Five-Phase ID Framework

Parker Lee · Instructional Design

Discover

Verify the problem before designing any solution

Questions I won't skip

01What behavior is missing or wrong — observable, on the job?
02What happens when someone gets it wrong? What's the real cost?
03Who decided this is a training problem — and did anyone push back?
04Can you show me what expert performance looks like, or introduce me to someone who does it well?
05Is the gap in knowledge, skill, or environment — process, tools, management clarity?
06What would change if training worked perfectly? Would you see it in a metric?
07What do learners already know? What's their context coming in?
08What's the real timeline, and is it realistic for the scope being described?

SME action-mapping script

→Ask what experts do, not what they know
→"Walk me through the last time this went wrong — specifically."
→"What would a great performer do at this decision point that a novice wouldn't?"
→Push vague answers: "What does that look like on the floor?"

AI + this phase

✓ Can help

Draft SME interview questions
Pressure-test the brief
Flag non-training problems

✕ Can't replace

The actual SME conversation
Org and culture context
Training vs. process judgment

Design

Pick the right modality. Structure before building.

Modality decision guide

Rise 360Async, self-paced knowledge transfer — onboarding, compliance, product content. Use when completion tracking matters.

StorylineBranching scenarios, scored simulations, complex interactions. Use when skill practice and realistic consequence matter.

HTML / codeLightweight reference tools, embeddable job aids, sales enablement. Faster, more accessible, easier to update.

BlendedComplex judgment calls, culture-shaping, sensitive topics. If the live debrief is the point, don't async your way out of it.

Job aidPure reference — process steps, checklists, decision trees. Never needs a quiz. If they need to memorize it, it's not a job aid.

Scenario structure

→Realistic trigger — not "imagine you are a rep..."
→Choices that aren't obviously right or wrong
→Realistic consequences — not just "try again"
→Debrief that explains, not just reveals the answer

AI + this phase

✓ Can help

Scaffold scenario text fast
Draft and critique objectives
Suggest modality given constraints

✕ Can't replace

Knowing your learners' context
Culture that makes scenarios real
Domain accuracy — SME review required

Build

Pre-build discipline prevents expensive rework

Pre-build

→Storyboard before opening Storyline or Rise
→Content QA with SME before build — not after
→Lock final brand assets before building
→Scope in writing — verbal agreements disappear

Build cadence

→Pilot section first — QA one piece before building everything
→Narration: script → record temp → record final
→In Storyline: master slides before content slides
→Test every branch, especially failure paths

Polish checklist

→Read every word aloud — awkward phrasing reveals itself
→Test on the device learners actually use
→One SME review pass on final content only
→Check accessibility: tab order, alt text, contrast

AI + this phase

✓ Can help

Draft narration scripts
Flag corporate-sounding language
Suggest alternatives for awkward phrasing

✕ Can't replace

Domain accuracy QA
Never skip the SME content review

Deliver

Manager prep is the most skipped, most impactful step

Launch sequence

1Pilot with 3–5 real learners — not leadership, not L&D. Track time-on-task and confusion points.

2Adjust based on pilot. Completion time, drop-offs, and qualitative quotes beat survey scores.

3Manager brief before learner launch: what's coming, why it matters, what to reinforce.

4Launch email: one sentence on why this matters to them, not to the company.

5Completion deadline: visible, specific, reasonable. "By end of quarter" is not a deadline.

Manager one-pager

→What the training covers (2 sentences max)
→3 questions to ask in a 1:1 after completion
→What good performance looks like in 30 days

AI + this phase

✓ Can help

Draft launch communications
Write the manager one-pager
Summarize pilot readout data

✕ Can't replace

The manager conversation
Advocacy from the right sponsor

Measure

Evaluation is a design constraint, not an afterthought

Kirkpatrick L1–L4

L1 · ReactionSurvey at end, 3–5 questions. Useful for spotting confusion. Not evidence of effectiveness.

L2 · LearningKnowledge checks in-course. If you're not adjusting design based on failure rates, you're not using L2 data.

L3 · BehaviorObservation or manager report 30–90 days post. Requires manager pre-commitment before launch.

L4 · ResultsBusiness metric change. Only promise if you have a clear, defensible causal path — attribution is almost never clean.

When L3/L4 attribution is credible

→You have a pre/post baseline on the metric
→You control for confounds: market, leadership, headcount changes
→Training was the only meaningful intervention in the window

AI + this phase

✓ Can help

Draft L1 survey questions
Suggest L3 behavioral indicators
Design evaluation canvas with sponsor

✕ Can't replace

Collecting the actual data
Causal attribution — that's a business call