Instructional design models: a practical reference

Instructional design has a long history of producing frameworks, and most practicing designers have opinions about which ones are most useful. The frameworks share a basic premise: that designing training well is a structured discipline rather than an instinctive one, and that the structure pays off in better training outcomes.

This guide walks through the major instructional design models you’ll come across in the field. For each, we cover what the model actually is, what kind of work it’s best suited to, and where it tends to fall short. By the end, you’ll have a working sense of when to reach for which framework and how the major options compare.

For the broader question of what instructional design is and what an instructional designer does, our guide to instructional design covers the discipline itself. This article goes deeper, specifically on the models.

What instructional design models are (and aren’t)

An instructional design model is a structured approach for designing training. Some models describe a process (the steps you go through to get from a performance gap to a finished course). Others describe principles (what good instruction looks like at the level of a single learning event). Still others describe taxonomies (frameworks for classifying objectives or measuring outcomes).

What the models share is a commitment to making design decisions deliberately rather than by instinct. What learners need to do, in what order, with what kind of practice, and how you’ll know it worked. Without a model, the choices still get made, but they’re often inconsistent across projects and hard to defend when stakeholders push back.

A few things to keep in mind before getting into specific models:

• Models are scaffolding. They give structure during the work, especially on complex projects. They don’t replace judgment.
• Most projects use more than one. A typical project might follow ADDIE as its overall process, write objectives using Bloom’s Taxonomy, structure individual lessons using Gagné’s Nine Events, and plan post-launch measurement using Kirkpatrick’s framework. The models complement each other.
• Real-world practice is messier than any model. Subject matter expert (SME) interviews don’t produce clean inputs. Stakeholders change requirements mid-project. Source material is incomplete. The models help you stay anchored when the project gets complicated, but they don’t make the complication go away.

Instructional design models at a glance

The table below summarizes the eight models covered in this guide. Each is expanded in detail in the sections that follow.

ADDIE: Analyze, Design, Develop, Implement, Evaluate

ADDIE is the most widely taught instructional design model and the one most stakeholders will recognize by name. It describes a five-phase, sequential approach: analyze the performance gap, design the solution, develop the materials, implement the training, and evaluate the results.

The five phases:

1. Analyze. Define the problem the training is supposed to solve. Identify the audience, their current capability, and the gap between what they can do now and what they need to be able to do. Determine whether training is even the right intervention.
2. Design. Translate the analysis into a design plan. Write learning objectives. Decide on modality (eLearning, instructor-led training, blended). Map the structure of the training, the major activities, and the assessment strategy. The output is usually a design document or treatment, before any storyboard work begins.
3. Develop. Build the actual materials. For an eLearning course, this is where the storyboard gets fleshed out into a course in an authoring tool. For instructor-led training (ILT), this is where the facilitator guide and participant materials get produced.
4. Implement. Deploy the training. Get learners into the course, run the live sessions, manage the rollout.
5. Evaluate. Look at how the training performed against the objectives set in the analysis phase. Adjust based on what worked and what didn’t.

ADDIE works well when project requirements are clear at the start, when the timeline allows for sequential phasing, when the project is large enough that the structure prevents corners from being cut, and when stakeholders need predictable phase gates for review and approval.

It struggles when requirements are likely to evolve mid-project (you commit to a design before discovering things during development that would change it), when the project is too small or too urgent to absorb the full sequential pass, and when prototyping early would help validate design choices.

ADDIE is the most widely taught framework in the field, which means most instructional designers can work fluently within it and most stakeholders recognize the phases.

SAM: Successive Approximation Model

SAM is an iterative alternative to ADDIE developed by Michael Allen specifically to address the problem of design assumptions getting locked in too early. Instead of one long sequential pass, SAM has the team build small prototypes early, test them, refine, and repeat.

The model has three phases:

1. Preparation. Brief, focused. Gather what’s known about the audience, content, and constraints. The phase ends with a single half-day “Savvy Start” workshop with stakeholders to align on the goals.
2. Iterative design. A repeating loop of design, prototype, and review. The team produces small samples (a single screen, a brief scenario, an interaction prototype), reviews them with stakeholders and ideally with target learners, and refines based on feedback.
3. Iterative development. The same loop applied to development. Build pieces, review pieces, refine pieces, build more.

SAM fits projects where the design choices benefit from learner feedback before being locked in. Soft skills, behavior change, anywhere the right approach isn’t obvious upfront. It also fits organizations comfortable with iteration and willing to participate actively rather than approving finished deliverables. The tradeoff is that stakeholders who need predictable phase-gate approvals struggle with the iterative rhythm, and well-defined stable content can leave teams iterating just to iterate.

