Reviewing a course

There are three points where a person has to say yes, and they are deliberately far apart in cost: the plan, before anything is written; the document, before anything is rendered; the lessons, before anything is delivered.

None of them times out and none of them proceeds on its own. A course stops and waits at each one — so all three need somebody rostered to give them, not just the last.

Review one: the plan, before anything is written

Starting a course does not write the course. It writes a plan — the sections, the lessons inside each one and the minutes each lesson should take — and then it stops. Open the course and there is a panel headed Approve the plan for this course where the document will later be. Nothing is generated until somebody presses approve; that press is what starts the spending.

What to look for

Read it as a table of contents someone will be taught from, not as a list. Does the order work for someone meeting this for the first time? Is a section carrying twice the lessons of every other one? Is anything missing that your people actually need, and is anything here that they do not?

Minutes matter as much as titles. The lesson lengths on this screen are the ones the course will be written and rendered to.

What you can do

Add a lesson, remove one, move one, rename one, set the minutes a lesson should take, and leave a written instruction on a single lesson or on the plan as a whole. Sections are editable in the same way: rename one, add one, remove one, move a lesson between them.

Everything else — the objectives, the assessment blueprint, the scripts — is written from this plan afterwards and is not editable here.

Nothing is saved as you type. There is no draft and no autosave: approving is the act.

Why this one is first

This is the cheapest correction in the whole process. Moving a lesson here is a keystroke. The same change after the writing pass is a re-generation of every lesson it touches, and after rendering it is that plus hours of compute.

Review two: the document, before anything is rendered

Everything is text at this point — every lesson script and its opening line, the student workbook and the quiz, all written against the plan you approved. Fixing something here costs a minute. Fixing the same thing after rendering costs hours of compute and another review pass.

What to look for

The things a machine cannot know. This is where your reviewer earns their afternoon:

• Figures that changed. A threshold, a limit, a rate, a deadline.
• Policies that were superseded. The document may be the current file and still describe last year’s process.
• Internal vocabulary. The right word for a thing is the word your people actually use, not the one in the manual.
• Steps that are technically correct and practically wrong — the escalation path everyone bypasses, the form nobody fills in.
• Anything that is true but that you would not want said this plainly on video, in your organisation’s name, to someone who joined last week.

The teaching, not just the facts. Does the lesson order make sense to someone meeting this for the first time? Is anything assumed that has not been taught yet? Is there a lesson doing two jobs that should be two lessons?

The quiz. Can each question be answered by someone who watched the lesson it belongs to, and only that lesson? A question that needs outside knowledge is a bad question even when it is a fair one.

Before you count them: the size rule is one question per three minutes of finished video, with a floor of eight and a ceiling of forty. So an hour carries about twenty, but a fifteen-minute course carries eight rather than five, and every course longer than two hours carries forty — a four-hour course earns eighty by the ratio and is given forty, a twenty-hour course earns four hundred and is given forty. A short-looking quiz on a long course is the ceiling, not a failed generation. If you need assessment that stays proportionate on a long course, say so before you order it: what that needs is an assessment broken into section checks and a final, and that is not something produced today. The full rule, with both reasons, is in The delivery kit.

What you can do

Edit anything in place, or send a lesson back to be written again. Regenerating one lesson does not disturb the rest of the course.

Confirm

Confirming locks the document and allows rendering to begin. Until you confirm, nothing is rendered. After you confirm, changing the text means confirming a new version — which is not difficult, it is just deliberate.

Between the last two: the automated checks

You do not do anything here. When rendering finishes, five layers of automated checks run before you are asked to look at anything. If a hard gate fails, the course goes back rather than forward.

See Automated quality checks for what they cover. The short version is that they cover the mechanical failures — a black frame, a frozen screen, narration that drifted from the script, a missing lesson — so that your review time goes on the content.

Review three: the lessons, before delivery

When the checks are green, the course is ready for review and someone on your team watches it.

What to look for

The joins. Does the opening land? Does each lesson end and the next begin without whiplash?

The voice. Does the narration sound like the person it is attributed to, and does it read like something a person would say rather than something a machine would write? The automated checks reject the obvious tells; whether it sounds like your instructor is a judgement only you can make.

The lips. If a presenter is on camera, does the sync hold. This is a human check on purpose.

The slides. Do they support what is being said, or are they decorating it? Does the text on screen agree with the words being spoken?

The whole thing, once, at normal speed. Not scrubbed. If nobody on your team is willing to watch it end to end, it is not ready to send to three hundred people who have to.

Approve

Approval is recorded against the person who gave it. That record is why the question “who signed this off” has an answer that is not somebody’s memory.

Why three reviews and not one

One review at the end would mean discovering a wrong figure after it had been narrated, rendered, subtitled and put into a workbook. One review at the start would mean shipping a video nobody watched. And a plan reviewed only once the lessons are written is a plan reviewed too late to be cheap — twenty-odd writing passes have already run against it.

Each is cheap in a way the others are not, and none of them replaces another. Plan for three people-shaped stops in your schedule, not one.