Staff Training & Adoption of a New System
A full-day session three weeks before anyone has a reason to use the system teaches almost nothing, and everybody involved knows it. What works instead is short, role-specific and delivered at the moment of need — plus one uncomfortable rule about who trains the storekeeper.
The standard training plan is a day in a room with a projector, everybody present, covering every module, three weeks before go-live. It exists because it is easy to schedule and easy to report as complete. Retention from it is close to zero, and this is not controversial — ask anyone who has sat through one what they remembered a fortnight later.
The reason is not that the training was bad. It is that people cannot learn a procedure they have no current use for. The storekeeper listening to a demonstration of receiving in a room with no delivery in it is watching a film about his job.
Train at the point of need, not on the plan's schedule
The single change that transforms adoption is moving each session from "before go-live" to "the first time you actually have to do this". The storekeeper learns receiving standing at a real delivery. The cashier learns the till on the first morning with real customers, quietly, with the supervisor beside them. The buyer learns requisitions the first time they need something.
This is more work to organise than one day in a room and it is the difference between a system used and a system endured.
-
One role, one task, twenty minutes
Not "inventory training" — "how to receive a delivery". A session covering four tasks teaches one of them, usually the first, and nobody can tell you which because the questions ran out before the end.
-
On their own screen, with their own login
Watching someone else's screen builds recognition, not capability. They will recognise the button and not find it. Every session should end with them having done the thing, on their account, with their permissions — which is also how you discover the permission you got wrong.
-
With real data, from day one
Practice items called "Test Item 1" teach the clicks and none of the judgement. Use a real delivery, a real supplier, a real item — with the understanding that the record stands, because that is the point.
-
Repeat it on day three and day ten
The second pass is where the real questions arrive, because by then they have hit something the first pass did not cover. The third is where it becomes theirs. Budget three passes per task and you will actually need two.
-
Have them teach the next person
The most reliable retention test available, and it staffs your own support: whoever trained a colleague is now the person that colleague asks at 4pm rather than you.
The question is never "have they been trained". It is "who do they ask when it goes wrong at 4pm on a Friday" — and if the answer is the vendor, adoption has not happened.
The uncomfortable rule about who trains the storekeeper
Not the vendor. Not the IT person. The person who will be asking the storekeeper for the figures.
This is unpopular because it costs a manager real hours, and it is the highest-leverage decision in the whole adoption plan. A vendor teaching the buttons transmits procedure. A manager teaching the same task transmits something else entirely — that this is how the work is now done, that the figures will be looked at, and that the person asking understands the process well enough to notice when it has been skipped. The last of those is worth more than the training.
People do not adopt systems. They adopt the expectations of whoever checks their work.
What to do about the person who will not use it
There is usually one, occasionally two, and they are often good at their job — which is precisely why they resist. They have a working method, built over years, and the new system is slower for them specifically because they were fast at the old way.
Three things work, in this order. First, find out whether they are right: sometimes the resistance is a genuine workflow problem and the resister is the only person paying enough attention to notice. Second, make the old path unavailable rather than discouraged — the paper book gets collected, on a date, publicly. Third, if it is still happening in week six, it is a management conversation and not a training one, and continuing to treat it as training is how it becomes permanent.
Watch for the helpful workaround
The most damaging resistance is not refusal — it is someone maintaining a private spreadsheet "so the reports come out right". It looks like diligence, it is often praised, and it means two records exist and the system's one is being quietly corrected by hand. Find these in week three by asking, without any edge in your voice, what else people are keeping.
What is available to lean on
Resources that exist, and how to use them properly
- The Academy — structured courses per module with assessments and a verifiable certificate. Best used as reinforcement after a real session, not as a substitute for one; a certificate proves someone completed a course, not that they can receive a delivery.
- The help centre — task-level articles for the "how do I again?" moment, which is the moment that actually decides whether someone asks you or gives up.
- The in-app assistant — genuinely useful for finding where something lives. Be clear about what it is: keyword-matched guidance over the product, and it does not read your organisation's data. It will tell you where the dead-stock report is; it will not tell you what is in it.
- Your own one-pager per role — the four things this role does, in your own words, with your own thresholds on it. The vendor cannot write this because it contains your policy, and it is the document people actually keep.
Measuring adoption without pretending
Attendance and certificates measure activity. Three things measure adoption, all of them observable in the first month.
| Signal | What good looks like | What it tells you |
|---|---|---|
| Entry lag — the gap between a thing happening and being recorded | Same day, by the person who did it | Yesterday's movements entered tomorrow by someone else means the system is a reporting chore, not a working tool. |
| Who is entering | The person doing the work | If one clerk enters everything, you have digitised the paper book and added a step. |
| Questions changing shape | From "where do I click" to "should this be a transfer or an adjustment" | The second kind of question means someone has internalised the model and hit a genuine judgement call. That is adoption. |
Our take
Cancel the all-day session. Run twenty-minute, one-role, one-task sessions at the moment of need, on the user's own login, with real data, three times each — and have the manager who will ask for the figures do the teaching rather than the vendor. Then watch entry lag rather than attendance, because attendance is a number that only ever goes up.
Train at the delivery, not in the boardroom
Courses with assessments and certificates, task-level help articles and an in-app assistant for the "how do I again?" moment — as reinforcement behind the sessions your own managers run, which is the part that decides whether any of it lands.
See plans & pricingFrequently asked questions
How much training does a new operations system need?
Less than most plans allocate and spread far wider. Twenty minutes per role per task, repeated about three times, delivered at the moment the person first has to do the thing. A single all-day session covering every module three weeks before go-live is easy to schedule, easy to report as complete, and retains almost nothing — because people cannot learn a procedure they have no current use for.
Who should deliver the training?
The manager who will be asking for the figures, not the vendor and not IT. A vendor transmits procedure; a manager transmits the expectation that the work is now done this way and that the numbers will be looked at. It costs real management hours and it is the highest-leverage decision in the adoption plan.
Should we train on test data or real data?
Real data, from the first session. Practice items called "Test Item 1" teach the clicks and none of the judgement, and they let people avoid the part that matters — that the record stands. Use a real delivery, a real supplier and a real item, on the user's own login with their own permissions, which is also how you find the permission you got wrong.
What do we do about a member of staff who refuses to use it?
First check whether they are right — resistance is sometimes the only person paying enough attention to a genuine workflow problem. Then remove the old path rather than discouraging it: the paper book gets collected, on a stated date, publicly. If it is still happening in week six it is a management conversation, and continuing to treat it as a training problem is how it becomes permanent.
How do we know whether adoption is real?
Three signals. Entry lag: are things recorded the same day by the person who did them, or entered tomorrow by a clerk? Who is entering: one clerk keying everything means you have digitised the paper book and added a step. And the shape of the questions: when they move from "where do I click" to "should this be a transfer or an adjustment", someone has internalised the model. Attendance and certificate counts measure activity, not adoption.
Can the in-app assistant answer questions about our data?
No. It matches keywords against guidance about the product, so it is genuinely good at telling you where something lives and how a feature works. It does not read your organisation's records — it will point you at the dead-stock report and cannot tell you what is in it. Useful to say out loud during training, because people assume the opposite and then distrust it.
