Instructional Design Cannot Repair a Bad Capability Claim at the End

Your course has a cinematic opening.

Accessible captions.

Tasteful motion graphics.

And an exit survey with enough agreement emojis to power a quarterly business review.

It also has a completion record.

Somebody now wants to put “capable of incident triage” into the skills system.

This is where the course starts being asked to do something it was never designed to do.

Instructional design is not a capability claim.

It is the disciplined work of selecting objectives, activities, media and learning conditions.

Assessment is the disciplined work of asking whether a result can support a particular decision about performance.

A completion record is an institutional wrapper around participation in the experience.

Three related objects.

And the LMS button labelled Award skill does not know the difference between them.

Start with the decision that needs evidence

The first sentence is not “the module should be engaging.”

It is this.

What will this result let somebody decide, and under what conditions was it produced?

Then name the population, the decision maker, the consequence, and whether the decision can be reversed.

A short diagnostic at entry can identify missing prerequisites.

It is not secretly a grade, a promotion signal, or permission for independent operation, merely because it returned a number.

A more consequential decision needs a correspondingly credible chain of evidence.

That sentence gives the objective a job.

In Dick and Carey’s ten-step systems approach, the team writes performance objectives with a performance, a condition and a criterion.

Then it develops criterion-referenced assessment instruments.

Then, and only then, it develops instructional materials.

The assessment is step five.

The materials are step seven.

This is occasionally received as an administrative inconvenience by the team holding a nearly complete storyboard.

It is actually the moment the programme decides what it is claiming.

ADDIE says the same thing in another dialect.

Analysis, Design, Development, Implementation, Evaluation.

Task inventories become measurable terminal and enabling objectives during Design, before the media production machinery starts humming.

The military IPISD specification makes that machinery unusually explicit, with 19 standardized sub-tasks and formal review gates.

The gates are not there because instructional designers enjoy paperwork.

Nobody enjoys paperwork.

They are there because a course cannot acquire a missing criterion by becoming more beautifully narrated.

A good objective can still be attached to the wrong evidence

The usual failure is tidier than a bad course.

The outcome says analyse.

The final check asks candidates to recognise a definition.

Both concern the same topic.

The evidence does not concern the same performance.

That is not alignment.

That is a shared noun wearing a lanyard.

An assessment blueprint makes the mismatch visible before your reporting system makes it permanent, which it will.

Content or task areas down one side, performance demands across the other: recall, interpretation, procedure, explanation, judgment, production.

Your weighting follows the decision, not the calendar.

Equal treatment time is not a defensible rationale.

Neither is the fact that the instructional team has already made several videos about the subject.

Format follows evidence too.

A one-best-answer item can compare judgment efficiently across a large group.

A short constructed response can show a chain of reasoning.

A performance station can show execution under stated conditions.

A portfolio can show work over time.

Oral questioning can show explanation, provided the prompt and the rubric do not drift with whoever is asking.

The format is not the capability.

It is a way of eliciting evidence about the capability.

So when a client handover requires observed performance or an artefact, a multiple-choice check does not become more valid through elaborate distractors, a timer, and a progress bar reading “Almost there.”

It is still evidence of what the item elicited.

The rubric is part of the claim

Tasks arrive with their scoring rules, an intended answer or exemplar, a rationale, and a subject-matter check.

For performance work, you define the observable dimensions, the evidence, the levels and the non-compensable critical errors before anyone begins.

Levels have to name visible features.

“Good” and “excellent” are not level descriptors.

They are wishes with formatting.

The distinction matters most where a total score can hide a decision you would not actually make.

A safety-critical criterion is conjunctive: it has to be met, and no amount of good work elsewhere converts a critical omission into an acceptable result by arithmetic.

The same discipline applies to the reported pass.

A claim states the routine, the critical criterion, the condition and the rule.

Under these observed conditions, this result supports a decision that the trainee can complete this task with no critical omission.

“Completed Incident Triage Foundations” identifies a catalogue object.

It is not the same kind of statement.

The design work is genuinely good, which is the problem

None of this argues against polished courses, learning objectives, learner feedback, or instructional designers who can make a difficult procedure comprehensible before lunch.

Instructional design solves real problems.

It makes learning coherent.

It sequences practice.

It catches accessibility barriers before they become an accidental test of keyboard speed or platform familiarity.

It gives your dispersed workforce a repeatable experience, rather than a collection of charmingly contradictory manager explanations.

For complex work, the 4C/ID model is especially clear about why content alone is insufficient.

It combines four components: authentic whole learning tasks, supportive information for the non-routine aspects, just-in-time procedural information, and part-task practice for routines that need automaticity.

