Starting at step three. Buying a training feels like progress and is, but without steps one and two you can't explain afterwards why that training was the right one.
What happens then is the same at nearly every organisation. A programme was purchased, there's an attendance list, and the question of whether it was sufficient gets a hesitant yes. That's not bad faith: the list feels like evidence. It just proves attendance, and the obligation is about literacy.
The repair is cheaper than redoing it. You don't have to repeat the training, you add the reasoning underneath. A five-minute measurement per employee still gives you the role analysis, and repeating it 60 days later gives you the difference that carries the rest of your file.
The second mistake is making the file heavier than necessary. The standard is proportionality, not completeness. Five documents that hold up are stronger than a forty-page policy nobody checked, and they can be produced in an afternoon.