The ID Toolbox
← Guides

How long does it take to build an hour of e-learning?

The published benchmarks answer between 49 and 716 hours, which is a way of saying the question is incomplete. What the research found, what actually drives the number on your project, and how to produce an estimate you can defend.

Updated September 15, 2026

Somebody wants a number before lunch. The honest answer is that the published research spans 49 to 716 hours of work for one finished hour of e-learning, which is not evasion, it is the actual finding. The spread is the useful part: it tells you the question is missing the information that decides the answer.

What the research says

The numbers everybody quotes come from the Chapman Alliance study of 2010, which surveyed around 250 organizations. It reported hours of total work per finished hour of learning, by how elaborate the course is:

Level 1Pages, text, graphics, simple quiz. Little or no audio.49:1 to 79:1
Level 2Narrated, custom interactions, scenario questions.184:1
Level 3Simulation, branching, heavy media and animation.490:1, up to 716:1

Two cautions before anyone pastes those into a proposal. They are from 2010, and they cover the whole project, not the build: analysis, design, development, testing and project management. Benchmarks that look smaller, like the roughly 120 hours per finished hour that IconLogic reports for authoring tools, are measuring a narrower slice of the same work.

ATD has revisited the question several times since, with Robyn Defelice’s research, and the direction of travel is telling: the later work moved away from “one hour of learning” toward time per unit, because most modules people actually build now run about twenty minutes, and the 2023 update warns against using any benchmark without the context of the organization that produced it. Christy Tucker keeps a good summary of the studies if you want the sources themselves.

Why the range is that wide

Because “an hour of e-learning” describes the output, and almost nothing about the output predicts the work. These do:

  • Whether the content exists. Turning a finished policy document into a course is a different job from working out what the policy means with three people who disagree.
  • How many review rounds, and who signs. Two rounds with one owner is a schedule. Four rounds with legal, compliance and a regional committee is a different project with the same word count.
  • Media. Stock images and text-to-speech versus a shoot, a voice actor and custom illustration. This is the single biggest swing after review cycles.
  • Interactivity. Not the number of clicks but the number of distinct things you have to build, test and then fix on three browsers.
  • Accessibility. Captions, alt text, keyboard testing, focus order, a screen reader pass. Real hours, routinely left out of estimates and then done anyway.
  • Translation. Every language adds production, re-recording or subtitling, and layout rework where the text expands.

A worked example

Half an hour of finished course, standard complexity, meaning narrated with custom layouts and a few real interactions, plus professional voice-over. Running that through the estimator:

  • Base rate for standard work is 127 hours per finished hour, so half an hour is about 64 hours
  • Professional voice-over adds 15 percent, taking it to about 73 hours
  • Across the range that is roughly 52 hours if everything goes well and 104 if it does not
  • Split by phase: 10 percent analysis, 20 percent design and storyboard, 45 percent development and media, 20 percent QA and revisions, 5 percent project management

The phase split is what makes the number survive a conversation. Nobody argues with 73 hours as a lump. Plenty of people will argue that 15 hours of QA is generous, and that argument is worth having, because it is really a conversation about how many review rounds they intend to ask for.

Making an estimate you can defend

  • Quote a range, never a point. A single number is heard as a promise and a range is heard as professional. Give the low, the likely and the high.
  • Write the assumptions next to it. Two review rounds. Content supplied by a named date. Stock images. English only. The assumptions are what turn a later change into a change request rather than an argument.
  • Anchor to your own last three projects. Any benchmark beats guessing, and your own history beats any benchmark. If you have never tracked it, start now: hours by phase, per project, is a five-minute-a-week habit that pays for itself the first time somebody challenges a quote.
  • Ask the scoping question first. Sometimes the honest answer is that the hour of e-learning should not exist. Does This Need To Be A Course? walks the eight questions that settle it, and a job aid delivered in a week beats a course delivered in a quarter.

Questions people ask

Has AI made this faster?

For some of it, clearly: first drafts, alt text, caption cleanup, scripting variations, images you would have bought from a stock library. For the rest, not yet, because the constraint on most projects has never been typing speed. It is getting a decision out of a subject matter expert and surviving four rounds of review. Any estimate that assumes AI removed those has skipped the part that actually takes the weeks.

Why do vendor quotes for the same course vary so much?

Usually because they are quoting different work. One has assumed your content is ready and another has assumed they will write it. One included accessibility and another did not. One expects two review rounds and another three. Compare assumptions rather than totals, and make them itemize.

Is Rise faster than Storyline?

For content that fits Rise, yes, often substantially, because you are not building layouts. For anything with custom interaction, simulation or precise visual design, the time comes back with interest. The Authoring Tool Selector scores a specific course against sixteen criteria if the choice is genuinely open.

How long should the course itself be?

Shorter than the request usually assumes. If you are working back from a script, the Narration Timing Converter turns word count into runtime at a chosen speaking rate, which is the fastest way to show somebody that their forty-minute content is twenty minutes of course.

Read next

Tools in this guide