Selling Courses Online: Choosing and Building Your LMS

Two businesses asked us the same question last quarter from opposite ends of the scale. A solo educator earning about $40,000 a year from three courses asked whether she needed a "real" LMS, while a corporate training company managing tens of thousands of seat licenses asked whether it could stay hosted. Both were really asking the same thing: which parts of a course business should you rent, and which parts should you own.

This is the comparison we run before anyone touches a lesson player: what hosted platforms cost at your volume, what a custom LMS costs to build and run, how delivery UX decides completion, and how payments, drip, certificates and community separate a storefront from a training operation. We split the decision into two options and three pieces, then close with migration triggers. Pricing below is published where stated; everything modeled from engagements is labeled illustrative.

Here is the short version before we dive deep.

Key Highlights

• The global e-learning market sits at about $251 billion in 2020, heading to roughly $458 billion by 2026 (Research and Markets) — demand is not the constraint; delivery is.
• Hosted platforms list at roughly $40–$400 a month, with lower tiers adding transaction fees up to 10% (published list pricing, subject to change).
• Payment processing runs 2.9% + 30¢ (Stripe's published rate) on both paths — the real gap is plan fees, transaction fees and app rent.
• At $10K monthly course revenue a hosted stack costs about 8% of revenue; by $250K a month that falls to roughly 3.2% (illustrative model).
• Three-year TCO for an academy at $870K annual course revenue: **$168K hosted versus ~$224K custom** (illustrative) — custom is a business-model bet, not a savings plan.
• Published MOOC research puts completion at 3–13% (Jordan, 2014; Reich and Ruipérez-Valiente, Science, 2019); structured delivery lifts our model to 46–68%.
• Certificates and evidence are the B2B product: 93% of organizations are concerned about retention, and learning is the top retention lever (LinkedIn Workplace Learning Report 2023).
• Replatforming later costs roughly $12K under 1,000 learners, ~$100K at 10,000–50,000 (illustrative) — pricier every quarter you wait.

The 2026 Baseline: Demand Is Not the Problem — Delivery Is

Online course demand keeps growing, and the money is not small. Research and Markets put the global e-learning market at about $251 billion in 2020, projected to reach roughly $458 billion by 2026 — about 10.6% a year. Interpolating the middle years on that trajectory gives the curve below.

Market growth, however, is not growth in your academy: every topic is crowded, and learners judge the experience as much as the content. The sellers who compound share three traits — delivery students finish, payment mechanics that match how buyers purchase, and provable evidence that training happened.

To keep the arithmetic honest, we will use one illustrative academy throughout: Harbor Compliance Academy, a typical professional-training business with 24 courses, about 3,500 enrollments a year, a $249 average order, and roughly $870K in annual course revenue split between individuals and B2B seat contracts. Rescale the assumptions to your own numbers.

Key takeaway: The market will keep growing without you — what moves your revenue is how well delivery, payments and evidence work, and that starts with choosing the right container.

Option 1: Hosted Course Platforms — Renting the Whole Stack

Hosted platforms — the Teachable, Thinkific and Kajabi class — bundle course authoring, checkout, video hosting, learner accounts, email and basic certificates into one subscription. You can be selling within days, which for a first course or lean catalog is almost always the right first answer: the rent buys speed.

The pricing has two layers, and the second one surprises people. Plans list between roughly $40 and $400 a month, many lower tiers add a transaction fee of up to 10% of course revenue, and top tiers usually waive it. Processing applies either way at the standard 2.9% + 30¢ (Stripe's published online rate), so the honest question is what the whole stack costs as a share of revenue as you grow:

Read that curve in two halves. At small scale the plan fee dominates, so growth is your discount; by $50K a month the floor is payment processing at roughly 3%, which nobody escapes on either path. The model assumes a 0%-fee plan — sit on a 5%-fee tier and add up to five points at every step, at which point hosted stops being cheap.

What the rent buys is real: zero maintenance, a roadmapped feature pipeline and an app ecosystem. What it costs is control — the delivery experience is the platform's opinion, the data model is fixed, and several business models simply do not fit:

• Seat-based B2B contracts — purchase orders, wallets and per-tenant pricing usually need apps or manual invoicing.
• Compliance-grade evidence — verifiable certificates, audit trails and CE-credit reporting are thin or absent.
• Multi-brand or multi-tenant storefronts — one brand per account is the assumption, and workarounds pile up.
• Deep learner flows — content tied to your product data or HR systems lives outside the platform's model.

Key takeaway: Hosted platforms win the first year for almost everyone — just price them as a percentage of revenue, not a monthly fee, and know which of the four hard bullets you will eventually need.

Option 2: The Custom LMS — What Building Costs and Buys

A custom LMS means you own the learning database, delivery engine, storefront and admin panel instead of renting them. The build is the headline cost, but the honest comparison is three years of total cost of ownership — here is that table at Harbor's scale (illustrative):

Year by year, custom front-loads a build and then runs on maintenance and chosen feature work, while the hosted stack climbs slowly as plan, apps and retainers grow.

At this scale the hosted stack wins the raw money math: roughly $168K versus $224K over three years (illustrative). The gap flips in three situations — a 5% transaction-fee plan adds up to ~$44K a year, heavy B2B seat administration pushes app and retainer costs up, and platform workarounds consume staff time that never shows on the platform invoice.