SAM and ADDIE are often presented as opposites, but in practice many instructional designers blend them. A hybrid approach uses ADDIE’s overall structure for stakeholder communication and project planning, while building iterative prototyping into the design phase for parts of the project where the design direction isn’t clear yet.

Bloom’s Taxonomy

Bloom’s Taxonomy, originally published in 1956 and revised in 2001, is a framework for classifying learning objectives by the type of thinking they require. It’s not a process model. It’s a tool for one specific part of design work: writing objectives.

The revised version organizes thinking skills into six levels, from simpler to more complex:

1. Remember. Recall facts and basic concepts. (Define, list, identify.)
2. Understand. Explain ideas or concepts. (Describe, summarize, classify.)
3. Apply. Use information in new situations. (Execute, implement, demonstrate.)
4. Analyze. Draw connections among ideas. (Differentiate, compare, examine.)
5. Evaluate. Justify a decision or position. (Critique, judge, defend.)
6. Create. Produce new or original work. (Design, construct, develop.)

The practical use is keeping learning objectives honest. “Understand the company’s data privacy policy” is at level 2 (and probably not what you actually want learners to be able to do). “Identify a privacy risk in a customer interaction and respond appropriately” is closer to levels 3–4, and produces a much different course.

Bloom’s helps any time you’re writing learning objectives. It’s especially useful when stakeholders default to “understand” or “be aware of” objectives that won’t translate into observable behavior change. The taxonomy gives you a vocabulary for pushing back constructively.

Where Bloom’s falls short is as a complete design framework. It’s a tool, not a process. It doesn’t tell you how to structure a course, how to design practice, or how to assess. Pair it with a process model.

Gagné’s Nine Events of Instruction

Robert Gagné’s Nine Events is a framework for structuring a single learning event so it actually teaches. It’s not a project-level model; it’s a lesson-level checklist. The nine events:

1. Gain attention. Open in a way that engages the learner and orients them to the topic.
2. Inform learners of the objective. State what they’ll be able to do at the end.
3. Stimulate recall of prior learning. Connect the new material to what learners already know.
4. Present the content. Deliver the new information.
5. Provide learning guidance. Help learners process and integrate the new material.
6. Elicit performance. Have learners practice or demonstrate the new skill.
7. Provide feedback. Tell them how they did and why.
8. Assess performance. Check whether the learning objective was met.
9. Enhance retention and transfer. Help learners apply the new learning beyond the course.

Gagné’s helps as a quality check on individual lessons or modules. If a learning event is missing several of the nine events, it’s likely to underperform regardless of how polished the production is.

It falls short as a project-level framework. The events tell you what each lesson should include, but they don’t help you decide what lessons should exist or how they should sequence. Use Gagné’s inside a larger framework like ADDIE.

Dick and Carey’s Systems Approach

The Dick and Carey model treats instructional design as a system in which all components are interconnected. The instructor, the learner, the materials, the environment, and the content all interact, and design choices in one part affect all the others. The model has ten interrelated steps that emphasize the relationships between components rather than just the sequence of work:

1. Identify instructional goals
2. Conduct instructional analysis
3. Analyze learners and contexts
4. Write performance objectives
5. Develop assessment instruments
6. Develop instructional strategy
7. Develop instructional materials
8. Design and conduct formative evaluation
9. Revise instruction
10. Design and conduct summative evaluation

Dick and Carey works well for large, complex projects where the interactions between elements are non-trivial. The model is heavily used in academic and government contexts where rigor is required. For most corporate projects, the level of formal rigor is overkill. The underlying systems thinking is valuable, but the formal model can slow projects down without adding proportional benefit.

For most practical work, the principle to take from Dick and Carey is more useful than the model itself. Design choices interact. Changing one element (the audience, the modality, the assessment) usually requires revisiting others.

Merrill’s First Principles of Instruction

David Merrill’s First Principles, articulated in 2002, is a synthesis of what instructional design research consistently shows works. It’s not a process model; it’s a set of principles that effective instruction has in common, regardless of which process produced it. The five principles:

1. Problem-centered. Learning is most effective when learners are engaged in solving real-world problems, not abstract exercises.
2. Activation. Learning happens when learners’ existing knowledge is activated as a foundation for new knowledge.
3. Demonstration. New knowledge needs to be demonstrated, not just stated.
4. Application. Learners need a chance to apply the new knowledge themselves.
5. Integration. New knowledge has to be integrated into the learner’s existing world (their job, their life) for it to stick.

