The SAM model is an instructional design process built on short cycles of design, prototyping, and review. SAM stands for Successive Approximation Model. Michael Allen and the team at Allen Interactions developed it as an alternative to linear course development. L&D teams use it when a program has to ship quickly and get better with each round of feedback.
This guide covers how SAM works, the difference between its two versions, how it compares with ADDIE, and how to apply it to training that ends in real practice.
What is the SAM model?
SAM treats every early version of a course as a draft meant to be tested. Allen Interactions puts it plainly in its overview of rapid instructional design with SAM. “Most simply put, SAM is an iterative prototyping process through which prototypes evolve into the final product.”
That idea changes how a project feels. Stakeholders react to something they can click through or try within the first days of the project, and the design improves after each review.
The model also shifts attention toward the learning experience and learner motivation. Content organization still matters, but the question the team keeps asking is whether the experience changes what people do. Michael Allen laid out the full approach in the book Leaving ADDIE for SAM, which is still the standard reference for teams adopting it.
SAM1 and SAM2: the two versions
There are two versions of the process, and the right one depends on project size.
SAM1 is the simpler version. It runs a loop of evaluate, design, and develop, and the loop usually repeats about two times before the product is complete. Small projects with a short content list and a small team fit SAM1 well.
SAM2 is built for larger projects with more content, more stakeholders, and more development work. It breaks the work into three phases and adds structure around reviews and releases. Most enterprise programs, such as a new hire curriculum or a certification for a whole sales org, end up on SAM2.
The three phases of SAM2
Preparation
The team gathers background information quickly. That includes the business goal, the audience, existing materials, and any constraints on delivery. Preparation is meant to be short. The point is to get enough context to make the first design session productive.
Iterative design
Iterative design opens with the Savvy Start. This is an intensive kickoff, usually one or two days, where the design team and key stakeholders brainstorm, sketch early designs, and test assumptions together. Allen Interactions describes the guiding question of the session as “why shouldn’t we do this?” instead of “what should we do?”
After the Savvy Start, the team builds rough prototypes and reviews them in small steps. The prototypes are intentionally unfinished. Allen Interactions advises working breadth before depth, so the team sketches every content area before polishing any single one.
Iterative development
Development starts with a Design Proof, a functional demonstration that tests both the design and the production process. Then the team produces three releases. The alpha is a complete version with all content in place. The beta fixes the problems found while validating the alpha. The gold release is the final version, and it ships once the beta performs as expected.
| Phase | What happens |
|---|---|
| 1. Preparation | Quickly gather the business goal, audience, existing materials and delivery constraints |
| 2. Savvy Start | Intensive kickoff where designers and stakeholders brainstorm, sketch and test assumptions |
| 3. Iterative design | Build rough prototypes and review them in small steps, breadth before depth |
| 4. Design Proof | A functional demonstration that tests both the design and the production process |
| 5. Alpha, beta, gold | A complete version, then fixes from validation, then the final release |
Teams rolling SAM out for the first time often lean on the Leaving ADDIE for SAM Field Guide from ATD, which adds checklists, templates, and job aids for each phase.
SAM model vs. ADDIE
The comparison comes up in almost every L&D team that considers SAM. The ADDIE model moves through analysis, design, development, implementation, and evaluation in order. Each phase produces a deliverable that feeds the next one, and learners usually see the program near the end.
SAM runs those activities in loops. Analysis, design, and development overlap, and stakeholders review working prototypes from the first week. Evaluation happens throughout the project, so problems surface while they are still cheap to fix.
Neither model is right for every project. ADDIE fits stable content with strict review requirements, such as regulated compliance courses where a documented sign-off at each phase matters. SAM fits content that changes often or depends on how learners respond. Product launches, new sales messaging, and conversation skills all fall into that second group. Teams that need speed without losing quality often pair SAM with rapid eLearning development practices.
Why SAM suits skills-based training
Conversation skills are hard to design on paper. A script that reads well in a document can fall apart the first time a rep tries it with a skeptical buyer or a frustrated customer. Successive approximation catches that early because people try the experience while it is still a prototype.
