Training Modules: What They Are and How to Build One

What a training module actually is

A training module is the best way to deliver a given piece of knowledge to a given individual. That is the definition I use, and it is deliberately narrow. It is the most effective way to ensure that a given person actually learns what the material proposed to teach.

Most of what gets called a training module fails that test. It delivers a piece of knowledge to everyone, in one modality, at one length, and the objective is implied rather than set out. That is content. A module is content that has been shaped around one person learning one thing.

The rest of this guide is about the decisions that shape it: whether you need it, how long it should be, how to cut a big topic into modules, where teams get stuck, and how you know it worked.

Start by deciding whether you need training at all

There is a practice that more senior folks follow, and it is the one thing I would push hardest. You start your whole building process by defining whether you need training at all. Sometimes the answer is that you don’t.

This gets skipped because it is uncomfortable. A request arrives, it names a module, and building the module is the visible response. But a lot of what arrives as a training request is a process problem, a tooling problem, or a communication problem, and a module aimed at it will be completed by everyone and change nothing.

There have been some ChatGPT apps built for exactly this, and other AI apps that walk you through the concept. ADDIE is also a good framework to help you decide, because its first phase is analysis rather than design. If the analysis says the gap is not a knowledge gap, you have saved yourself the entire build.

What separates a good module from a bad one

There are several reasons a module fails, and two of them account for most cases.

The modality is wrong

The first is not choosing the right modalities. If the audience expects a different modality, let’s say they expect video but you deliver text only, it is not engaging. That is the problem. It is not a content quality problem and no amount of rewriting fixes it, because the mismatch happened before a word was written.

Modality is a decision about the audience, not about the topic. The same compliance policy can be a document for the legal team who will search it and a video for the frontline staff who will watch it once. Both are correct. Where the answer is video, the format and length decisions that follow are covered in the training videos guide.

The quality does not respect how the audience learns

The second is when the quality is not detailed enough, or the quality does not respect what the audience wants and how the audience usually learns that given topic. Then there are several problems at once: the module is too shallow for the people who need depth and too abstract for the people who needed a worked example.

“How the audience usually learns that topic” is the part teams skip. Engineers learn a new system by breaking it. Sales teams learn objection handling by hearing someone else handle the objection. A module that ignores the existing habit is fighting it.

How long a module should be

There is definitely a tendency towards micro learning, and micro learning can cause some problems. Cutting everything to three minutes because short is fashionable produces modules too thin to carry a complex topic.

Otherwise the length should always follow the complexity and the expected time the learners have to sit down and consume the material. That is different for each role. A field technician between jobs and an analyst at a desk do not have the same window, and the same module cannot serve both well.

So length is an output of two inputs, complexity and available time, and neither of them is a house style you can set once.

How to cut a big topic into modules

There are standard frameworks for this, and they are worth reading rather than inventing your own. The practical answer is to keep sticking to a micro learning approach: showing bite sized pieces, not just because of the concept and the ease of consumption, but also for adaptive learning practices, which AI genuinely enables now, and which is where companies are pushing their learning experiences.

That second reason is the one that gets missed. Bite sized pieces are not only easier to consume. They are the unit that adaptive delivery needs. If your onboarding is one ninety-minute block, there is nothing for a system to adapt: you can serve it or not serve it. If it is fourteen pieces, you can skip the four this person already demonstrated and go deeper on the two they failed.

The evidence for chunking is older than the trend. John Sweller’s cognitive load theory, developed in the late 1980s, distinguishes intrinsic load (the difficulty of the material), germane load (the effort that actually builds understanding) and extraneous load (the burden added by how you presented it). Working memory is severely capacity-limited, so complex schemas “may be broken into individual ‘subschemas’ and taught in isolation, to be later brought back together.” That is chunking, described in the 1980s, and it is why the module boundary should fall where a subschema ends rather than where the slide deck did.

The frameworks worth knowing

ADDIE

I follow the ADDIE approach. It’s a really good one, and it is worth detailing here because it is the closest thing the field has to a default.

ADDIE is five phases:

1. Analysis. Clarify the instructional problem, establish the goals and objectives, and identify the learning environment and what the learner already knows.
2. Design. Set the learning objectives, the assessment instruments, the exercises, the content structure, the lesson plan and the choice of media.
3. Development. Create and assemble the content, integrate the technology, build it, test it and revise on feedback.
4. Implementation. Prepare the facilitators and the learners, get the tools and the delivery in place, and make sure everything works.
5. Evaluation. Formative evaluation runs throughout; summative evaluation tests the finished thing against criteria and user feedback.

Florida State University developed the framework in 1975 to describe the processes involved in an instructional systems development programme for military interservice training. The acronym itself came later and its origin is genuinely obscure: Molenda’s 2003 paper points out that practitioners cite the underlying military model constantly but never use its actual initials, so it cannot be the source of the name.

