people who told me, directly, when I had not landed it.

Watching someone act on your explanation

There is a feedback loop in mentoring that you do not always get in the work itself. You explain something, and then you watch the person try to use it. If your explanation was good, they move forward. If it was not, they stall in a way that points straight back at the gap you left. It is immediate and it is honest.

That loop trained an instinct I now use constantly: to imagine the moment after my work lands, when someone has to act on it without me in the room. If I can picture them getting stuck, I have not finished. The explanation is not done when it is correct. It is done when someone else can carry it on their own.

Why it compounds

The reason I keep doing it is that the skill it sharpens is the one that does not go out of date. Tools change, models change, the specific techniques I teach will be replaced. But the ability to take something genuinely complicated and make it clear enough for a busy person to act on is the part of this work that has only become more valuable as the mechanical parts get automated.

Mentoring is the cheapest, most direct training I have found for it. You sit with someone who does not yet understand, you find the words that get them there, and you watch whether it worked. Then you bring that same instinct back to the next thing you hand to someone who has to make a real decision. The students get a hand. I get a better grip on the one skill I most want to keep.