A real scenario: the courses sell, the platform gets stuck
A design teacher with 30k YouTube subscribers wanted to turn free content into paid courses. She started on an overseas SaaS, hit about NT$400k revenue in three months — but fees plus payment processing ate nearly 15%, student data sat with the platform, and she couldn't customize the learning flow. Her real question was one thing: when should you move from "renting a platform" to "building your own"? This piece answers that decision and lays out a full plan for a self-built LMS.
When to build vs when not to
| Good fit for a self-built LMS | Not a fit (use SaaS first) |
|---|---|
| Stable paying audience; want to own members and data | Still testing whether the course sells |
| Need custom flows (quizzes, certificates, community, gated tiers) | Only 1-2 simple courses |
| Volume large enough that SaaS fees > amortized build cost | No ops budget, don't want to manage tech |
| Must integrate existing systems (CRM, LINE OA, invoicing) | Need "selling this week" timelines |
Alternatives matrix
| Option | Pros | Cons | Cost tier |
|---|---|---|---|
| SaaS (Teachable etc.) | Selling in days, no ops | Fees, data not yours, hard to customize | $ tens-hundreds/mo + cut |
| WordPress + LMS plugin (LearnDash etc.) | Cheap, mature plugins | Weak performance/video protection, plugin risk | NT$50k-150k build |
| Custom build (Laravel + streaming) | Fully custom, own your data, deep integration | High upfront, needs ops | NT$350k-800k + maintenance |
Full process (5 stages, with time, deliverables and tools)
- Stage 1: Requirements & course structure (1-2 wks) — course count, unit structure, quizzes/certificates, payment method. Deliverables: MoSCoW feature list, wireframes (Figma).
- Stage 2: Design & technical selection (1-2 wks) — UI, streaming (Mux/Cloudflare Stream), payments (Stripe/ECPay). Deliverables: designs, architecture diagram.
- Stage 3: Development (4-8 wks) — members/permissions, courses/units, signed video streaming, payments, progress tracking, admin. Deliverable: testable staging.
- Stage 4: Testing & payment UAT (1-2 wks) — payment cases, refunds, video permissions, cross-device. Deliverables: test report, UAT sign-off.
- Stage 5: Launch & migration (1 wk) — domain, SSL, data import, monitoring (UptimeRobot). Deliverables: launch checklist, ops handover.
Real cost breakdown
Build fee NT$350k-800k by flow complexity; video streaming billed by views (Mux-style services start at a few cents per streamed minute, a popular course can be a few thousand to over NT$10k/month); payment fees — ECPay card ~2.75-3%, Stripe ~2.9% + fixed; SSL usually free (Let's Encrypt). The most overlooked hidden cost is ongoing maintenance (NT$8k-30k/month) and the fact that per-transaction SaaS cuts scale up with revenue. One reminder: SaaS fees don't get cheaper as you grow; an amortized build does.
Reality vs client imagination
| Clients think | What actually happens |
|---|---|
| Just upload the videos | Video protection (signing, watermark, streaming) is the heaviest part |
| Build it and buyers come | The platform is just a cash register; traffic and conversion are another fight |
| Payments are a quick integration | Refunds, invoices, reconciliation and subscription renewals are the devil in the details |
| Launch is the finish line | Completion rate, repeat purchase and support are where operations begin |
Common traps and how to avoid them
- mp4 straight on the server: that's open download. Use streaming + signed short-lived URLs.
- No dynamic watermark: leaks are untraceable. Show a student-email watermark to deter re-sharing.
- Only testing the happy path: refunds and double-charges blow up post-launch. Test failure and edge cases.
- No invoice automation: manual invoicing collapses at volume. Integrate an e-invoice API in v1.
- No completion analytics: you can't see where students drop. Instrument unit-completion and drop-off from day one.
Success metrics + 90-day roadmap
- Day 30: zero payment reconciliation errors, video LCP < 2.5s, a baseline completion rate for the first cohort.
- Day 60: reorder units by drop-off, add quizzes and reminders, lift completion 10-20%.
- Day 90: introduce repeat/advanced courses, evaluate App push notifications, check whether LTV supports the amortized build.
Decision checklist
- ☐ I already have a stable paying audience (not still testing)
- ☐ I need custom learning flows (quizzes/certificates/community)
- ☐ SaaS fees per year now exceed a three-year amortized build
- ☐ I want to own member and learning data
- ☐ I need to integrate existing systems (CRM/LINE OA/e-invoice)
- ☐ I have a monthly ops budget (NT$8k+)
- ☐ Video protection is critical to my content value
- ☐ I can accept an 8-14 week build
- ☐ I (or an outsourcer) will run operations and support post-launch
- ☐ I understand the platform is a cash register; traffic is another fight
More checks, better fit for building; fewer than half, validate on SaaS first.
FAQ
How much and how long for a self-built course platform?
An MVP with payments, members, video protection and progress tracking runs about NT$350k-800k over 8-14 weeks, depending on course count, live streaming and whether you need an App. Validating on SaaS costs tens-hundreds of dollars monthly and sells in days but takes a cut. Validate first, then invest.
Can students download and leak the videos? How to protect?
No 100% download prevention, but you can make piracy uneconomical: HLS streaming + signed short-lived URLs, plus a dynamic watermark showing the student email. Far safer than mp4 on a server, and the heaviest part of a self-build.
I have few courses — is building worth it?
Usually not while testing the market — use SaaS. Building pays off when you own member assets, need custom flows, or volume makes SaaS cuts exceed the amortized build. Rule: when SaaS fees/year exceed a three-year amortized build, build.
Do I need an App, or is a website enough?
Usually an RWD site is enough — learn in a mobile browser, cheap and fast to update. An App's value is push reminders (higher completion) and offline viewing. Validate on web first, add an App once you have steady paying users and completion is key.
Call to action
Not sure whether to rent or build for your case? We offer a free 30-minute technical consult to translate your course structure, payments and video-protection needs into a clear timeline and quote. ScriptWalker builds long-lived course platforms with Laravel + Flutter, from NT$350k.
- Email: [email protected]
- Phone: 0916-224-047
- LINE: @ufv9089p