eLearning Development Costs: What a Course Actually Costs to Build

What Actually Goes Into the Cost of an eLearning Course?

Six lines. Most budgets contain three of them, and the three that get left out are the ones that overrun.

1. The authoring tool. The licence that lets somebody build the course. The only line with a published price, which is why it gets treated as the cost of the project when it is usually the smallest part of it.
2. The build time. The hours somebody spends turning source material into lessons, questions and structure. The largest line in almost every project, and the one that varies most by tool.
3. The subject matter expert's time. The hours of the person who actually knows the content: explaining it, reviewing it, and correcting the three things the draft got wrong. Charged to nobody, costed by nobody, and routinely the bottleneck that sets the delivery date.
4. Media. Photography, video, voiceover, illustration, screen recordings. Historically the line that could double a project on its own, and the line that generative tooling has changed most.
5. Delivery and hosting. Getting the finished course in front of learners and recording that they did it. Either a platform fee, an LMS seat count, or both.
6. Maintenance. Every revision for as long as the course stays in use. Absent from essentially every initial budget and, over a three-year life, frequently larger than the original build.

Notice what is not on the list: the learners' time. It is real and it is often the biggest number of all, but it belongs to the return calculation rather than the development budget. Our training ROI guide is where that side is worked out, and mixing the two is how a cheap course gets approved for an audience that cannot afford to sit it.

Why Cost Per Finished Hour Has Stopped Being a Useful Number

For two decades the industry costed eLearning in dollars per finished hour of content. The unit made sense when a course was a linear sequence of screens, when the build was the dominant cost, and when an hour of content was roughly comparable between one project and the next.

It has three problems now, and all three are getting worse.

1. An hour of content is no longer the deliverable. If a programme is eight four-minute lessons released over a quarter, the denominator describes a thing you never built. Costing it per finished hour produces a number that is arithmetically correct and operationally meaningless.
2. The build is no longer the dominant cost. When generation takes minutes, the expensive hours shift to deciding what the course should contain and getting the content verified by somebody who knows. Per-hour pricing hides that shift entirely.
3. The ratio was always a range, not a figure. Published cost-per-hour figures bundle together simple page-turners and bespoke simulations. Any average across those two is an average of two different industries.

The replacement is not a better average. It is counting your own six lines, which takes about twenty minutes and produces a number you can defend line by line when somebody asks where it came from.

How Much of the Budget Is the Authoring Tool?

Less than most people expect, and it is the only line with a published price, which is exactly why it absorbs an unfair share of the argument. These are the figures the vendors publish. Anything a vendor does not publish is marked as not published rather than estimated, because an invented number is worse than an absent one.

Vendor pricing in this table was last verified in July 2026. Check the current page before you build a business case on any of it, including ours.

Three things that table is doing that a feature comparison would not.

1. It separates authoring from delivery. They are different purchases and different cost curves. A tool that is cheap to author in can be expensive to deliver from, and the pricing page rarely presents both halves at once.
2. It shows where the cost scales with people. QuikAuthor and Articulate 360 charge for authors. SC Training and TalentCards charge for learners. At 50 learners those look similar. At 2,000 they are not the same purchase.
3. It names what is not published. Five platforms in the category publish no pricing at all, which means any budget including them has a line in it that depends on a sales conversation you have not had yet.

How Much of the Budget Is Delivery?

This is the line that decides the three-year number, and it turns on one question: does the price follow authors or learners?

Per-learner pricing is honest and predictable for a stable, fully-trained workforce. It gets expensive in exactly three situations, and they are common ones.

1. High turnover. You pay for a seat, the person leaves, you pay for the replacement's seat too.
2. Seasonal or surge headcount. A retailer tripling for December pays December prices for an audience that trains once.
3. Occasional learners. The contractor who sits one induction a year costs the same as the manager who trains monthly.

