Learn ยท Choosing what to build
Video, documentation or simulation: choosing by training task
Each format is right for something. The mistake is choosing by what the team knows how to make. Choose by the task: what a wrong action costs, how often the screens change, and whether the learner has to do it or only recognise it.
Documentation for tasks people look up. Video for tasks people need to recognise. A simulation for tasks people have to perform, where a wrong action has a cost and the interface is where the mistake happens. These formats can complement each other.
Three questions that decide it
| Ask | Documentation | Video | Simulation |
|---|---|---|---|
| Does a wrong action reach a customer, a balance, a record or a permission? | Rarely the right answer | Shows it, cannot check it | The learner makes the choice and is told why it is wrong |
| Will the learner do this once a week, or once a year? | Once a year: they look it up | Either | Once a week: they need the habit |
| How often do the screens change? | Cheap to update | Re-shoot | Re-record the task, or describe the change in a sentence |
| Does the learner need to recognise a control, or find and use it? | Recognise | Recognise | Find and use |
| Do you need the LMS to know they did it? | Viewed | Watched | Completed every step |
Documentation: for tasks people look up
Reference material wins when the task is rare, when the learner is at the screen with the document beside it, and when the answer is a value rather than a procedure: which code, which field, which threshold. It is the cheapest format to keep true, because a changed label is a changed word. It is the wrong format for anything the learner has to do fluently under pressure, because reading and doing are different skills and a document confirms exposure, not ability.
Choose it when: the task is infrequent, the cost of a wrong action is low or reversible, and the learner will have the document open while they work.
Video: for tasks people need to recognise
A recording of the task shows pace, sequence and what the screens look like when things go right. It is quick to make and persuasive to watch. Two limits matter. A video cannot tell whether the viewer could do the task; watching someone choose Solved is not the same as choosing it. And a video is fixed to the day it was shot: one moved button and the whole thing is subtly wrong, which viewers notice and stop trusting. The cost is not the first recording but the fifth re-shoot.
Choose it when: the learner needs to recognise the flow rather than perform it, the product is stable, and the point is orientation rather than competence.
Simulation: for tasks people have to perform
An interactive simulation puts the learner on a copy of the real screens and checks what they do. It is the only format where a wrong action produces a reason instead of a consequence. It costs more to make than a document, and historically far more than a video, because the interface had to be rebuilt in an authoring tool. That is the cost Shadow Capture is built to remove: the screens come from one recording of the task, so a simulation starts with the workflow itself. Allow time to review, edit and test it before publishing.
Choose it when: a wrong action has a cost, the task is performed often enough that fluency matters, and there is a decision inside it that people get wrong on the live system without any error appearing.
The same task, three ways
Take closing a support ticket. As documentation: a table of statuses and when to use each. As video: an agent working a ticket, three minutes, no way to know if the viewer absorbed the Pending versus Solved distinction. In an illustrative simulation, the learner writes the reply, sends it, and then meets the status dropdown, where choosing Pending produces this:
"That leaves the ticket in the queue. Pending is for tickets waiting on the customer or another team, and this one is neither. Payments have confirmed the re-issue and Rosalind has been answered, so the right status is Solved."
Illustrative feedback for a fictional support ticket
The documentation and the video both contain that rule. Only the simulation finds out whether the learner has it.
A way to sort your backlog
- List the tasks people ask about. Support tickets, onboarding questions, the things an expert keeps demonstrating on calls.
- Mark the ones where a wrong action reaches something real. A customer, money, inventory, a permission, a record other people rely on. Those are simulation candidates.
- Of those, keep the ones with a decision inside. A status, a stage, a priority, an approval. Pure data entry with no choice is a hint away from fine; the decision is where the training pays.
- Send the rest to documentation, and make the reference easy to reach from the screen where the question comes up.
- Reserve video for orientation: what the product is, what a good day in it looks like, before any of the tasks.
The expense approval example shows how a recorded workflow can give learners decisions to practice.