How to build an app like Udemy: features, tech stack, and cost

Most founders who want to build something like Udemy are not trying to compete with 62 million students globally. They are building a course marketplace for a specific professional niche, a training platform for their industry's certification programs, or an internal learning system for a network of clients who have outgrown generic tools like Teachable. Udemy's generic model creates real friction for those use cases -- its revenue share, content policies, and discovery algorithms all serve the mass market, not your vertical.

A course marketplace MVP costs $25,000--$60,000 and takes 14--20 weeks. A full two-sided platform with instructor tools, payout management, and mobile apps costs $70,000--$140,000 over 24--36 weeks. The single biggest cost driver is video infrastructure, not the student UI.

Two platform models: different architecture, different business

Before scoping anything, decide which business you are building. The architecture, cost, and timeline differ by roughly 40%.

Open marketplace (Udemy model): any qualified instructor can apply, publish courses, and earn a share of every sale. The platform earns a revenue share. You need creator onboarding, course submission and approval workflows, a self-serve course builder, instructor analytics, and automated payout management. According to McKinsey's 2023 learning industry report, open marketplace platforms achieve 3--4x the user acquisition rate of curated platforms, but curated platforms retain learners 60% longer.

Curated platform (Masterclass model): you invite creators. The catalogue is small, production value is high, and access is typically sold via subscription. You skip most of the creator self-service infrastructure. Build costs are lower -- roughly 30--50% less -- but your operational model depends on creator partnerships rather than platform network effects.

Most teams want the open marketplace because it scales without managing creator relationships individually. Most teams also underestimate what that model actually requires.

How does a course marketplace make money?

Udemy's revenue model is a tiered split. The platform keeps 37% of every sale generated by its own organic traffic and promotions. When an instructor drives the sale themselves -- via their own coupon code or affiliate link -- the platform keeps only 3%. The instructor keeps the rest in both cases.

Your marketplace can set any split. Common structures:

• Flat platform cut: 20--30% on every sale regardless of traffic source. Simple to explain to instructors.

• Traffic-tiered split: lower platform fee when the instructor drives the student, higher when the platform does. Mirrors Udemy's model. Rewards instructor marketing effort.

• Subscription access: students pay a monthly or annual fee for unlimited access. The platform distributes revenue to instructors based on watch minutes. Masterclass and LinkedIn Learning use this model.

• B2B licensing: sell team or company access at a flat annual price. Udemy Business generates significant revenue this way. Simpler to monetize early if your target audience is corporate training buyers.

Unit economics matter at small scale too. At 500 monthly course purchases averaging $40 each, a 30% platform cut generates $6,000 per month in gross revenue before payment processing fees. Stripe's standard rate (2.9% + $0.30 per transaction) costs roughly $730 of that. The remaining $5,270 needs to cover hosting, video delivery costs, and support before the platform is profitable. Plan for that math before you set your pricing.

Who builds a course marketplace instead of using Udemy or Teachable?

Four specific company types find Udemy and off-the-shelf alternatives insufficient.

Professional associations and certification bodies that issue industry-recognized credentials. Teachable and Thinkific do not support proctored assessments, third-party credential verification, or the accreditation audit trails these organizations need. A custom platform lets them issue verified certificates that integrate with LinkedIn and employer HR systems.

Corporate training departments at mid-market companies where the training catalogue is internal-only and the buyer (the L&D team) needs single sign-on, manager-level completion reporting, and integration with the company's HRIS. Udemy Business works, but at $300--$500 per seat annually and with no way to host proprietary content, it becomes expensive and limited as headcount grows.

Niche professional verticals -- construction safety, medical device compliance, financial services continuing education -- where the content is tightly regulated. Udemy's general-purpose platform does not support jurisdiction-specific access controls, required completion verification for regulatory audits, or the specialized quiz formats some certifications require.

EdTech startups building a content marketplace for a specific creator community where the audience already exists (e.g., a fitness creator network, a coding community, a design platform) and the founders want the platform economics rather than sending 30--37% to Udemy on every sale.

What features do you need to build, and when?

The common mistake: planning the student experience end-to-end and leaving the instructor side for "phase two." Instructors are your supply. Without a working instructor product, you have no catalogue. No catalogue, no students.

The phasing below keeps the instructor path on track from week one.

Cross-platform mobile (one codebase for iOS and Android) saves $15,000--$25,000 compared to building native apps separately. We build cross-platform unless there is a specific camera or performance reason not to. For course platforms, there is rarely such a reason.