QuikAuthor counts active monthly learners, so somebody who did nothing in a given month does not appear in that month's figure. Pro and Team each include up to 200, and beyond that the rate falls as the audience grows:

• 201 to 500 learners: $1.00 per learner per month
• 501 to 750 learners: $0.80 per learner per month
• 751 to 1,000 learners: $0.65 per learner per month
• 1,001 to 2,500 learners: $0.50 per learner per month

One mechanic in this category is worth knowing about before you model anything, because it is not what the words suggest. Articulate's Reach 360 Pro is $3,600 per year all-in and includes 1,200 active-learner credits. A credit is not a person. One credit is consumed per learner per 30-day period in which they train, and deleting a learner does not return the credit. A workforce of 400 people who each train in four separate months consumes 1,600 credits, not 400.

What Does the Internal Time Cost?

Nobody invoices for it, which is why it is missing from most budgets and present in every project. Count it in hours and price it at whatever your organisation uses for loaded internal cost. Six buckets:

1. Scoping. Deciding what the course is for and who it is for. Usually half a day, and the cheapest half-day in the project, because it is the one that prevents building the wrong thing. Our training needs analysis guide is the method.
2. Source gathering. Finding the current version of the procedure, the policy or the deck. Longer than anyone expects in an organisation where the current version lives in three places.
3. The build. Turning source into lessons, questions and structure. The line that tool choice moves most, and the one where the difference between tools is measured in days rather than percentages.
4. Subject matter expert input. Explaining the content, then reviewing the draft. The SME is usually senior, usually busy, and usually the reason a two-week project takes six. Our tribal knowledge guide covers getting more out of less of their time.
5. Review and sign-off. Legal, compliance, brand, and the manager who asked for it. Budget for two rounds, not one, and name the approvers before you start.
6. Publishing and assignment. Exporting, uploading, enrolling, testing on a real device. Small, but never zero, and the step where a Friday deadline is actually lost.

Two of the six are where AI changes the arithmetic rather than shaving it. The build collapses when a PDF, a recording or a deck can be turned into structured lessons with draft knowledge checks automatically. SME input shrinks when the expert can talk for fifteen minutes into a phone and have that become a draft, instead of writing a document that somebody else then rewrites. The other four stay roughly as they were, which is the honest version of what the technology does.

The Line Almost Nobody Budgets: Keeping It Current

Build cost is a one-off. Maintenance is a subscription you pay in somebody's time, and over a three-year life it routinely exceeds the original build for any content that moves.

Five triggers that force a revision, all of them normal:

1. The underlying policy or procedure changes. The most common and the most urgent, because the training is now wrong rather than merely dated.
2. The system in the screenshots changes. Any course containing a user interface has a built-in expiry date set by somebody else's release schedule.
3. The regulation changes. Externally imposed, non-negotiable, and usually with a deadline attached.
4. The organisation changes. New brand, new structure, new job titles, a merger. Cosmetic, mandatory, and tedious in proportion to how much media the course contains.
5. The evidence says it is not working. The best reason to revise and the one least often acted on, because it requires somebody to have been looking at the per-question results.

Two questions to ask any vendor before you buy, both about maintenance rather than building:

1. How long does a small change take? Not a rebuild. One sentence, one question, one screenshot. If the answer is "re-export and re-upload the package", your maintenance cost is set by the delivery pipeline rather than the edit.
2. What happens to the media when the text changes? If narration was recorded by a person, correcting a sentence means a studio. If it is generated from a script, it is an edit. That single difference is worth more over three years than most feature comparisons.

How Do You Build the Number for Your Own Project?

Eight steps, about twenty minutes, and the output is a figure with a visible derivation rather than a range copied from an article.

