An instructional design model is a repeatable way of running the work: what you find out first, what you build from it, how you test it, and how you decide it is done. There are a dozen worth knowing, and the choice between them comes down to four questions: how stable is the content, how fast do you need a first version in front of people, how much access do you have to the people who know the subject, and how expensive is a wrong first draft.
Most lists of "instructional design models" blur four different kinds of thing together. A learning theory says how people learn. A taxonomy classifies what you want them to learn. A set of principles says what good instruction contains. A process model says how to run the project. Bloom's Taxonomy appears on almost every list of models, and it is not one: it is a way of classifying objectives that every model uses. Sorting the field into those four kinds is the first thing this guide does, because a designer who can say which kind of thing they are reaching for is already ahead of most of the room.
First published January 13, 2024. Revised September 2026: reorganized into four categories, three process models added (Action Mapping, Agile and LLAMA, Design Thinking), the 24/7 Teach cycle added and set beside ADDIE and SAM, and originators named throughout.
Four kinds of thing that get called a model
| Kind | What it answers | Examples in this guide |
|---|---|---|
| Learning theory | How do people learn? | Behaviorism, cognitivism, constructivism, connectivism |
| Taxonomy | What kind and depth of learning am I asking for? | Bloom's Taxonomy, Webb's Depth of Knowledge |
| Principles and frameworks | What should the instruction itself contain? | Merrill's First Principles, Gagné's Nine Events, Keller's ARCS, Universal Design for Learning |
| Process model | How do I run the project from need to launch? | ADDIE, SAM, Dick and Carey, Kemp, Backward Design, Action Mapping, Rapid Prototyping, 4C/ID, Agile and LLAMA, Design Thinking |
You use one from each row on every project, whether or not you name them. The process model is the one you choose deliberately; the other three are the ones you should be able to defend.
Learning theories, briefly
These are covered in depth in Lesson 3 of our Instructional Design Foundations course. For the purpose of choosing a model, the summary is this. Behaviorism treats learning as a change in observable behavior shaped by practice and feedback; it is why every model has an "elicit performance and give feedback" step. Cognitivism treats learning as building and organizing mental structures; it is why cognitive load, chunking and worked examples matter. Constructivism treats learning as something the learner builds from experience, which is why problem-centered and project-based models exist. Connectivism (George Siemens) treats learning as forming and navigating networks of people and sources, and is the theory behind social and networked learning designs.
No process model belongs to one theory. ADDIE can produce a constructivist course; SAM can produce a behaviorist one. The theory shows up in what you build, not in which model you ran.
Taxonomies: Bloom's and Webb's, and why they are different ladders
Bloom's Taxonomy, in the revised form by Lorin Anderson and David Krathwohl, names the kind of thinking an objective asks for: Remember, Understand, Apply, Analyze, Evaluate, Create. It is indispensable for writing objectives and for checking that an assessment asks for the same kind of thinking the objective promised.
Webb's Depth of Knowledge (DOK), from Norman Webb, names how deep and complex the thinking has to go: recall (DOK 1), skill and concept (DOK 2), strategic thinking (DOK 3), extended thinking (DOK 4). The two are not the same ladder. A learner can "create" a poster that restates given facts (Create verb, DOK 1 to 2) or "create" an original argument from conflicting sources (Create verb, DOK 3 to 4). Rigor lives in what the task demands, not in the verb. Hiring managers who know the field will test whether you know that.
Principles and frameworks: what good instruction contains
Merrill's First Principles of Instruction (M. David Merrill). Learning is promoted when instruction is problem-centered, when it activates prior knowledge, when it demonstrates the skill, when learners apply it, and when they integrate it into their own world. Use it as a checklist against any lesson: which of the five is missing is usually why the lesson is weak.
Gagné's Nine Events of Instruction (Robert Gagné). Gain attention, state the objective, recall prior learning, present the content, guide the learner, elicit performance, give feedback, assess performance, enhance retention and transfer. It is a sequence for a single lesson or session and it is at its best for instructor-led or virtual instructor-led training, where the events map to what the facilitator actually does.
