InnerWeb
Skip to main content

Step-by-step training tests: good practices

An OHS training passed from a sheet of paper on which a colleague wrote the correct answers checks nothing. The OHS trainer can change this by building a step-based training with random questions and scoring on the server side. Below are the practices for creating such a training in InnerWeb, from dividing the material to the test settings.

75
questions in the pool of the sample training
Source: sample calculation
25
questions solved by the participant
Source: sample calculation

Two ways of running a training

In InnerWeb a training can follow a classic or a scenario-based flow. In the classic flow, the participant reviews the materials before the test, in a chosen order (video and PDF), and then solves a single final test. One material can be marked as reviewed after the test, for example a card with evacuation rules, which one returns to after passing.

In the scenario flow, the training consists of steps. Each step is a material, meaning PDF pages or a time range of a video, random questions from a defined range and the required number of correct answers. A failed step repeats the fragment of the material, so the participant returns to content they did not master, instead of going through the whole course from the beginning.

The choice between them depends on the content. A short introduction for visitors and suppliers works best in the classic layout. A longer training for workers of external companies, covering several risk areas, gains from being divided into steps.

Rule 1: divide the material by areas, not by time

A step should correspond to one area of knowledge: evacuation routes, vehicle traffic, work at height, hazardous substances. Do not divide the training by video duration, because then the questions fall on random fragments. A participant who failed the step on vehicle traffic returns to the fragment on vehicle traffic, not to a random minute.

You set the scope of the material in a step as PDF pages or a video time range. This makes it possible to use one long film in many steps, each with its own questions, without cutting it into separate files.

Sample training: question pool and test per step

Sample training: question pool and test per stepFive steps of the training, each with a pool of 15 questions and a test of 5 questions.051015questionsQuestion pool (Evacuation): 15 questions15Test per step (Evacuation): 5 questions5EvacuationQuestion pool (Vehicles): 15 questions15Test per step (Vehicles): 5 questions5VehiclesQuestion pool (Machines): 15 questions15Test per step (Machines): 5 questions5MachinesQuestion pool (Substances): 15 questions15Test per step (Substances): 5 questions5SubstancesQuestion pool (Reporting): 15 questions15Test per step (Reporting): 5 questions5Reporting
  • Question pool
  • Test per step
Source: Sample calculation: 5 steps, pool 15, test 5

Rule 2: a question pool larger than the test

Random questions from a defined range make sense only when the pool is larger than the number of questions in a step. For example, a step with a pool of 15 questions and a test of 5 questions gives different sets at each attempt. A participant who received set A cannot pass knowledge of set B to a colleague.

The required number of correct answers should be lower than the number of questions only when the plant deliberately allows for an error. For critical topics, such as evacuation or work in a hazard zone, consider requiring all answers to be correct. For auxiliary topics, 4 out of 5 is a reasonable threshold.

  • A question pool per step larger than the number of questions in the test, for example 15 to 5.
  • A requirement of all correct answers for critical topics.
  • A threshold of 4 out of 5 for auxiliary topics.
  • Questions describe situations from the plant, not definitions from a code.

Rule 3: scoring on the server side

The scoring of a passed scenario-based training takes place on the server side, and the player does not receive the scoring of the answers. This has two practical effects. First, one cannot check in the browser tools which answers are correct. Second, the participant cannot manipulate the result, because the server counts it, not their phone.

For the trainer this means that questions and answers must be configured in the panel, not in a file that could circulate among participants. It is worth running a trial pass on your own account to check how the training looks from the participant's perspective.

Treat with caution the claim that random drawing and server-side scoring solve the problem of copying. A participant can still talk to a colleague at the tablet. The training's resistance to such practices is increased only by the fact that the questions differ and concern situations, not memorized formulas.

Rule 4: shuffling the answers

The order of answers in a test is shuffled by default, and each download of the training is a new attempt with a new shuffle. The setting can be turned off for a training. We recommend leaving shuffling on in almost all cases. The exception is questions in which the order follows logic, for example the sequence of actions for an emergency shutdown of a machine.

The very fact of shuffling answers defeats the simplest copying method, which is passing on a string of letters such as "B, A, C, D". The participant must know the answer, not its position.

After turning shuffling off, the results should be checked periodically for suspiciously identical answers within the same group.

Example layout of a training

A sample introductory training for external companies may have 5 steps: evacuation, vehicle traffic, work near machines, hazardous substances, reporting incidents. In each step there is a pool of 15 questions, a test of 5 questions, a threshold of 4 correct answers, and in the evacuation step all 5. In total, the participant solves 25 questions from a pool of 75, and two people rarely get the same set.

Such a training takes longer than a single test at the end, but it gives the trainer information about which area the participants fail in. If 30 percent of people do not pass the step on hazardous substances, the problem is the material, not the participants.

Checklist

Video

Frequently asked questions

How does a classic training differ from a scenario-based one?

The classic one has materials before a single final test. The scenario-based one consists of steps with their own material, random questions and the required number of correct answers.

What happens when a participant fails a step?

The failed step repeats a fragment of the material, so the participant returns to the content they did not master.

Are the answers shuffled?

By default yes, and each download of the training is a new attempt with a new shuffle. The setting can be turned off for a training.

Who scores the test?

The server. The player does not receive the scoring of the answers, which makes it harder to manipulate the result.

Let's talk about conditional admission

We will show, using your plant as an example, how trainings and qualifications become a condition for entering the site. Book a short conversation or a demo.

Let's talk