The learner meets whole tasks of increasing difficulty, and the scaffolding fades toward autonomous performance.

That is excellent instructional design.

It does not mean that viewing the sequence proves autonomous performance at work. It tells the team how to build toward it.

Agile models are useful for a different reason.

A savvy start produces low-fidelity functional prototypes in a day or two, then iterative loops move through a design proof to alpha, beta and gold.

Fast feedback finds a confusing interaction before a large release.

It cannot turn an unspecified authorization decision into a valid one.

The design deserves credit for the thing it is evidence of: a considered learning experience.

The mistake begins when that evidence gets promoted into proof that a different thing happened.

Pilot the response process, not just the slide deck

The course preview goes very well.

The expert approves the examples.

The programme manager enjoys the animations.

The authoring tool produces a completion badge with a reassuring amount of blue.

None of that tells you what your candidates think the task requires.

Dick and Carey’s formative sequence has three stages for exactly this reason.

One-to-one work with three to five representative learners.

Small-group trials with eight to twenty.

Then a field trial with thirty or more, in the real operating environment.

The small-group stage checks whether learners can execute independently, and whether the item-level evidence supports the intended mastery threshold without external coaching.

At field trial, server bandwidth, instructor fidelity, administrative handoffs and the live LMS all become part of the evidence.

A beautiful assessment that behaves differently when the VPN has opinions is measuring more than the intended capability.

In a thirty-minute pilot, three candidates stopping at the same instruction are evidence about the instruction.

They are not a small rebellion against the course.

So your pilot records completion time, clarification requests, navigation errors, resource use and interruptions.

Then the team can compare the intended response process with the one actually produced.

Then lock it.

Version, instructions, permitted resources, accommodation process, scorer materials, incident process.

Item 3.2 and rubric 2.1 make a later review possible.

“Latest final revised” makes a later review into archaeological fieldwork.

Satisfaction is a useful report about satisfaction

The post-course reaction score is often the cleanest number on your dashboard.

Understandably.

It measures reaction: satisfaction and perceived utility.

Nearly every programme collects it.

It is useful.

A course your learners cannot navigate, understand or tolerate has a real design problem.

It is not a report of workplace performance.

The next level up measures knowledge and skill acquisition, through pre- and post-tests or rubric-scored simulation.

Roughly three-quarters of programmes get that far.

The level after that measures on-the-job application, through supervisor audits at 30, 60 and 90 days, or peer observation.

Implementation falls to something like a fifth of programmes.

The last level measures organisational results.

Defect rates.

Sales growth.

Retention.

Fewer than one programme in ten reaches it.

That drop is not a moral indictment of the people running learning.

Observing workplace behaviour and results costs more than sending a survey link.

It is a reminder that those reports make distinct claims.

A 95 percent pass rate on the learning assessment does not contradict a later finding that fewer than a fifth of graduates apply the procedure at 60 days.

It tells you to investigate management reinforcement and workplace friction, rather than congratulating the completion report for solving a different problem.

The platform deserves separate inspection too.

The System Usability Scale runs ten items, five positive and five negative, and produces a composite out of 100.

Its baseline average is 68.

A score above 80 sits in the top tenth.

Interface friction can make an assessment partly about navigation, when navigation was not in the claim.

The learner may have enjoyed the course.

The system may have been usable.

The learner may also be unable to perform the task under the conditions that matter.

All three can be true at once, which is inconvenient for a single-field report.

Abundant content, unchanged claim

Generative production makes content abundant.

A team can now produce a module, a voiceover, a knowledge check and a completion signal at a speed that would previously have required a request form, a vendor, and somebody discovering the brand-font file was missing.

Which makes the distinction more urgent, not less.

An AI system will happily ingest course titles, objective labels, quiz scores, completion dates and learner reactions.

They are structured.

They look like workforce data.

It can infer a skills record, recommend an assignment, or flag a person for promotion, with no visible malice and a beautifully formatted confidence score.

But a label is not evidence.

A completion is not performance.

And a satisfaction score is not authorization.

So your record has to retain the decision claim, the condition, the task format, the scoring rule, the administration version, the result and the limitation.

A qualified result against a stated operational threshold, under one standard administration, is a qualified result.

“Competent” is a much larger claim, and your database should be required to explain what it means by it.

The durable object is the defined performance a learner can demonstrate, under the conditions that matter for the decision.

The course is the learning experience.

The completion record is a participation wrapper.

The objective label is a design wrapper.

The score is evidence only to the extent that its task, conditions and interpretation support the decision.

Build excellent learning experiences.

Measure their usability.

Ask learners what worked.

Then decide, before the catalogue arrives, what has to be true before you write a capability claim into a system that will remember it for years.