Two quiet failures sink good courses: an outcome that’s taught but never measured, and evaluation that stops at “did you like it?” I map coherence before launch and measure real behavior after — then feed what I learn back into the next version.
Two things go wrong and nobody notices until it’s expensive. First, incoherence: an objective gets taught but never practiced, or assessed but never actually taught — the course looks complete and isn’t. Second, evaluation dies at the smile sheet: we learn that people enjoyed the training and never find out whether anyone did anything differently. Both are fixable, but only if you look on purpose.
Every outcome mapped against whether it’s taught, practiced, and assessed. The holes jump out — and get fixed before a learner ever hits them. This is constructive alignment made into a checklist.
| Outcome | Taught | Practiced | Assessed | Status |
|---|---|---|---|---|
| Name the behavior and its impact | ✓ | ✓ | ✓ | Aligned |
| Frame a behavior-based expectation | ✓ | — | ✓ | Gap: never practiced |
| Hold the standard under pushback | — | — | ✓ | Gap: assessed, never taught |
Row two needs a practice activity; row three is worse — we’re testing something we never taught. Both are cheap to fix now and painful to discover from learner failures later.
Kirkpatrick’s four levels. Most programs stop at Level 1 because it’s easy — but the value lives at Levels 3 and 4, where behavior and business results actually change. I design the measure for each level in from the start, not bolted on after.
Did they find it useful and relevant? The smile sheet — necessary, not sufficient.
Did they actually gain the skill? The aligned assessment answers this.
Are they doing it on the job weeks later? Observation, manager input, work samples.
Did the business metric move — the number that started the whole request?
The evaluation data isn’t a report card filed away — it’s the input to the next iteration. Where Level 2 is strong but Level 3 is weak, the gap is transfer, and I redesign practice, not content. That’s the SAM mindset: small, evidence-driven revisions instead of a monolithic rebuild every few years.
AI generates the alignment matrix from a course’s outcomes, activities, and assessments and flags the gaps automatically — the tedious coherence check, done in seconds. What stays mine is interpreting it: deciding whether a flagged gap is a real hole or an intentional choice, and which evaluation signal matters for this program. The tool surfaces; I judge.
Mapping alignment before launch means no learner is the first to find a gap. Evaluating past Level 1 means I can show whether the training changed behavior — the evidence stakeholders actually want. And feeding it back means each version is measurably better than the last.
Frameworks in play: constructive alignment / curriculum mapping, Kirkpatrick’s four levels, and the SAM iterative loop. The assessment that generates the Level 2 evidence: Proving the learning actually happened →