Start with a real problem, end with real application. Courses built on Merrill feel different — learners do more, apply more.
Use for: Applied, problem-centred learning experiences.
Framework · Evaluation
Kirkpatrick
4-level training evaluation framework
L1 ReactionL2 LearningL3 BehaviorL4 Results
Not a design model — an evaluation framework. Everyone cites it. Very few orgs measure beyond L1 (smile sheets).
Use for: Measuring training effectiveness after delivery.
The Core Mantras
ADDIE
"A blueprint before you build."
Without Analysis, you're building a solution to an undefined problem. The plan is the work.
SAM
"Show, don't tell — then show again."
Rough prototype in front of SMEs beats a polished storyboard they'll half-read. Show early, fix cheap.
4C/ID
"Whole tasks, not just parts."
If the job demands the whole thing under pressure, train the whole thing under pressure — not the steps in isolation.
Bloom's
"What will they DO with this knowledge?"
Forces you to name the verb. If you can't, the objective isn't measurable — and you can't design a test for it.
Gagné's 9
"Did they practice before you told them how they did?"
Content → quiz is testing, not training. Learners must attempt the skill before they see feedback.
Merrill's
"Start with a real problem, end with real application."
If the learner never does anything with the content, it's a presentation — not instruction.
Kirkpatrick
"Agree on what success looks like before you build."
Work backwards from the real-world result. Completions prove nothing. Behaviour change does.
Which one do I reach for?
"What's my design process for this project?"
→ ADDIE if waterfall · SAM if iterative
"Is the skill genuinely complex — whole-task, high-stakes?"
→ 4C/ID. Otherwise, ADDIE or SAM is enough.
"Are my learning objectives measurable?"
→ Check against Bloom's — fix the verb if it's vague.
"Does my course sequence feel thin or rushed?"
→ Map it against Gagné's 9 — find the missing event.
"Are learners doing enough, or just watching?"
→ Apply Merrill's — put the problem first.
"How will I prove this training worked?"
→ Define Kirkpatrick levels with the stakeholder before you build.
ID Cheat Sheet · Interview Edition
7 tools. Know them cold.
3 give you a design process. 4 give you a lens to think through. Click any card — the same example runs through all of them.
Running example →Sales objection-handling course · 50 new B2B reps · ships next month
The 3 True ID Models — give you a design process
ID Model · Sequential Process
ADDIE
The classic linear design model
AnalyzeDesignDevelopImplementEvaluate
"Blueprint before you build."
Tap to expand ↓
In our example
Before touching a slide, Analyze first. Interview the sales manager. Are reps failing on objections — or is it actually product knowledge? You might be building the wrong course entirely.
Use when
Clear scope. Waterfall delivery. Defined stakeholder handoffs. Client hands you a brief and expects a finished product.
🏗️
Memory device
The architect analogy. No architect pours concrete before drawing plans. ADDIE is your drawing phase. Skipping Analyze = building the wrong building. The A is the most important letter.
Not for
Fast-moving projects where requirements will shift. If SMEs keep changing direction, switch to SAM.
ID Model · Agile / Iterative
SAM
Successive Approximation Model
PreparationIterative DesignIterative Dev
"Show rough early. Fix cheap."
Tap to expand ↓
In our example
By Day 5, show 3 sales reps a rough scenario — not a storyboard. A real prototype. Their reaction reshapes everything before you've built 90% of the course.
Use when
SMEs are available and willing to give iterative feedback. You have room to pivot after early reactions.
🔄
Memory device
SAM = Show, Adjust, Make. Not show-when-perfect. Show rough, get feedback, adjust, make the next version. The value is in the loop — not the final launch.
Not for
Projects where SMEs are unavailable or where feedback turnaround is too slow to make iteration worthwhile.
ID Model · Complex Skills
4C/ID
Four-Component Instructional Design
Learning TasksSupportive InfoProcedural InfoPart-task Practice
"Train the whole task under real pressure."
Tap to expand ↓
In our example
Objection handling is a whole-task skill. Reps must handle a live objection under pressure — not recite the LAER steps. The course puts them in full role-plays from session one.
Use when
Complex job where training parts in isolation fails: clinical, technical, leadership, legal. The whole task must be trained — not just the components.
