Successive Approximation Model (SAM)

Design Process · All design processes
Evidence · 6 claims (4 for, 1 mixed, 1 against) · 12 studies (4 quant-synthesis, 3 review, 2 causal, 2 design, 1 theoretical), q1–q4 · 4 of 12 report an effect size · 3 claims rest on one study

Description

SAM is an iterative design process for courses, formulated by Michael Allen as an explicit alternative to running ADDIE as a waterfall. Its argument is empirical rather than philosophical: the phases of a systematic process are all sound, but performing each one completely before starting the next means the first real evidence about whether the design works arrives after the budget is spent.

SAM keeps the same activities and re-orders them into short cycles. Each cycle produces something reviewable — a sketch, a prototype, a working module — and each review is allowed to change the design. Convergence is by successive approximation: the design is deliberately rough early, and the roughness is the point, because a rough artifact is cheap to reject.

The process opens with a Savvy Start: a short, intensive session in which stakeholders, subject-matter experts and designers sketch and critique candidate designs together, rather than the designer interviewing them and going away to write a specification. Later cycles alternate design–prototype–review, with development running behind design rather than after it.

Where the scope itself is unsettled, SAM is the wrong tool and Design Thinking is the right one: SAM iterates within an agreed problem, and assumes the goals are broadly known even when the design is not.

Implications

Context

Requirements

• Stakeholders who will show up repeatedly. The process substitutes recurring short reviews for one long specification, and collapses to a waterfall without them
• Tolerance for visibly unfinished work. A reviewer who reads a rough prototype as a poor product will drive the team back to specifying up front
• A team that can prototype quickly — storyboards, clickable shells, one built module
• A change budget. Iteration is only real if review findings are allowed to alter the design after work has been done

Constraints

• Scope creep is the standing risk. Each review is an invitation to add, and without an agreed scope the cycles do not converge
• Poor fit for a fixed, auditable derivation chain. Where an accreditor needs objective-to-item traceability, the systematic process produces that artifact and SAM does not, by design
• Rework is real cost, not waste avoided. SAM trades late expensive rework for early cheap rework; where the analysis is genuinely reliable up front, that trade loses
• Depends on review quality. Reviews that collect preference rather than performance evidence steer the design confidently in the wrong direction — the same failure Feedback is most effective when directed at the task and process rather than the self. [+S] names for learners
• Distributed or asynchronous teams lose most of the Savvy Start's value, which comes from co-present sketching and immediate critique

Grain Size

• Course
• Module
• Programme (run per course, coordinated across)

Target Goals

• Designs where the requirement is understood but the right solution is not
• Projects with real delivery deadlines, where a partially-complete but working design beats a complete specification
• Redesigns of courses that are known to be failing, where the failure's cause is contested

Target Learners

• Any; SAM constrains the design process, not the audience. Its learner-facing leverage is indirect — the early cycles put a prototype in front of real learners sooner, so learner evidence reaches the design while it can still be acted on

Theory

Supporting

• Design Layers Theory — a design decomposed into layers that can be revised at different rates is what makes iteration tractable
• Designerly Stances

Contradicting / Qualifying

• Behaviorism — the criterion-referenced tradition assumes objectives are fixed before design begins, which SAM deliberately relaxes

Claims

• Participant feedback in macro cycle one prompted renaming the Administrator standard to Coordinator and reordering the standards [+W]
• The prototyping exercise promoted exchange of ideas for HLC development and enhanced a shared image of the HLC among team members [+W]
• Scheduled design cycles and peer review most successfully evoked iteration; the resources message and chat sessions were less successful [+W]

Supporting

• Feedback is most effective when directed at the task and process rather than the self. [+S] — the standard each review cycle has to meet
• Goal setting improves performance. [~M] — iteration does not remove the need for a target; it defers how precisely it is specified

Contradicting

• Guidance becomes more necessary as task complexity and learner inexperience increase. [-M] — a design team without instructional expertise iterates without a criterion, and repeated approximation converges on whatever the loudest reviewer prefers

Design

Sequence

1. Preparation — gather background and known constraints; keep it short. This is information collection, not analysis-to-completion
2. Savvy Start — stakeholders, subject-matter experts and designers sketch, critique and reject candidate designs in one session. Expect several designs to be discarded; that is the deliverable
3. Iterative design — cycles of design → prototype → review. Each cycle is short and produces something a reviewer can react to
4. Iterative development — cycles of develop → implement → evaluate, overlapping the design cycles rather than following them. Alpha, beta and gold releases mark convergence, not phases
5. Roll out, with the evaluation instrumentation already built in — see Continuous Improvement of Learning Materials for what happens next

Elements Used

• Learning Objectives — sharpened across cycles rather than fixed at the start
• Formative Assessment
• Iterative Learning

Affordances

• Formative Assessment

Personalization

• Two-phase variant. Allen distinguishes a lighter form for small projects — preparation plus a single iterative design-and-develop phase — from the fuller form that separates design and development cycles. Choose by whether development is expensive enough to deserve its own loop
• Where stakeholders cannot convene, replace the Savvy Start with a sequence of short paired reviews and accept slower convergence; do not replace it with a written specification, which reintroduces the failure the process exists to avoid
• Where an audit trail is required, run SAM for the design and generate the objective-to-assessment map as a release artifact at gold, rather than as the design's starting point

Related Processes

• Systematic Instructional Design — the process SAM was formulated against; its phases survive, its ordering does not
• Design Thinking — iterates the problem statement as well as the solution; SAM keeps the problem fixed
• Continuous Improvement of Learning Materials — the same iterative logic applied after release, driven by usage data
• Learner Experience Design — supplies the evaluation methods the review cycles need

Examples

• Corporate e-learning production — the setting SAM was developed in, where fixed delivery dates and shifting subject-matter make a full up-front specification unaffordable
• Course redesign with a live cohort — successive rounds released to consecutive cohorts, each informed by the last
• Replace monolithic comprehensive studies with smaller, agile research cycles that test one conjecture at a time

Key Sources

• Cullen. Ch. 26 in Design for Learning. EdTech Books. https://edtechbooks.org/id/agile_design
• Allen, M. W. (2012). Leaving ADDIE for SAM. ASTD Press.

Page metadata

Field	Value
type	process
id	successive-approximation-model
title	Successive Approximation Model (SAM)
description	An iterative, agile course design process that replaces a single analyse-build-evaluate pass with short rounds of design, prototype and review, starting from a collaborative Savvy Start and converging on a design by repeated approximation rather than by up-front specification.
status	review
grain_size	course
author	Michael W. Allen
generated	by claude/unspecified on 2026-09-02
trust	unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.