The value of ADDIE is not the sequence, which is obvious. It is that analysis comes first and evaluation is a phase rather than an afterthought. Most failed modules skipped one or both.

Gagné’s nine events of instruction

Robert Gagné set out nine events in The Conditions of Learning, first published in 1965. They work as a checklist for the inside of a single module rather than for the programme around it:

gain attention; inform the learner of the objectives; stimulate recall of prior learning; present the stimulus; provide learning guidance; elicit performance; provide feedback; assess performance; enhance retention and transfer.

If a module feels flat and you cannot say why, walk it against that list. Usually two or three events are simply missing, and it is almost always “elicit performance” and “provide feedback.”

The kind of module most teams underuse

Interactivity, asking the learner questions and changing the module based on the answer, is definitely popular. AI enables this, and I think the market will go in that direction.

Underused is the second half of that sentence. Plenty of modules ask questions. Far fewer change based on the answer. A quiz that reports a score and moves everyone to the same next screen is assessment, not adaptation, and the difference is whether the answer alters what happens next.

That is the gap worth closing first, because it is now cheap to close. Interactive video with branching turns a question into a fork rather than a checkpoint, and the fork is what makes it adaptation rather than assessment.

Where teams actually get stuck

Getting the knowledge out of the people who hold it

Teams get stuck on gathering the core knowledge, the part that has to come out of subject matter experts, or out of the core retainers of knowledge in the organization.

This is the bottleneck, and it is a scheduling problem disguised as a content problem. The expert has the knowledge and no time. The instructional designer has the time and not the knowledge. Every week the module does not exist is a week that gap stays open, and the usual fix, a ninety-minute interview that someone then transcribes, costs the expert exactly the thing they have least of.

Note also that “core retainers of knowledge” is not the same set as your formal SMEs. The person who actually knows how the escalation works is often not the person whose title says they own it.

Production, scale and localization

Secondly there is the production side: using the right AI software, deploying it at scale, localizing it. All of these are development aspects.

This used to be where budgets died. A single module in one language was a production project, and a second language doubled it. That is the part that has genuinely changed. Turning documents into video modules, applying a brand kit, and localizing into a hundred languages without re-recording is now a workflow rather than a procurement cycle, which is what makes the modular approach affordable in the first place.

How you know a module worked

Start with the lagging indicators and work your way back from there. So stop the vanity KPIs of engagement rate. Of course they are important to a certain extent, but you have to build them up towards the lagging indicators. Otherwise L&D will not become a strategic function.

Concretely: name the business number the module is supposed to move before you build it. Time to competency for a new hire. Incident rate for a safety programme. Win rate for sales enablement. That is the lagging indicator, and it is the one the business already tracks and already cares about.

Then choose the leading metrics that feed it, and report them as leading metrics rather than as results. Completion rate is not an outcome. Quiz score is not an outcome. They are early evidence that the outcome might arrive, and presenting them as the outcome is exactly the habit that keeps L&D out of the rooms where budget is decided.

Software for building training modules

Which tool you pick matters less than any of the decisions above, but the tool decides how fast you can act on them. Prices below were checked against each vendor’s own published pricing page on 28 August 2026. Anything a vendor does not publish is marked as such rather than estimated.

One note on a name you will still find in older lists, including an earlier version of this page. EdApp was retired on 31 March 2026 and access was removed on 20 April. The capability moved rather than closed: SafetyCulture bought EdApp in 2020, renamed it SC Training in 2024, and has since folded training into its wider platform, which now trades as Mitti. If a 2026 listicle still lists EdApp as something you can go and buy, it has not been checked since March.

Video-led modules sit alongside all of these rather than replacing them. Colossyan turns documents and decks into avatar-led video modules, adds interactive questions and branching, and exports SCORM 1.2 or 2004 so completions and quiz results report back into whichever platform above you already run.

Where the module is the unit rather than the video, that is Colossyan Learn: course authoring built around the video, with page structure, built-in quiz questions and branching in one editor. That is the gap between producing a video and producing a module. The full set of programme types this covers is on the training use cases page, and if you are choosing the platform layer rather than the content layer, the corporate training platforms comparison goes deeper on that decision.

Where this is going

Adaptive learning is the only way to go, and the only way for L&D to break out of the non-strategic approach to metrics and business measurement. It will help learning and development folks be much closer to the business value and be able to measure the outcomes better. I think that is the direction, for sure.

The connection back to modules is direct. Adaptive delivery needs something to adapt, and what it adapts is modules: bite sized, single-objective, tagged against an outcome. Build the programme as one long course and there is nothing to personalise. Build it as modules and the personalisation is a configuration rather than a rebuild.

That is the case for taking module design seriously, and it is a stronger one than it was two years ago, because the production cost that used to make modularity expensive has mostly gone.