One session is one round

Many people assume a machine only needs to read through the data once and it is done. The reality is the opposite: it studies the same material over and over, exactly the way we revise.

In the field, each complete pass through all the data is called an epoch. And the data is not force-fed all at once either; it is split into small batches of a few dozen to a few hundred examples, so the machine draws lessons gradually after each batch rather than trying to swallow everything in one go. A figure to make it concrete: a common image recognition model called ResNet-50 typically has to study an enormous library of images, over 1.2 million of them, around 90 times to become good enough. Not once and done, but many rounds, each one a little better.

Developing people in an organisation seems to me exactly the same. One training session, however carefully prepared, is only one round. Real skill forms through many repeated rounds: do it, get feedback, adjust, do it again, on small batches of work just big enough to observe and correct in time. Expecting someone to be fluent after a single session is probably an expectation placed slightly wrong.

A capability you want your team to be genuinely solid in: how many rounds is it being repeated over, or has it stopped at exactly one session?

To your desk

A machine does not get better from a single pass, but from many rounds over small batches. Building our own skills follows the same rhythm.

This week, instead of running one very long training session and hoping the team is good immediately, try picking one specific skill and splitting it into several short rounds: practise one small situation each week, take feedback, repeat the following week. A steady loop, however short, usually carries people further than one large effort that happens only once.

The question worth asking is not “have we taught it” but “have we repeated it enough rounds”.