The stronger argument for building is business-model, not cost: purchase-order seat contracts, evidence pipelines that survive audits, white-labeling, and integrations your platform will never ship. We watched exactly that ceiling break a training business in our LMS platform for corporate training case study, where an off-the-shelf LMS gave way to a multi-tenant platform serving 210,000 employees across 40 countries.

Key takeaway: Custom does not save money at most course-selling scales — it buys capabilities that change what you can sell, so justify it with contracts you cannot sign today, never with license fees.

Piece 1: Course Delivery UX Decides Whether Anyone Finishes

Completion is the invisible revenue lever in course businesses. Published MOOC research lands between 3% and 13% completion (Jordan, 2014; Reich and Ruipérez-Valiente, Science, 2019), and while a paid course is not a MOOC, the problem is identical: self-paced content with no structure loses almost everyone. Renewals, testimonials and word of mouth flow from the students who finish.

Structure changes the number, and structure can be engineered:

Our illustrative model quadruples unstructured self-paced completion with three mechanics. Deadlines and drip create a schedule students synchronize to instead of a binge-and-bail archive, cohorts add the social contract of not falling behind, and manager tracking puts corporate compliance training at the top of the range.

Mobile delivery deserves its own line, because study time happens where students are. Our e-learning language app case study shows the stakes: a web-only language school was losing 84% of new members in week one until offline mobile lessons with streak mechanics turned the commute into study time. Hosted platforms give responsive web and, on some plans, a branded-app add-on; custom gives native offline video and push on one account database.

Key takeaway: Completion is an engineering outcome, not an audience trait — drip, deadlines, cohorts and mobile access are features you choose, and each is worth revenue multiples of its cost.

Piece 2: Payments, Drip, Certificates and Community

Payments are easy on both paths for one-off consumer sales, which is why the comparison usually stops there. The differences appear in B2B mechanics: seats bought against a purchase order, drawdown wallets that decrement as employees enroll, per-tenant pricing, net-30 terms and tax invoices finance can process. Hosted platforms approximate these through apps or manual billing; a custom LMS models them natively.

Drip content works the same way: hosted drip runs by calendar or login, which covers most solo sellers, while compliance and cohort products need logic-based unlocks — assessment passed, manager approved, cohort schedule. Drip rules encode your teaching method, and your method is your product.

Certificates are where training businesses leave hobby territory, because in B2B the certificate is the product. LinkedIn's 2023 Workplace Learning Report found 93% of organizations concerned about retention, with learning the top lever — employers buy proof, not videos. Hosted platforms give template certificates on higher plans; custom gives verifiable credentials, QR validation, renewal workflows and CE-credit reporting:

Community is the quiet completion multiplier: bolt-on forums generate noise, while community tied to course progress, cohorts and tenants generates accountability. Treat it as a delivery feature, not a marketing extra.

Key takeaway: Payments, drip, certificates and community are not feature checkboxes — they are the mechanics of your business model, and each is a reason the hosted-versus-custom decision shifts as you grow.

Piece 3: Integration, Migration Triggers and the Verdict

An LMS rarely lives alone once real revenue flows: B2B accounts want CRM sync, learners want single sign-on, completion events should reach HR systems, and finance wants reconciliation without spreadsheet archaeology. The failure pattern is the one in our CRM platform for a recruiting firm case study, where 240 recruiters lost nine hours a month each re-entering data between tools that never talked — training admins accumulate the same tax, one export at a time.

Migration is the other cost that grows quietly — every quarter adds learners, enrollments and history that must survive the move:

Write your triggers down and review them quarterly — they convert a vague itch into a checklist with numbers attached:

The verdicts we give clients collapse into one table:

One discipline cuts the cost of any path: own your data layer, with clean exports of learners, enrollments, progress and certificates wherever you live.

Key takeaway: The platform decision is not permanent but it is expensive to reverse — write the triggers down, review them quarterly, and let the checklist rather than a vendor demo fire the move.

What Your Business Should Do

Platform decisions reward sequence and punish vibes. Here is the 90-day plan:

Days 1–30: Get the real numbers.

• Pull 12 months of course revenue split B2C versus B2B, average order value, refunds and enrollment counts.
• List your current stack with dollar amounts: platform plan, every app, transaction and processing fees, retainers.
• Score your business model against the hard-fit list: seat contracts, evidence demands, catalog complexity, mobile dependency.

Days 31–60: Run the two-path comparison.

• Build the three-year TCO table for your real volumes: hosted stack versus a scoped custom quote.
• Quantify one completion bet — drip, cohorts or mobile — in revenue terms using the Piece 1 model.
• Check every migration trigger in the table above against your last two quarters of numbers.

Days 61–90: Commit and instrument.

• Pick the path the arithmetic says, and write down the thresholds that would reverse it.
• If staying: kill unused apps, re-tier your plan, and set a quarterly take-rate review.
• If moving: commission the data-migration and certificate-preservation plan before any design work starts.

The Bottom Line

The honest 2026 answer: hosted platforms win the money math for most course sellers up to seven figures of revenue, while custom LMS platforms win the business-model math for seat-based B2B training, evidence-driven compliance and mobile-dependent catalogs. Good outcomes come from running the take-rate, TCO and completion analysis before aesthetics or a demo gets a vote.

If you want that analysis run on your real numbers, that is our daily work. Book a free strategy call and we will model the three-year math for your course business together — or send us a message with your questions and we will reply within one business day.