🏛️
Memory device
A table with 4 legs. Remove any one leg and the whole thing collapses. You can't skip Supportive Info or Part-task Practice and expect learners to perform the real skill.
Not for
Most corporate eLearning. It's the most demanding of the three models — overkill for awareness training or simple process walkthroughs.
The 4 Frameworks, Taxonomies & Principles — give you a lens
Taxonomy · Cognitive Levels
Bloom's
6-level classification of learning
RememberApplyCreate
"Name the verb — then write the objective."
Tap to expand ↓
In our example
"Understand objection handling" ✗ — too vague, untestable. "Handle the top 5 objections using LAER in a live role-play" ✓ — Apply level. Now you can design an actual assessment.
Use for
Writing measurable objectives. Pick the verb first, from the right level. "Understand" is never a Bloom's verb — it's a hedge.
🪜
Memory device
Start from the top, work down. Ask "what should they DO with this?" first — that lands you at Apply or higher. Then design everything below it to support that goal. Most IDs start at Remember and never climb.
Common mistake
Bloom's is NOT a design process. It's a vocabulary for writing objectives. Don't use it to sequence a course.
Principles · Instruction
Gagné's 9
9 Events of Instruction
1–5: Setup6: Practice ← skipped7–9: Confirm
"Did they attempt it before seeing the answer?"
Tap to expand ↓
In our example
Event 6 (Elicit Performance): before showing reps the ideal objection response, make them handle it first. Their attempt reveals the gap. Most designers skip from Event 5 straight to Event 7.
Use for
Auditing a course that feels incomplete. Map your existing sequence against the 9 events — missing events explain why learners feel unsupported.
🎯
Memory device
The missing step is Event 6. Most designs go Content (4) → Guidance (5) → Feedback (7). They skip the attempt. Gagné's core insight: learners must perform before receiving feedback — not after watching content.
Common mistake
Treating it as a rigid 9-step checklist. It's a sequencing lens. Events can combine, split, or compress — what matters is the logic, not the count.
"Start with the real problem. End with real application."
Tap to expand ↓
In our example
Open with a recording of a real failed objection call — no slides. Let reps feel the problem (Activation: "what did you try last time?"). Then demonstrate the expert model. Problem always comes first.
Use for
Applied, problem-centred learning. If learners do more than watch in your course, you're probably already applying Merrill's — just not naming it.
🥪
Memory device
P-A-D-A-I. Problem, Activation, Demonstration, Application, Integration. The key is the first and last: if there's no real problem at the start and no real application at the end, it's a presentation — not instruction.
Common mistake
Treating Demonstration (P2) as lecturing. Merrill means showing the expert performing the task — not explaining slides about it.
Framework · Evaluation
Kirkpatrick
4-Level Training Evaluation
L1 ReactionL3 BehaviorL4 Results
"Agree on what success looks like before you build."
Tap to expand ↓
In our example
L1: did reps find it useful? L2: can they name LAER? L3: did call quality scores rise in month 1? L4: did close rates move? Only L3 and L4 prove it actually worked.
Use for
Pre-build stakeholder conversations. Define L3 and L4 metrics before you design anything — not after. If you can't measure it, you can't prove ROI.
📏
Memory device
Like it? Learn it? Do it? Did it matter? L1 → L4. Work backwards: define L4 first (what business result matters?), then figure out what L3 behaviour change would produce that result. Most orgs never leave L1.
Common mistake
Kirkpatrick is NOT a design model. It tells you how to measure impact — not how to build better learning. Use it during scoping, not development.
Which one do I reach for?
"What's my design process for this project?"
→ ADDIE if waterfall · SAM if iterative
"Is the skill complex — whole-task, high-stakes?"
→ 4C/ID. Otherwise ADDIE or SAM is enough.
"Are my learning objectives actually measurable?"
→ Check against Bloom's — fix the verb if it's vague.
"Does my course feel thin or rushed?"
→ Map it against Gagné's 9 — find the missing event.
"Are learners just watching, not doing?"
→ Apply Merrill's — put the real problem first.
"How will I prove this training worked?"
→ Define Kirkpatrick L3/L4 with the stakeholder before you build.