ADDIE Model

The five phases of ADDIE

1. Analysis. Figure out what problem you are solving and whether training can solve it. Who is the audience, what do they need to do differently, what do they already know, and what constraints apply (time, budget, shifts, languages, devices)? A training needs analysis is the core output.

2. Design. Write measurable learning objectives, choose the format mix (classroom, e-learning, practice, on-the-job), outline the sequence, and decide how you will assess whether people learned it. This is the blueprint, often a design document of a few pages.

3. Development. Build the materials: slides, facilitator guides, e-learning modules, videos, job aids, assessments, roleplay scenarios. Pilot them with a small group and fix what does not work.

4. Implementation. Deliver the training: schedule sessions, prepare trainers, set up the LMS, enroll learners, handle logistics.

5. Evaluation. Check whether it worked. Formative evaluation happens during each phase (reviews, pilots). Summative evaluation happens after delivery, often using the Kirkpatrick model to look at reaction, learning, behavior and results.

Where ADDIE came from

ADDIE grew out of instructional systems design work for the US military in the 1970s, most often traced to a model Florida State University developed for the armed forces. The acronym itself became common later, and the original was more linear and detailed than the five-word version people use now. Today it is less a strict procedure than a shared vocabulary for the stages of instructional design: almost every corporate L&D team uses its phases, even when they work in a more iterative way.

Worked example: ADDIE for a new hire program in a claims center

An auto insurer's claims intake center is losing too many new hires before day 90, and new hires make twice as many coding errors as tenured staff. Leadership asks L&D to rebuild the four-week new hire program.

Analysis (2 weeks). The designer reviews error data, sits with six new hires and four team leaders, and interviews people who quit. Findings: the classroom covers every claim type equally, but 70% of real calls are three claim types. New hires get almost no practice with live system screens before nesting.

Design (1 week). Objectives are rewritten around the three common claim types. The program changes to two weeks of classroom with daily system practice, then two weeks of nesting. Each week ends with a scored simulation.

Development (4 weeks). The team builds 12 practice scenarios in a training copy of the claims system, a facilitator guide and quick-reference cards. Two tenured agents and a new class pilot the first week.

Implementation. The next two classes go through the new program. Trainers get a half-day session on running the simulations.

Evaluation. At 90 days, early attrition and coding errors are compared with the previous four classes. The designer also reviews which scenarios trainees fail most and feeds that back into development.

ADDIE vs SAM and agile approaches

The main criticism of ADDIE is that it can be slow and linear: months of analysis and design before anyone sees a draft, and problems discovered only at implementation. The SAM model (Successive Approximation Model) responds by building rough prototypes early and iterating in short cycles with stakeholders and learners.

In practice the gap is smaller than the debate suggests. Most teams run ADDIE iteratively: quick analysis, a prototype in development, a pilot, revisions. We'd use a fuller ADDIE for high-stakes or regulated content, where the audit trail of analysis and design decisions matters, and a SAM-style approach for fast-changing content like product updates, where a good enough module next week beats a perfect one next quarter.

Common mistakes when using ADDIE

• Skipping analysis because the request seems obvious. A manager asks for a communication course; analysis shows the real problem is an incentive that rewards short calls. Training would not have fixed it.
• Objectives nobody can measure. Understand the refund policy is not an objective. Process a partial refund correctly in under three minutes is.
• No pilot. Content that looks fine in review often fails with real learners. Pilot with a small group every time.
• Treating evaluation as a smile sheet. Post-session surveys tell you whether people liked it. Look at assessment scores and job performance too.
• Letting the project stall in design. Set time boxes for each phase, especially for programs that must launch with a hiring wave.

When ADDIE is the right choice

ADDIE fits best when the training is large, will run for a long time, carries compliance risk, or must be consistent across many sites and trainers. New hire programs for high-volume roles, certification programs, and regulated training like anti-money-laundering or healthcare privacy usually justify the structure. For a one-off 30-minute update, a lighter process is fine as long as you still answer the analysis questions: who, what must change, and how will we know.