Merrill’s works as a checklist for whether a design (regardless of which framework produced it) is doing what effective instruction does. If a course doesn’t activate prior knowledge, doesn’t demonstrate the new material in context, and doesn’t give learners a chance to apply it, the principles say it’ll struggle to work. That holds whether the course was built using ADDIE, SAM, or anything else. The principles describe what good instruction looks like rather than providing a workflow, so they sit alongside a process model.

Kirkpatrick’s Four Levels of Evaluation

Kirkpatrick’s model isn’t a design framework at all. It’s an evaluation framework: a way of measuring whether training is working. It’s included here because most instructional design models stop short on evaluation, and any complete approach to designing training has to include a way of knowing whether the training did its job.

The four levels:

1. Reaction. Did learners enjoy the training? Did they think it was relevant? (Measured by post-training surveys.)
2. Learning. Did learners actually acquire the knowledge or skills? (Measured by assessments, demonstrations, knowledge checks.)
3. Behavior. Are learners doing the job differently afterward? (Measured by observation, performance data, manager feedback.)
4. Results. Did the changed behavior produce the business outcome the training was supposed to drive? (Measured by metrics tied to the original problem.)

Levels 3 and 4 matter most because most organizations measure level 1 (satisfaction surveys) and sometimes level 2 (post-course quizzes), then stop. The training-to-performance gap exists in the space between levels 2 and 3. A course can score 4.7/5 on satisfaction (level 1), pass everyone on the assessment (level 2), and produce no observable behavior change on the job (level 3 failure). Without measuring at level 3, the failure is invisible.

Where Kirkpatrick’s falls short is in measurement cost: the higher levels are harder and more expensive to measure. Level 4 in particular requires data integration with business systems and patience to wait for results to surface. Many organizations find level 3 measurement the right balance of cost and signal.

Action Mapping

Cathy Moore’s Action Mapping isn’t always classified as a major instructional design model, but it’s worth knowing about because it represents a focused alternative to the more comprehensive frameworks. The premise: the goal of training isn’t to transfer information; it’s to change behavior. Everything else in a course exists to support that goal. If a piece of content doesn’t directly help learners do something differently, cut it.

The Action Mapping process:

1. Identify the business goal.
2. Identify the actions people need to take to achieve the goal.
3. Identify what’s preventing those actions (often this isn’t lack of knowledge, but process problems, tools problems, or motivation problems).
4. For the gaps where training is the right intervention, design realistic practice activities.
5. Add only the information learners need to perform those activities.

Action Mapping helps when stakeholders are pushing for content-heavy courses (“we need to cover everything in the policy”) and the training is at risk of bloating into reference material. The framework gives you a vocabulary for cutting.

It falls short for purely informational training where there’s no specific behavior to change. Compliance awareness, general onboarding orientation, and similar content has less for the framework to add.

How to choose between models

Most instructional designers don’t pick one model and use it for everything. The skilled practice is matching the model to the project. A few rough heuristics:

• For most corporate training projects, ADDIE or a SAM-ADDIE hybrid is the right backbone. ADDIE for projects with clear requirements and traditional review gates, SAM (or hybrid) for projects where iteration would help.
• Bloom’s belongs in every project, used to write learning objectives that pass the “can we observe whether the learner did this?” test.
• Gagné’s Nine Events is worth running through any individual lesson or module that’s underperforming or that you’re worried about. It surfaces missing structure quickly.
• Kirkpatrick’s belongs in the evaluation conversation from the start of every project. Even if you can’t realistically measure at level 4, knowing what you’re aiming for shapes the design from day one.
• Merrill’s First Principles is a useful sanity check: does the design activate prior knowledge, demonstrate, give learners a chance to apply, and integrate? If not, fix that before launching.

For a closer look at the design process itself, including how a curriculum gets built across multiple courses, see how to design a curriculum.

How Custom Learning chooses between models

Neovation Custom Learning is your full-service, instant L&D capacity, providing expert instructional designers, eLearning developers, and project managers who turn your organization’s raw expertise into interactive, scalable custom training. In practice, that means we use ADDIE, SAM, and hybrid versions of both, depending on the project. For most corporate training work, Custom Learning starts with an ADDIE-style structure and brings in SAM-style iteration during the design phase for parts where the design direction isn’t obvious yet. We use Bloom’s Taxonomy in objective-writing, Gagné’s events at the lesson level, and design with Kirkpatrick’s framework in mind so that what gets measured after launch matches what the training was designed to do.

Picking one model and forcing everything through it produces worse outcomes than matching the framework to the work. A high-stakes compliance course needs different structure than a soft-skills behavioral change program. The model should serve the project. Request a quote when you’d like to discuss how we’d approach a specific project, or browse our case studies to see what these frameworks look like in client engagements.