Keller's ARCS (John Keller). Attention, Relevance, Confidence, Satisfaction: the four conditions for motivation. Reach for it when the content is fine and the learners still are not doing it. Most "engagement" problems are a Relevance problem in disguise.
Universal Design for Learning (CAST). Multiple means of engagement, of representation, and of action and expression. It is a framework for designing so that the widest range of learners can use the instruction without retrofitting, and in 2026 it is a hiring expectation rather than a specialty.
Process models: how to run the work
ADDIE
Analysis, Design, Development, Implementation, Evaluation. The origin is commonly traced to instructional systems work at Florida State University for the United States Army in the 1970s; the acronym came later and has no single author, which is why nobody is cited for it. ADDIE is a phase model: each phase produces the inputs for the next, and evaluation closes the loop. It is the common language of the field, and every interview will assume you can speak it.
Best for: content that is stable, projects where the deliverable must be right the first time (compliance, safety, regulated training), and organizations that need a documented, auditable process.
Watch out for: treating it as strictly linear. Practitioners iterate inside every phase and loop back; a designer who runs ADDIE as a waterfall produces a course that is finished and wrong.
SAM (Successive Approximation Model)
Michael Allen's answer to ADDIE's tendency to defer testing until the end. SAM runs a short preparation phase (a "savvy start" that gathers stakeholders and sketches ideas fast), then iterative design and development, each cycle producing a testable prototype that is reviewed and refined: alpha, beta, gold. The core move is putting something in front of people early and often.
Best for: eLearning where the brief is still moving, stakeholders who cannot say what they want until they see it, and teams comfortable with review cycles.
Watch out for: iteration without a stopping rule. SAM needs a defined "gold" or it never lands.
Dick and Carey
Walter Dick, Lou Carey and James Carey's systems model, from The Systematic Design of Instruction. It is ADDIE with the joints exposed: identify the goal, analyze the instruction and the learners and context, write performance objectives, develop assessment instruments, develop the strategy, develop the materials, run formative evaluation, revise, run summative evaluation. Its distinguishing feature is that assessments are written before the strategy and materials, so alignment is designed in rather than checked afterwards.
Best for: large programs, formal education, anything where the relationship between objective, assessment and activity has to be demonstrable.
Watch out for: the overhead. It is the right amount of process for a curriculum and too much for a job aid.
Kemp
Gary Morrison, Steven Ross and Jerrold Kemp's model, from Designing Effective Instruction. Nine elements (instructional problems, learner characteristics, task analysis, objectives, content sequencing, strategies, message design, development, evaluation) arranged in an oval rather than a line, because you may start anywhere and revisit anything. It is the model that treats the design as a whole rather than a sequence.
Best for: projects with diverse learners and shifting constraints, and designers who already know the phases and want permission to move between them.
Watch out for: the same freedom becoming drift. Kemp assumes a designer who can hold the whole picture.
Backward Design (Understanding by Design)
Grant Wiggins and Jay McTighe. Three stages: identify desired results, determine acceptable evidence, plan learning experiences. The order is the point. You decide what the learner will be able to do, then what would prove it, and only then what you will teach. Designers who run it stop producing activities that nobody can connect to an outcome.
Best for: curriculum, K-12 and higher education units, and any program that will be judged on measurable outcomes.
Watch out for: assuming the "desired results" are handed to you. In corporate work, stage one is the analysis you did not get to skip.
Action Mapping
Cathy Moore's method, and the most widely used approach in corporate learning that most academic lists leave out. Start with a measurable business goal. Identify what people need to do differently to reach it. Design practice activities that rehearse those actions in realistic conditions. Add only the information required to do the activities. The model's signature move is the question it asks first: is this a training problem at all, or is the gap in tools, process, incentives or environment?
Best for: performance problems in organizations, stakeholders who ask for "a course on X," and any project where the temptation is to dump content.
Watch out for: the honest answer being "no training needed." Action Mapping will tell a stakeholder that, and the designer has to be willing to deliver the message.