1. Define the deliverable in minutes, not hours. "Six lessons of about four minutes" is a scope. "An hour of eLearning" is a unit of measurement pretending to be one.
2. Count the source material you already hold. Every approved document, recording or deck that can become content is build time you do not have to spend. Count them individually.
3. Estimate build hours per lesson, then multiply. Use your own last project, not an industry figure. If you have never measured one, build the first lesson and time it before estimating the rest.
4. Book the SME hours explicitly, with names. Hours, not goodwill. An unbooked SME is the single most reliable cause of a slipped date, and a slipped date is a cost.
5. Price the media honestly. Decide now what is generated, what is stock and what has to be made. The third category is the only expensive one, and it is the one that quietly grows.
6. Add two review rounds. Not one. Nobody has ever had one.
7. Add the licence and the delivery cost at your real learner count. Both halves, at the number of people who will actually train in a month, not the number on the payroll.
8. Add a maintenance line for three years. Estimate the revisions per year from the five triggers above and multiply by the cost of a small change. If that number is uncomfortable, you have found the thing that should drive the tool decision.

A Worked Example

Figures below are illustrative and the internal rate is a placeholder. Substitute your own loaded hourly cost and the shape of the answer holds.

A compliance refresher, eight lessons of three to four minutes, built from a policy PDF that already exists, for 180 staff in two countries, with a three-year life. Assume an internal loaded cost of $50 an hour, which is the one number here you should replace first.

The same content commissioned from outside is a different order of magnitude: a custom compliance course from a training consultant costs $5,000 to $15,000, and that figure buys the build rather than the licence, so the maintenance line stays with you either way or comes back as a change request.

Three things the example is deliberately showing.

1. The licence is more than half the year-one total. That happens when the build is small, and it is the case where a cheaper tool genuinely is cheaper. It stops being true the moment build hours grow.
2. The media line is zero because of a decision, not a discount. Choosing generated narration and stock imagery over commissioned media is what removed it, and the trade-off is real.
3. Maintenance over three years is $800, more than the build. Which is the general case, not a quirk of the example.

Where eLearning Budgets Are Actually Wasted

Six failure modes, in rough order of how much money they account for.

1. Building a course for a problem training cannot solve. The whole budget, spent correctly, on the wrong intervention. Still the largest single category of waste in corporate learning.
2. Scoping to the whole workforce by default. "All staff" avoids an argument and multiplies every learner-priced line. The actual audience is almost always narrower.
3. Commissioning media before the content is signed off. Recording a voiceover against a draft means paying for it twice.
4. Treating the pilot as the product. The first version of anything is a draft. Budgeting as though it is final guarantees the revision is unfunded.
5. Buying for the peak headcount. Licensing for the December count and paying it in February.
6. Paying a platform for authoring you never use. Large suites are often bought for one capability and billed for twelve. Check what was actually used after the first year.

How This Fits With QuikAuthor

Being honest about this: QuikAuthor changes two of the six budget lines substantially, one slightly, and three not at all.

The build line is the one it moves most. The AI Asset Converter takes a PDF, a Word document, a PowerPoint deck, an audio file or a video and turns it into structured lessons with draft knowledge checks, which you correct rather than write. Compliance Builder does the same for approved policy documents and writes questions traceable to specific requirements in the text. Both produce a draft that opens in the full lesson builder, so the saving is in the first draft rather than in the editing.

The media line is the second. Narration is generated from your script in premium text-to-speech voices with synced captions created automatically, so a corrected sentence is an edit rather than a studio booking. AI Avatar Video puts a presenter in a course without a camera, metered at 2 credits a month on free and 10 on Pro. AI Translation covers 20 languages from one build, including video transcripts and subtitles, which replaces a per-language production cycle for the text.

The delivery line moves a little, because the pricing follows authors rather than heads up to 200 active monthly learners, and because SCORM 1.2 and 2004 export is included rather than gated behind an enterprise tier as it is at two of the competitors in the table above.

Scoping, SME time and sign-off are not changed by any tool, including this one. They are the lines that decide whether the course should exist, and no software has an opinion about that.

Frequently Asked Questions

How much does it cost to develop an eLearning course?

