Most training reports in Malaysian multi-site operations look healthy. Completion sits at ninety-four, ninety-seven, sometimes a hundred per cent. Certificates are on file, attendance sheets are signed, and the quarterly summary goes to management without much comment. Then a mystery shopper visit, an internal audit or a customer complaint shows that one outlet has been doing something differently from the other eleven, and has been doing it that way for weeks.
The report was not wrong. It was answering a different question from the one the operations manager needed answered. A training completion rate records that an employee opened a module and reached the end of it. Whether the changed step is now happening on the floor is a separate question, and most training systems are not built to answer it.
Malaysian regulation makes this distinction plainly. Section 15(2)(c) of the Occupational Safety and Health Act 1994 requires an employer to provide “such information, instruction, training and supervision as is necessary to ensure, so far as is practicable, the safety and health at work of his employees.” Four things are listed, and training is only one of them. Supervision appears separately because delivering the material was never treated as sufficient on its own. The same pattern appears in food service: Regulation 30(1) of the Food Hygiene Regulations 2009 requires food handlers to hold a Certificate of Food Handlers Training, and Regulation 11(1) requires the proprietor to keep those records. That certificate is good evidence that a person attended an approved course at some point. It is not evidence that last month’s revised allergen procedure is being followed at the Ipoh branch this week.
Why the usual responses do not close the gap
When execution turns out to be inconsistent, three responses are common, and none of them address the cause.
The first is to send the material again. This assumes the problem is exposure. If the employee already completed the module, another copy of it adds nothing; it tells them what they were told before, in the same form, at the same level of generality.
The second is to raise the pass mark. This makes the assessment harder without making it more relevant. An employee can score ninety per cent on questions about what a procedure says and still not recognise the situation on the floor where it applies.
The third is to add another attendance sign-off. This produces more paperwork proving the same thing the completion report already proved.
There is reason to take the underlying problem seriously rather than treat it as an administrative nuisance. A 2025 study by Md Ali and Watanabe in Food Safety and Risk examined 225 food service establishment inspection records in Malaysia from 2022 to 2023. Outbreak-associated premises showed an 18.71 per cent non-compliance rate against 13.38 per cent for routine inspections, and food handler training emerged as a highly influential node linking multiple other compliance failures. Training is connected to almost everything else that goes wrong. That is precisely why knowing it was delivered tells you so little on its own.
Step 1: Write down the observable action before you build the training
For each affected role, write one sentence: who does what, at what point, and what a supervisor would actually see. A revised allergen procedure might require the service crew to ask one new question at the point of order, the kitchen team to check a different ingredient list, and the outlet manager to decide what happens in an uncertain case. If you cannot write that sentence for a role, the training has nothing to be verified against.
Step 2: Test the decision, not the definition
“What is the allergen declaration requirement?” checks recall. “A customer asks whether the sambal contains shrimp paste and the prep sheet is missing — what do you do?” checks the decision the employee will actually face. A competency assessment built from real situations tells you where to look; one built from definitions only tells you who revised.
Step 3: Collect one piece of evidence from the shift itself
Good evidence is a photograph of the corrected shelf label, a completed opening check that now carries the new step, or a short supervisor confirmation against the observable action written in step one. Keep it to one item. Evidence collection fails when it is heavy enough that managers start completing it from memory at the end of the week. Where employees operate equipment, prepare food or perform care procedures, this step should include direct supervisor observation, because a module cannot verify a physical technique.
Step 4: Read the results by site and by shift, never as one organisational figure
An organisation at ninety-six per cent can contain one outlet at fifty-five. The averaged number is usually the reason the inconsistency stayed invisible for weeks.
Step 5: Re-check after the attention has moved on
Two to six weeks after rollout, sample the same question again. Compliance in the first week is often a product of supervision rather than of habit, and the difference between the two only shows up later. This is also the point at which refresher training can be aimed at the sites that need it, rather than reissued to everyone.
What changes once you measure execution
Sites stop drifting apart quietly, because the drift becomes visible while it is still small. Gaps surface within days of a rollout rather than at the next audit. Conversations with outlet managers become specific: you are discussing one step at one branch rather than a general impression that training is not working. Training records show what employees did, not only what they were assigned. And the amount of manual chasing falls, because the follow-up is directed at the sites that need it instead of everyone.
Where Wisteria fits
Wisteria supports this sequence directly. Wisteria Scan checks connected sources and flags documents that may require training, so a revised SOP does not sit unnoticed. From the flagged material, AI prepares draft flashcards and scenario questions, which your trainer reviews and approves before anything is published: automatically found, human approved. The approved version then reaches the affected teams on the devices they already use, so every shift receives the same material. Assessments check understanding rather than attendance, and results can be read by site and by shift. When the source document changes again, affected training can be reviewed and reassigned to the people it applies to.
A specific next step
If you would like to see how this would work on your own material, send us one SOP or training document you have revised in the last six months. We will review it and show you a set of sample flashcards and scenario questions built from your content, so you can judge the quality before deciding anything.
Carmen Lee — Instructional Designer, Wisteria
Six years in the training industry, and someone who cares about frontline learning.
