The ADDIE Model, Explained

What do ADDIE and SCORM have in common? 

They were both commissioned by the US Department of Defense, albeit nearly 30 years apart. Not what you were expecting, is it? (Although it certainly explains a lot.)

In the 1970s, the US Armed Services had a widening capability gap. They needed new recruits, of varying educational backgrounds, to be able to operate sophisticated equipment. Their ad hoc training approach wasn’t sufficient. 

So they partnered with Florida State University to develop a standardized approach to training. The result was the Interservice Procedures for Instructional Systems Development, which had 19 steps grouped under 5 phases: Analyze, Design, Develop, Implementation, and Control. 

Turns out, they were on to something. By the 1980s, instructional designers in the private sector, and eventually education, had adopted the framework. They ditched the unnecessarily complex steps, and instead adapted the five phases, with one change. They decided to swap out Control for Evaluate (same meaning, just a little less, well, regimented).

And that’s how ADDIE (Analyze, Design, Develop, Implement, Evaluate) came to be.  

‍Analyze

The first phase of ADDIE is conducting a needs assessment.

The term needs assessment was coined by Roger Kaufman in 1972 at Florida State University. (Kaufman was a colleague of the team behind ADDIE, and helped build the instructional design and learning technologies department.) He described a needs assessment as the identification of the gaps between the current results and the desired or required results. In other words, a needs assessment is analyzing the gap in observed outcomes.

Here's why that's so important. Let's say you've been asked by IT to redesign an orientation session. IT insists they need 45 more minutes with every new hire to cover how to use your company's tech stack. (Just what every new hire wants in their first week.)

According to Kaufman, and ADDIE, your job at this phase is to ask IT where they're seeing gaps in desired behavior, and what evidence supports that, such as:

• Are they fielding too many questions in a new hire messaging channel?
• Are new hires unable to request the tools needed for their roles?
• Are new hires responsible for an uptick in cybersecurity incidents? 

In this phase, you're validating whether or not a training problem exists. If you skip this step, you risk creating something based on assumptions, which means you may struggle to demonstrate its value.

Design

Let's say IT comes back to you with data supporting their claims. New hire messages to IT support about the HRIS have increased 30% since last quarter, concentrated on two tasks: enrolling in benefits and setting up payroll direct deposit. After some investigating, you find out the HRIS's UI substantially changed since the IT orientation training was originally designed.

Your next task is to translate this into learning objectives that are specific, observable, and measurable. (This is also known as the ABCD method, Audience, Behavior, Condition, Degree, developed by Robert Mager in 1962.) For instance, you might write two objectives: 

1. After participating in the IT orientation session, new hires will be able to navigate the HRIS platform to elect their benefits within the first 30 days of employment, without IT support.
2. After participating in the IT orientation session, new hires will be able to navigate the HRIS platform to set up their direct deposit by the end of their first week, without IT support.

Taken together, these objectives also identify two types of measurement data: an org-wide baseline (the volume of IT tickets, measured now and compared against after the training rolls out), and individual completion data (tracked by HR and Payroll teams).

In this phase of ADDIE, you’ll also identify the instructional strategy (how the training will be delivered) and content structure (how the content gets sequenced). In this case, you might decide to create an IT orientation eLearning course, with a brief training video showing how to navigate the HRIS.

Develop

Now it's time to build the training based on the blueprint you've designed, which is easier said than done.

In my experience, this phase is where scope creep becomes an issue. While you've technically outlined everything you're going to do, things will inevitably change. A stakeholder will ask for one more thing to be included. An SME will have feedback about your instructional design approach. A budget or staffing resource shifts.

To successfully develop any learning experience, you need to remain focused on the original capability gap you identified. Continuously ask yourself throughout this phase, "are we building the right training?"

In the case of this IT orientation, developing a five-minute training that shows two common workflows should be relatively straightforward. You'd need to work with SMEs to get updated screen recordings, and fact-check the final version. But this task could easily balloon with last-minute asks about other HRIS workflows that should be included, or tangential IT tasks for new hires related to onboarding.

