How this course works
Twelve sessions in two tracks of six. Designer is this one - designing the learning. Delivery is track B - getting it used inside an organisation and proving it changed something. You can take either alone; together they are the whole job.
★ What you walk out with tonight. One objective for Daybreak's data-contracts training that passes all four checks, your own objective linted and fixed, and a reliable way to tell an objective from a topic.
LiveThe running project2 min▶
Daybreak keeps breaking its own dashboards. A producer renames a column, the nightly load writes nothing, cron reports success, and by 08:40 thirty-four downstream assets are wrong. It has happened three times this year.
Somebody has decided the answer is "a training session on data contracts for the engineering team", and it has landed on you. Track B session 1 asks whether that is even a training problem - a question worth asking before you build anything. This track assumes the answer came back partly yes, and gets on with designing it.
The request as received: "Can you do an hour on data contracts so this stops happening." That is a topic and a duration. It is not an objective, and the difference is tonight.
Self-studyThe designer track, in six sessions2 min▶
| Session | What the design gains |
|---|---|
| a1 · tonight | One objective that can be failed, with a condition and a criterion |
| a2 | The assessment, written before any content exists |
| a3 | A sequence that respects how little anybody can hold at once |
| a4 | Activities that produce the behaviour rather than exposure to it |
| a5 | Questions that test the objective, and feedback that changes the next attempt |
| a6 | Materials, a job aid, and a plan for the room you actually get |
If you only do three, do a1, a2 and a4. The objective decides everything, the assessment stops you drifting, and the activities are where the learning actually happens.
If nobody can fail it, it is not an objective
6 min liveThat is the whole test, and it disqualifies most of what gets written. "Understand", "know", "be aware of", "appreciate" - none of them describe anything you could watch somebody do, which means none of them can be practised or checked.
LiveThe four parts, and why the last two go missing5 min▶
- Who. Not "participants". The engineer who owns a producing service. Different people need different objectives from the same session, and pretending otherwise is how you end up serving nobody.
- An observable verb. Classify, write, run, diagnose, choose. Something you could stand behind them and watch.
- A condition. Given what? A real pull request, a live incident ticket, a schema diff. This is the part that decides what the practice material has to be, which is why skipping it produces a session with no materials in it.
- A criterion. How well? No more than one error in ten, within five minutes, against the checklist. This is the part that decides whether the assessment is possible at all.
Condition and criterion are the two that get dropped, and they are dropped for the same reason: they force you to decide something you would rather leave open. An objective without them is comfortable, unfalsifiable, and useless for designing anything.
"Participants will understand the importance of data contracts and be aware of the new schema change process."
What do you build? Anything. Which is what happens: an hour of slides about why contracts matter, a walkthrough of the process, some nodding, and three weeks later somebody renames a column. The objective could not tell you to include a single exercise, because it never named anything anybody has to do.
"Given a proposed schema change on a service they own, the producing engineer will classify it as safe or breaking, and for a breaking change draft the deprecation notice, with no more than one misclassification in ten."
Now the session is obvious: ten real schema changes, sorted; a deprecation notice written and reviewed; the two ambiguous cases argued about. You have your assessment, your practice material and your run time, from one sentence.
Four checks, run on your actual text
5 min liveThis one is measured rather than modelled. It reads what you typed, looks for non-observable verbs from a fixed list, checks for a condition and a criterion, and reports what it found by name.
LiveWhat the linter cannot do, said plainly4 min▶
Both word lists are chosen, not derived. Sixty-five observable verb stems, inflected automatically; nineteen non-observable ones. That means:
- A good objective can fail a check. If you used a perfectly observable verb that is not on the list, it will say it found none. Look at what it says, not just at the tick.
- A bad objective can pass all four. "Given a laptop, the participant will attend the session within 60 minutes" passes every check and teaches nothing. The checks catch common failures, not every failure.
- It has no opinion about whether the objective is worth having. That is session a2, and it is the harder question.
