Training that survives contact with the job
A single session before go-live is the standard approach and the one most reliably forgotten by the time it is needed.
The usual training plan is one session for everybody, delivered shortly before launch, covering the whole system. It is efficient to deliver and most of it is gone within a week, because it was learned before there was anything to attach it to.
Teach what people do, in the order they do it
Comprehensive system tours are the wrong shape. Most people need to do three or four things, and everything else is noise that makes the three or four harder to find. For a concrete example of remote workforce management after purchase, this overview can be used to identify the behaviours and measures that rollout should actually change.
Build training around roles: here is what you do, here is how, in the order you will do it. Ten focused minutes per role beats an hour of coverage, and it can be delivered by someone who is not an expert in the whole system.
Training delivered a fortnight before someone uses a feature is forgotten. Delivered the day they need it, it sticks — which argues for short, repeated, role-specific sessions rather than one event.
Write the one-page version
For each role, a single page: the three or four tasks, the steps, and where to go when stuck. Not the vendor's manual — yours, in your language, referring to your process and your field names.
This is a couple of hours of work per role and it is what people actually use. It is also what new starters get in six months, when the training session is long forgotten and nobody remembers what was said in it.
Use the pilot group as the trainers
People who ran the trial know the product, know the local process, and are peers rather than management. Training delivered by them is better received and more accurate about the awkward parts.
It also distributes the support load after go-live, which is otherwise concentrated entirely on whoever led the implementation.
Plan for the second round
Everyone needs training twice: once to be able to start, and once a few weeks later when they have hit real situations and have real questions.
The second session is the one that produces competence and the one nobody schedules. It is short, it is mostly questions, and it is where the workarounds get caught before they set.
Do not forget the occasional user
The person who uses the system monthly will never become fluent and will forget between uses. They need a written guide, a simple path, and no expectation that they remember anything.
They are also, frequently, the source of the worst data, because they guess when they cannot remember. Designing for them specifically is worth more than it appears.
Build it into onboarding
After launch, training tends to stop, and every subsequent new starter learns by asking a colleague. Within a year the process being taught has drifted from the process that was designed.
Adding the role sheets to the standard onboarding checklist takes minutes and keeps the practice from decaying.
For practical operational guidance after deployment, CISA resources and tools provides public reference material.