Rapid Prototyping
Less a full model than a stance that several models adopt, described for instructional design by Steven Tripp and Barbara Bichelmeyer. Build a rough working version of the instruction early, test it with real learners, revise, repeat. The prototype is the analysis: what learners do with it tells you what the requirements are.
Best for: novel content, short timelines, and situations where nobody, including the client, can specify the requirements up front.
Watch out for: a prototype so polished that people mistake it for the product, and a prototype so rough that the feedback is about the roughness.
4C/ID (Four-Component Instructional Design)
Jeroen van Merriënboer's model for complex learning, developed with Paul Kirschner in Ten Steps to Complex Learning. Four components: learning tasks (whole, authentic tasks ordered from simple to complex), supportive information (the theory behind the tasks), procedural information (the just-in-time how-to), and part-task practice (drilling the routine pieces). It exists because complex skills do not decompose into a list of objectives without losing the thing that makes them complex.
Best for: professional and technical education where the job is an integrated skill: clinical work, engineering, aviation, software.
Watch out for: applying it to simple content. If the whole task fits on one screen, 4C/ID is scaffolding for a building that is not there.
Agile and LLAMA
Agile came from software and brought sprints, standups, backlogs and working increments. Megan Torrance's LLAMA (Lot Like Agile Management Approach) adapts it for learning teams: short iterations, a prioritized backlog of learning outcomes, a working deliverable at the end of each iteration, and stakeholder review built into the cadence. It is the model that treats "the requirements changed" as normal rather than as failure.
Best for: teams delivering continuously, products that update, and organizations already running agile elsewhere.
Watch out for: ceremony replacing design. A standup is not a needs analysis.
Design Thinking
Popularized by IDEO and Stanford's d.school: empathize, define, ideate, prototype, test. It is a human-centered problem-solving method rather than an instructional model, and designers borrow it for the front end, where it does two things well: it puts the learner's actual experience at the center of the analysis, and it generates several candidate solutions before committing to one.
Best for: the front end of an ambiguous project, and for learning experience design where the "learner journey" is as much the product as the content.
Watch out for: stopping at ideas. Design Thinking hands you a well-framed problem and a prototype; something else has to build and evaluate the instruction.
The model we run at 24/7 Teach
Every project in the Instructional Design Bootcamp runs eight stages: Plan, Design, Develop, Demo, Feedback, Reflect and document, Act, Launch. It is an iterative process model, and once you have read the list above you can place it. Here it is beside the two everyone knows.
| Our stage | ADDIE | SAM | What we do differently |
|---|---|---|---|
| Plan | Analysis and Design | Preparation (savvy start) | Analysis is bounded by five questions and a reversibility rule (minimum viable analysis), and the fifth question, what needs to be true for this to work, writes the checklist the Demo is judged against |
| Design | Design | Iterative design | Alignment is the criterion: does what you are making serve what you found |
| Develop | Development | Iterative development | Build to a minimum viable product that clears the results you set at Plan, and stop there |
| Demo | (Implementation, in part) | Prototype review | The first real test of whether the product cleared the bar. You do not certify your own product |
| Feedback | Evaluation (formative) | Review | Judged through four lenses: learning science, contextual relevance, instructional integrity, implementation feasibility |
| Reflect and document | (rarely explicit) | (rarely explicit) | Written so that someone who was not there can make a decision from your notes |
| Act | Revision | Next iteration | The changes you named actually appear in the next round |
| Launch | Implementation | Gold | Launch is the deadline, not the finish. The line is drawn, whatever is unfinished becomes the next iteration |
Where it borrows: the iterate-and-review spine is SAM's, and the insistence that analysis comes first and evaluation closes the loop is ADDIE's. Where it differs: Launch is a deadline that forces a decision rather than a state of completion; the three stages before it (Feedback, Reflect and document, Act) are the ones most models leave implicit and we make mandatory, because they are where a finished project turns into a higher starting point; and the analysis is deliberately minimal, sized by how reversible the decision is, because a first draft is now cheap and trusting one is expensive.