Practice-based programs also have more moving parts than a slide course. There is the scenario, the persona the learner talks to, the rubric used to score the conversation, and the feedback the learner sees afterward. Each of those pieces benefits from a few rounds of testing with real learners. Good learning journey design treats practice as the end point of a program, and SAM gives teams a way to tune that end point before launch.
AI roleplays make prototyping practice much faster. In Yoodli, an enablement or L&D team can build a roleplay around their own content, set the persona, and score responses against an org-defined rubric. A rough version is ready in time for the same review meeting where stakeholders discuss it. Yoodli Roleplays also support screen share, videos, and slides inside the conversation, so a prototype can include the materials a learner will actually use.
How to run a SAM project for a practice-based program
Here is how SAM maps to a program that ends in AI roleplay practice, such as onboarding or certification.
- Preparation. Collect the rubric your managers already use, the core product or process content, and examples of strong conversations. A prompt-based course creation workflow works best when these sources are ready before the first session.
- Savvy Start. Bring the program owner, a frontline manager, and one or two strong performers into the room. Build the first learning journey live. Yoodli Create generates a full learning journey from a single prompt, which gives the group something concrete to react to in the first hour.
- Iterative design. Run the rough roleplays with a small group of learners. Watch where they get stuck, where the persona feels unrealistic, and where the rubric scores do not match what a manager would say. Adjust and run them again.
- Design Proof. Finish one module end to end, including how learners will reach it. If the program lives in your LMS, test the delivery path now. Our guide to SCORM vs. LTI for AI training covers the tradeoffs.
- Alpha, beta, and gold. Release the full program to a pilot group, fix what they find, and then launch to everyone.
Scale raises the value of each round of testing. Google Cloud used Yoodli to certify 15,000+ employees on a new GTM pitch. When a program reaches that many people, a problem caught in a prototype saves far more time than a problem caught after launch.
How to measure each SAM iteration
Iteration only helps if the team knows what to change. Decide what each round of testing should tell you before it starts.
Early prototypes answer usability questions. Can learners find the practice, understand the instructions, and finish a roleplay without help? Later rounds answer skill questions. Are rubric scores rising across attempts, and do they line up with manager judgment of the same conversations?
Yoodli Measure shows progress across programs and against rubrics, which gives a SAM team real evidence to bring to each review. After launch, the Kirkpatrick model is a useful frame for evaluating reaction, learning, behavior, and results. Teams building a broader practice strategy can see how these pieces fit together on our learning and development solutions page.
FAQ
What does SAM stand for in instructional design?
SAM stands for Successive Approximation Model. It is an instructional design process developed by Michael Allen and Allen Interactions. The name describes how the work progresses: each prototype gets closer to the final product through repeated cycles of evaluation, design, and development. Stakeholders and learners review working versions along the way.
What is the difference between SAM and ADDIE?
ADDIE runs its five phases in sequence, while the SAM model repeats short design and development cycles. In ADDIE, learners usually see the program near the end. In SAM, stakeholders review working prototypes from the start, so feedback shapes the design early. ADDIE suits stable, heavily documented content, while SAM works better for programs that change often.
What is a Savvy Start?
A Savvy Start is the kickoff session that opens the iterative design phase of SAM2. It usually runs one or two days and brings the design team together with key stakeholders. The group brainstorms, sketches early designs, and tests assumptions quickly, so the project starts with rough prototypes everyone has already reacted to.
Is the SAM model an agile approach to training design?
Yes, SAM shares the core ideas of agile development. It works in short cycles, favors working prototypes over long documents, and builds in frequent review. Teams that already run agile sprints often find SAM easy to adopt, because the rhythm of build, review, and revise is the same.
Putting SAM to work
The SAM model gives L&D teams permission to show unfinished work early, and that is where most of its value comes from. Start with a short Savvy Start, get a rough version in front of real learners within days, and let their practice results decide what changes next. Programs built this way launch with fewer surprises and keep improving after the gold release.