A managed video hosting service saves 6--8 weeks compared to building your own transcoding pipeline. That time saving typically exceeds the annual hosting cost of the managed service at early scale.

What are the student-side features?

The student experience needs five things to work at launch: course discovery, a preview before purchase, the course player, progress tracking, and a completion certificate.

Udemy's 2024 Global Learning & Skills Trends Report found that students who watch a preview before purchasing complete courses at twice the rate of impulse buyers. Course completion rates on the platform average 15--20%. That gap between purchase and completion is the core retention problem in edtech -- the preview experience directly affects both conversion and long-term retention.

Course discovery and search

Students browse by category, filter by rating and price, and search by keyword. Under 200 courses, standard database search handles this fine. Above that, you need dedicated search infrastructure with indexing on title, description, tags, and instructor name. Budget an extra $15,000--$25,000 and 4--6 weeks for proper search at scale.

The course card matters: thumbnail, title, instructor name, rating (star + count), price, short preview line. Students scan a page of cards before deciding to click. Get this UI right in V1.

Preview and purchase flow

Students can preview the first one to three lectures before buying. This is the conversion mechanism. The preview must serve partial video access without exposing the full course to non-paying users -- that access control logic adds a week of work that teams routinely skip and then regret.

The purchase flow starts simple: one-time payment per course via standard checkout. Complexity arrives with coupons, bundle pricing, and subscription access. Build one-time purchase first. Every additional pricing model is a separate feature.

Course player

A weak player kills completion rates faster than bad content. The player needs quality selection (360p to 1080p), playback speed control (0.5x to 2x), progress tracking that resumes exactly where the student stopped, a notes panel alongside the video, per-lecture Q&A, and downloadable resources attached to lessons.

On mobile, offline download (saving encrypted video for airplane mode) is expected by users. It adds 4--6 weeks of development.

Progress tracking and certificates

The system tracks completion per lecture and per course. A lecture is "complete" when 80% of the video is watched or a quiz is passed. Course completion triggers a certificate -- an auto-generated PDF with a unique verification ID that employers can check via a verification URL. Simple to describe, mildly tedious to build: PDF generation pipeline, certificate template renderer, and a verification endpoint that stays live indefinitely.

What does the instructor-side product require?

Instructors are your supply. If building and managing courses is painful, they leave and your catalogue empties.

Course builder

The course builder is a drag-and-drop interface for structuring a course into sections and lectures, uploading video files, adding resources (PDFs, code files, slides), and creating quizzes. Video upload is where scope grows fastest. You need a chunked upload system for large files, a processing status indicator so instructors know when their video is ready, and a transcoding step that converts raw video into a format suitable for adaptive streaming. Managed video platforms handle transcoding for you. Building your own adds 6--8 weeks and only makes economic sense at very high volume.

Instructor dashboard

Instructors need to see total enrollments, revenue (gross and net), revenue by course, review ratings, and pending student Q&A. A simple enrollment trend over 30/90 days is expected from day one. Building this after launch creates an instructor support burden that eats team time fast.

Payout system

Every sale generates a split. The platform takes its cut (30--50% on open marketplaces), the instructor receives the remainder. This cannot be handled with a standard checkout flow. It requires a marketplace payment layer. Stripe Connect is the standard choice: instructors create a connected account, students pay the platform, and the platform routes instructor earnings on a weekly or monthly schedule.

Stripe Connect handles currency conversion, local bank account payouts in 40+ countries, and tax form generation (1099 in the US). It adds approximately 0.25% to payment processing cost on top of standard Stripe fees. Plan for 4--6 weeks to configure, test, and certify payout logic across the jurisdictions your instructors are based in.

What admin features does an open marketplace need?

Course approval workflow

Open marketplaces need a quality gate. Instructors submit a course, an admin reviews it (content quality, accurate description, proper structure), and approves or rejects with feedback. The submission queue, review UI, approval or rejection with notes, and instructor notification together take about a week to build. Skip this and you will spend months managing a broken inbox-based process instead.

Revenue split management

Admins configure the platform's revenue share globally and per instructor (for featured creators who negotiate custom terms). Build it as a configurable parameter, not a hardcoded number. It will change -- when you recruit a featured creator, run a promotion, or enter a new market.

Refund management

Most jurisdictions require some form of consumer refund rights on digital goods. Your admin panel needs a refund processing interface: trigger a refund, claw back the instructor payout if it has already been released, and update revenue reporting. Udemy offers a 30-day window. Budget 1--2 weeks to build refund logic that handles instructor clawbacks correctly.