An earlier version had forty verbs and rejected "classify" - a genuinely good objective scored badly. A tool that is wrong about your good work teaches you to ignore the tool. If you are building anything like this, err toward false passes rather than false failures.
Write one objective, four parts, and lint it
★ 22 min · everyone buildsOne objective. Not a set - one. Most training that fails is trying to do six things and achieving none of them, and the discipline of writing exactly one is how you find that out early.
Who. The specific person, named by their situation. "The engineer who owns a producing service" - not "participants", not "the team", and not two different roles at once.
Verb. Something you could watch. Run it past the failure test: could they do this and get it wrong? If not, find a different verb. "Understand" always fails this.
Condition. Given what? This sentence is also your materials list - if you cannot say what they will have in front of them, you do not yet have an exercise.
Criterion. How well? A count, a time, a checklist. If you cannot write one, ask what would make you say afterwards that it worked. That answer is the criterion.
Cut to one. If your objective has an "and" in it, it is probably two objectives that need different practice. Pick the one that matters most and write the other on the not-this list.
Write Daybreak's objective yourself first. The request was "an hour on data contracts so this stops happening". Turn it into one sentence with all four parts before you look at mine again.
Paste it into the linter above. Read what it says, not just how many ticks. The detail line tells you which verb it found and why it objected.
Now paste your own real objective. Something you wrote for actual training, or the learning outcome from a course you were sent on. Most people's first result is 1 of 4.
Fix the verb first. It is the cheapest fix and it usually forces the other three - once you write "classify" rather than "understand", the condition ("given what?") becomes an obvious question you have to answer.
Write the criterion by asking the afterwards question. "What would make me say this worked?" If the honest answer is "people seemed engaged", you have found something important about the request rather than about your writing.
Count your "and"s and split. Daybreak's has one - classify, and draft the notice. Those are genuinely two abilities that need two different exercises, and knowing that now is worth an hour later.
Read the objective and ask what it forces you to build. A good one is uncomfortably prescriptive. If you can still build anything you like, it is a topic wearing an objective's clothes.
If you take one thing from tonight. Ask "could somebody fail this?" of every objective you meet. It takes two seconds, it disqualifies most of them, and it is the entire discipline compressed into one question.
Before session a2
20 min- Lint three real objectives. Yours, or the learning outcomes from a course you have been sent on. Note the pattern - almost everybody is missing condition and criterion, and almost nobody is missing the verb entirely.
- Rewrite one to 4 of 4 and notice how much of the session it just designed for you.
- Find a real "and". An objective doing two jobs. Split it and decide which one you would drop if you only had 40 minutes.
- Ask a colleague the failure question. Show them an objective and ask "could somebody fail this?" Watch how fast it resolves an argument about scope.
- Bring your best objective to a2, because a2 writes the assessment for it before any content exists, and a weak objective makes that impossible in a way you can feel.
Three questions
Clear all three and this session gets stamped in your passport.
1 · Why is "participants will understand our data governance policy" not an objective?
Breadth is not the problem, testability is. Because nothing in it can be watched, the objective cannot tell you what exercise to build or what to leave out - so the session becomes whatever the expert felt like covering, which is exactly what happens in practice.
2 · Which two parts of an objective are most often missing, and why?
"Given what?" commits you to producing real practice material, and "how well?" commits you to a standard you can be measured against. Both are uncomfortable, and an objective without them is comfortable and useless for design.
3 · The linter's verb list was deliberately made long. Why does that matter?
An earlier version had forty verbs and failed "classify" on an otherwise perfect objective. False failures destroy trust in a checker much faster than false passes do, and this is worth remembering whenever you build anything that grades somebody's work.
What this session teaches, and where it came from
Daybreak and its training request are a teaching scenario; the linter's analysis is real. Full note in the course's source map.
Designer a1 cheat sheet · pin this
You have an objective that names an ability. Session a2 does the thing that feels backwards and is not: write the assessment before any of the content exists. Deciding how you would know deletes more slides than any other single move, and it is why so many courses contain material nobody can explain the purpose of.