There is no single figure, because "a course" covers a two-minute policy refresher and a forty-hour accredited programme. A usable number comes from counting six lines for your own project: the authoring licence, the build hours, the subject matter expert's hours, media, delivery at your real learner count, and maintenance across the course's whole life. The licence is the only one with a published price and is usually the smallest of the six, which is why quoting it as "the cost" is misleading in both directions.

Is cost per finished hour still a useful way to budget eLearning?

Not really, for three reasons. An hour of content is rarely the deliverable any more, because short lessons released over time are now the normal shape. The build is no longer the dominant cost, so a unit that measures build output misses where the money goes. And published per-hour figures average simple page-turners together with bespoke simulations, which makes the average a number from no real project. Counting your own six lines takes about twenty minutes and produces something you can defend.

What is the biggest hidden cost in an eLearning project?

Maintenance, followed closely by subject matter expert time. Maintenance is absent from almost every initial budget and, over a three-year life, frequently exceeds the original build for any content that moves. SME time is real, senior and unbilled, which is why it gets left out of the spreadsheet and ends up setting the delivery date.

Does AI actually reduce eLearning development cost?

It reduces two of the six lines and leaves three alone. The build line collapses when a document, recording or deck can be turned into structured lessons with draft knowledge checks automatically. The media line falls when narration is generated from a script rather than recorded, because a correction becomes an edit instead of a studio session. Scoping, review and sign-off are unchanged, and subject matter expert time shrinks only to the extent that the expert can talk rather than write.

Should I build eLearning in-house or outsource it?

Outsource when the content is stable, high-stakes and will not need frequent revision, or when nobody internally has the hours. Build in-house when the content is yours, changes often, or is specific to your policies and systems, because every revision of outsourced content is a new conversation and often a new invoice. For reference, a custom compliance course from a training consultant costs $5,000 to $15,000, and that buys the build rather than the licence.

How much of an eLearning budget is the authoring tool?

Usually a minority of it, and the proportion falls as the project grows. On a small build from an existing document the licence can be more than half of year one. On a project with substantial build hours, commissioned media and several review rounds, it is a rounding error next to the internal time. Published author-side prices range from no charge on QuikAuthor's free plan to $4,995 per year for 7taps Starter, with several platforms in the category publishing no pricing at all.

What is the difference between authoring cost and delivery cost?

Authoring cost is what you pay to build the course, and it usually follows the number of authors. Delivery cost is what you pay to put it in front of learners and record that they did it, and it usually follows the number of learners. They are separate purchases with different curves: a tool that is cheap to author in can be expensive to deliver from. QuikAuthor charges per author and includes up to 200 active monthly learners, with a tiered rate above that which falls from $1.00 to $0.50 per learner per month as the audience grows.

How do I budget for translating a course into other languages?

Historically this was a per-language production cycle: translate the text, re-record the narration, rebuild the layout for the new string lengths. The text and subtitle half of that is now a generated step. AI Translation in QuikAuthor supports 20 languages, translates the whole course including video transcripts and subtitles, and preserves your design, layout and formatting. Budget separately for in-country review, which is a human job and the one that catches the translation that is accurate and wrong.

Where This Guide Stops

Two limits, both outside what any cost model can settle.

A number does not tell you whether the course should exist. A perfectly costed budget for an intervention that was never going to work is the most expensive document in corporate learning, and the only defence against it happens before the costing, in the analysis.

And no model survives an organisation that will not record its own time. If nobody logs the hours spent on the last project, every estimate for the next one is a guess wearing a spreadsheet. The first useful thing most teams can do is measure one real build, end to end, and use that instead of anybody's published average, including the ones here.

Price Your Next Course Before You Commit to It

Take the next course on your list, count the six lines, and build the first lesson to check your estimate. QuikAuthor's free plan includes 25 courses, up to 50 active monthly learners and every AI feature, with no credit card and no expiry, which is enough to measure a real build rather than estimate one.

Related reading: Measuring Training ROI · Training Needs Analysis · Rapid Authoring Tools · Pricing