What are the hard parts that teams consistently underplan?

"Course platforms consistently underestimate the engineering cost of video delivery at scale. The transcoding pipeline is not a feature -- it is infrastructure that needs to be right before the first student watches the first lecture." -- Josh Nilsson, VP of Product at Mux, Video Developer Report, 2023

Video delivery: where most budgets break

A student in Mumbai watching a course recorded in London will abandon it if the video buffers. Getting reliable global video delivery right has direct revenue impact -- slow video means lower completion rates, lower completion means fewer reviews and repeat purchases.

The failure mode we see most often: teams start with a basic video upload to S3 and a standard HTML video tag. This works in development. It breaks for students on slower connections because raw video files are not optimized for adaptive streaming. The fix -- adding a proper managed video platform -- takes 2--3 weeks to retrofit but could have been done in the original build for the same cost. Teams that plan for it upfront save that retrofit time and the associated completion rate hit during launch.

According to Research and Markets, the global e-learning market is projected to reach $848B by 2030, growing at 17.5% annually. Video quality is a primary driver of learner satisfaction scores in that market.

Mobile video: a separate problem from web video

Mobile video has failure modes desktop does not: background audio (screen locked, still listening), picture-in-picture, offline download, and cellular data warnings before streaming. These are not edge cases -- they are how people use mobile learning apps daily. Budget 3--4 weeks specifically for mobile video, not just the React Native shell.

Payment splits across countries

An instructor in Brazil, a student in Germany, a platform incorporated in the US. Stripe Connect handles the mechanical payout routing, but you still need to configure which currencies you accept, how currency conversion affects the instructor's payout, what VAT rules apply (EU digital goods VAT is a compliance requirement, not optional), and how to generate correct tax forms per country. Budget 4--6 weeks for a payment system that handles international creators properly.

Build vs. Udemy: when does a custom platform win?

Keep using Udemy (or Teachable, Thinkific, Kajabi) when:

• Your catalogue is under 50 courses and you are testing product-market fit. The platform fee is cheap compared to the development cost.

• Your instructors are not interested in platform economics and just want distribution. Udemy's 62 million students are a real acquisition channel.

• You do not have a recurring audience to seed the platform. A marketplace with no students on day one has nothing to offer instructors.

• You need to launch in under 3 months. An MVP build takes 14--20 weeks minimum.

Build your own when:

• You are paying Udemy or Teachable more than $3,000--$5,000 per month in platform fees or revenue share. At that point, a custom build pays back within 18--24 months.

• Your content is regulated or proprietary -- medical, legal, financial -- and you need access controls, audit trails, or compliance documentation that off-the-shelf platforms do not support.

• You are selling corporate training licenses where SSO, manager reporting, and HRIS integration are table-stakes requirements.

• Your instructors are your product. If their satisfaction with the platform is central to your business model, you cannot afford to be at the mercy of Udemy's algorithm changes and commission adjustments.

Cost breakdown

RaftLabs has shipped SaaS platforms and marketplace products across 100+ engagements. The cost breakdown below reflects actual build complexity. The biggest variable: teams that use managed video platforms typically save 6--8 weeks compared to teams that try to build transcoding infrastructure themselves. That time difference alone often covers the annual hosting cost of the managed service.

What goes wrong in most builds?

Skipping the instructor side in phase one and never catching up is the most common failure pattern. The phased approach makes sense for validation, but some teams never build the real instructor product. They end up with a catalogue locked to a fixed set of courses and no path to marketplace growth.

Building a custom video player is the second most common mistake. Standard players handle 95% of use cases. Custom players are a multi-month project that rarely pays off.

Ignoring mobile video until late is the third. Mobile video has platform-specific behaviors that need dedicated testing time. Plan for it, do not discover it in QA week.

Hardcoding the revenue split is an operational tax you pay indefinitely. It will change when you recruit a featured creator, run a promotional period, or enter a new market. Configure it as a parameter from day one.

At RaftLabs, we've built SaaS products and marketplace platforms across industries. If you're planning a course marketplace or edtech platform and want a realistic scope and cost estimate, book a 30-minute scoping call with a founder -- no sales team, no follow-up sequence.

For related reading: how to build an e-learning platform covers the broader LMS landscape, and our SaaS app development cost guide has line-item cost benchmarks for common features.