At Synthesia, it's our mission to help people work better. L&D practitioners often come to us because they’re looking for ways to develop content more effectively and efficiently. Learn more about how one of our customers, Criteo, used AI video to improve their onboarding training development.

‍Implement‍

The IT orientation training video is published and ready to embed in your eLearning course. In this phase, you need to make sure the implementation of the new video is sequenced as originally designed. (If you were designing a live training, this phase is also where you'd make sure your facilitators are ready.)

This is your time to be a marketer and bring awareness to the training. After all, you went to all this effort, you want to make sure it's actually watched.

That includes deciding where to host the video and how to announce it: an enrollment email with reminders from your LMS, pinning the video in an onboarding hub, or even a calendar invite that blocks time for new hires with a link included. It also means figuring out the optimal timing. For this training, that might mean two nudges: a notification after new hires attend a benefits session from HR, and a reminder before their benefits enrollment deadline.

A successful implementation ensures you'll have sufficient data to evaluate the quality of the training. There's nothing worse than designing an objectively great training that no one actually watches or completes.

Evaluate

The final phase of ADDIE is also the one that gets skipped, or compromised the most. (I’m also guilty of this.) There’s so much pressure for L&D practitioners to consistently churn out content that evaluation is often a lower priority, contributing to what I call the readiness debt. 

The good news is that if you use ADDIE as a framework, you've already built the foundations for evaluating your training. For simple trainings, like this one, you can use the Kirkpatrick model for evaluation.

To do so, you start by determining a reasonable time frame, a month or even a quarter, after releasing the new training video. Then, you're going to collect the data from IT, HR, and Payroll.

You can complement this with feedback from new hires collected in onboarding surveys, that's your Level 1. Level 2 is whether new hires could actually enroll in benefits or set up direct deposit without escalating to IT, the same completion data you defined back in Design. 

A drop in IT tickets overall is Level 3, evidence new hires are applying that skill unprompted, beyond just the two tasks you measured directly. To get to Level 4, translate that into what it's actually worth: fewer support tickets means fewer hours IT spends on password resets and walkthroughs, time and cost you can put a number on.

Measuring more complex training may require another measurement approach, like the Phillips ROI method. 

How relevant is ADDIE?

At this point, you might be wondering if ADDIE is outdated. 

I think that’s the wrong question to ask. 

During my doctoral training, one of the things I was often asked to do was try on new perspectives, whether those were intellectual theories about how we acquire knowledge or pedagogical approaches. The goal wasn’t to accept any of these carte blanche, but rather to interrogate them, figuring out what we agreed with and what we didn’t. The decision didn’t matter, what did was the reasoning behind it. 

I’d challenge you to do the same with any learning and development framework. 

While you can certainly critique ADDIE for being too rigid, and created in a very different context over fifty years ago, you might be surprised at how it does (or doesn’t) challenge your current approach to creating learning experiences. ‍

ADDIE and AI ‍

While you’re at it, I’d recommend considering one other thing in this evaluative process: the role of AI in learning design. (I know, I work at an AI company, I can’t help it.)

In our 2026 AI in Learning & Development report, which was based on responses from over 400 L&D practitioners, we found that teams are primarily using AI to support the early phases of ADDIE. 

What that signals to me is a few things. First, because L&D practitioners spend most of their time in the Analyze, Design, and Develop phases of content creation, there's the greatest opportunity for AI to make these more efficient. Second, L&D generally isn't sure how to, or isn't comfortable, delegating more complex and strategic work, like measurement, to AI tools.

Whether or not you decide to use ADDIE, consider this something to reflect on. It may be a way for you to think through how you integrate AI into your tech stack and workflows, and why.

If you’re interested in learning more about how practitioners are (or aren’t) using ADDIE, consider joining our free community. You’ll find great conversation.