Why this and not one of the others: because the deliverable keeps changing (two years ago a course, today a system prompt or an agent) and the four things we assess on every project, analysis, alignment, evaluation and iteration, do not. A process model whose stages are judgment checkpoints rather than production phases survives the deliverable changing underneath it.
Choosing a model
| Situation | Reach for | Why |
|---|---|---|
| Compliance or safety content that must be right the first time and auditable | ADDIE, or Dick and Carey for a large program | Documented phases, alignment designed in, evaluation on record |
| The stakeholder asked for "a course on X" and you suspect the gap is not knowledge | Action Mapping | It asks first whether training is the answer, and designs practice rather than content |
| The brief is moving and the client cannot describe what they want until they see it | SAM or Rapid Prototyping | Something testable in front of people early; the prototype is the analysis |
| A two-week deadline and a team already working in sprints | Agile and LLAMA | Working increments and a backlog match how the organization already moves |
| A K-12 unit or a higher-education course judged on outcomes | Backward Design | Results, then evidence, then activities; nothing gets built that cannot be connected to an outcome |
| A complex professional skill (clinical, technical, integrated) | 4C/ID | Whole tasks from simple to complex, with theory and how-to attached to the task rather than front-loaded |
| An ambiguous problem where you do not yet know who the learner is or what they experience | Design Thinking for the front end, then one of the above | Empathy and problem framing before commitment |
| Diverse learners, shifting constraints, an experienced designer | Kemp | Enter anywhere, revisit anything, hold the whole design at once |
| A single lesson or facilitated session inside a larger design | Gagné's Nine Events, with Merrill's principles as the check | A sequence for one session, and a checklist for whether it teaches |
Two rules underneath the table. First, the model is a tool for the project, not an identity: a good designer runs ADDIE on Monday and Action Mapping on Wednesday. Second, when you are unsure, ask how expensive a wrong first draft is. If it is cheap, prototype (SAM, Rapid Prototyping, Agile). If it is expensive, analyze and align first (ADDIE, Dick and Carey, Backward Design).
Sources and originators
The names below are the originators of each idea and the works they are best known by. Publication years are deliberately not given here; add them from the primary source, not from memory or a secondary summary, before citing this guide in a course or a portfolio.
- Bloom's Taxonomy: Benjamin Bloom and colleagues; revised by Lorin Anderson and David Krathwohl, A Taxonomy for Learning, Teaching, and Assessing
- Depth of Knowledge: Norman Webb
- First Principles of Instruction: M. David Merrill
- Nine Events of Instruction: Robert Gagné, The Conditions of Learning
- ARCS: John Keller
- Universal Design for Learning: CAST
- ADDIE: instructional systems development work at Florida State University for the U.S. Army; no single author
- SAM: Michael Allen, Leaving ADDIE for SAM
- Dick and Carey: Walter Dick, Lou Carey and James Carey, The Systematic Design of Instruction
- Kemp: Gary Morrison, Steven Ross and Jerrold Kemp, Designing Effective Instruction
- Backward Design: Grant Wiggins and Jay McTighe, Understanding by Design
- Action Mapping: Cathy Moore, Map It
- Rapid Prototyping in instructional design: Steven Tripp and Barbara Bichelmeyer
- 4C/ID: Jeroen van Merriënboer; with Paul Kirschner, Ten Steps to Complex Learning
- LLAMA: Megan Torrance, Agile for Instructional Designers
- Design Thinking: IDEO; Stanford d.school
- Connectivism: George Siemens
Discussion question
Take a brief you are working on now, or this one: a regional hospital network wants "a course" for new charge nurses because handoffs between shifts are going wrong, the nurses are experienced clinicians with no time for a long program, and the network has never run one before.
Which model would you run, and why? Name one alternative and say specifically why it fits this brief less well. In the Instructional Design Foundations course this is Lesson 7's constructed response, and it is read by a mentor rather than posted to a comment thread.