Hosted Store Platform or Custom-Built E-commerce Site? A 3-Year Cost Comparison: Under NT$8 Million in Annual Sales With Standard Workflows, a Platform Usually Wins

Short answer: If annual sales are under NT$8 million and you run standard retail, a hosted store platform costs about NT$150,000–400,000 over 3 years and is usually cheaper. Above NT$8 million, with at least 2 workflows the platform cannot handle (tiered dealer pricing, custom bundles, real-time ERP sync), a NT$250,000–600,000 custom build pays back.

A 6-person handmade skincare brand paid NT$168,000 last year for its platform: annual fee, 7 add-ons and a 1% revenue share. The owner wanted to build a custom site to save money and got a NT$450,000 quote. Run the numbers and the two options differ by only NT$50,000 over 3 years, and a move can still cost SEO traffic. The real question is not "which is cheaper" but "which workflow is holding you back".

When a Custom Site Fits × When It Does Not

Bottom line: consider building your own only when both revenue and workflow complexity cross the threshold.

Alternatives Matrix

Bottom line: there is a middle path, platform plus custom integration, and it is often the cheapest compromise.

Take Shopify as an example: its official pricing page states that the Basic plan charges an extra 2% per transaction if you do not use Shopify Payments. Revenue-based fees like this are what make platform costs climb as sales grow.

Decision Tree: Three Questions

Bottom line: answer in order, and if any answer points to "platform", hold off on a custom build.

• Question 1: Are annual sales above NT$8 million? No → use a hosted platform. Yes → go to question 2.
• Question 2: Are there at least 2 workflows the platform cannot handle? No → stay on the platform and add plugins if needed. Yes → go to question 3.
• Question 3: Is the platform's 3-year cost higher than a custom build's 3-year cost? Yes → build your own. No → choose platform plus custom integration.

Cost formula: platform 3-year cost = (annual fee + add-on fees + sales × revenue-share rate) × 3; custom 3-year cost = build cost + annual maintenance × 3 + hosting × 3. Payment fees are about 2.75–2.8% either way, so you can leave them out of the comparison.

Full Process Breakdown (Custom Route)

Bottom line: a custom build takes about 12 weeks from planning to launch, with 2 weeks reserved for data migration.

• Week 1 | Requirements and cost audit: List current order workflows, add-ons and the last 12 months of platform invoices. Deliverable: a 3-year cost model in Google Sheets.
• Weeks 2–3 | Planning and design: Organize the feature list and MoSCoW priorities in Notion, and wireframe product, cart and checkout pages in Figma. Deliverable: wireframes and spec.
• Weeks 4–9 | Build and integrations: Build the admin and order flow in Laravel, and integrate ECPay payments, convenience-store logistics and e-invoicing. Deliverable: staging site.
• Weeks 10–11 | Data migration: Import products, members and order history, and build a 301 redirect map for old URLs. Deliverable: migration report.
• Week 12 | Acceptance and launch: Run 10 real card payments, refunds and store pickups each. Put the site behind Cloudflare for CDN and protection and monitor it with UptimeRobot. Deliverable: launch checklist.

True Cost Breakdown

Bottom line: the build cost is visible, but maintenance and migration are where people under-budget.

• Build: a standard custom store costs NT$250,000–600,000 (admin, payments, logistics, e-invoicing).
• Hosting and CDN: about NT$6,000–24,000 a year.
• Maintenance: about NT$40,000–80,000 a year for updates, security patches and small changes.
• Payment fees: about 2.75–2.8% for credit cards, on a platform or a custom site alike.
• Hidden costs: an OV SSL certificate at about NT$3,000–10,000 a year if you choose one, SMS verification at about NT$0.7–1 per message, SEO traffic loss during the move, and 2–4 hours a week from an internal point of contact.

Reality vs Expectations

Bottom line: a custom site saves revenue share, not workload.

• Expectation: no more monthly fees. Reality: maintenance and hosting still cost NT$50,000–100,000 a year.
• Expectation: traffic is unaffected by the move. Reality: with incomplete redirects, organic search often drops 20–40% in the first 1–3 months.
• Expectation: a custom site is always faster. Reality: without image compression and caching, LCP can still exceed the 2.5 seconds web.dev recommends.

Common Traps and How to Avoid Them

Bottom line: most failures come from bad math and a rushed move, not the technology.

• Comparing build cost instead of 3-year total cost → run both sides to year 3 with the formula above.
• The domain is registered to the platform or vendor → move domain, payment and cloud accounts into the company's name before signing.
• No 301 redirects for old URLs → build a redirect map following Google's site move guide and keep redirects for at least 1 year.
• Member passwords cannot be migrated → send a password-reset email at launch with a repurchase coupon.
• The spec just says "same as now" → screenshot every platform feature and tag it must-have, later or drop.

Success Metrics + 90-Day Roadmap After Launch

Bottom line: protect traffic and conversion first, then confirm the saved revenue share is real.

• Day 30: zero 404 errors, organic traffic back above 80% of pre-move levels, checkout failure rate under 2%.
• Day 60: conversion rate no lower than before the move, repeat-buyer share back to its previous level.
• Day 90: compare with platform-era invoices to confirm the monthly revenue share and add-on fees actually saved, and plan phase-two features.

Decision Checklist

Bottom line: if 7 or more of these 10 are "yes", a custom build is worth evaluating.

• ☐ Annual sales exceed NT$8 million
• ☐ Total platform fees exceed NT$120,000 a year
• ☐ At least 2 workflows the platform cannot handle
• ☐ Real-time sync with ERP or POS is needed
• ☐ You have B2B or dealer pricing
• ☐ You can budget NT$250,000 or more for the build
• ☐ You can budget NT$40,000–80,000 a year for maintenance
• ☐ Someone internal can act as point of contact
• ☐ The domain is registered in the company's name
• ☐ Products and customers have been proven for over a year

FAQ

Bottom line: owners most often ask about the threshold and the move.

At what annual revenue does a custom e-commerce site make sense?

Usually above NT$8 million a year, plus at least 2 workflows the platform cannot handle. Revenue alone is not a reason to build.

If I move off a store platform, can I take my members and orders with me?

Mostly yes: products, members and orders export as CSV. Passwords and loyalty points usually need a reset or conversion, so check export fields first.

Can I start on a platform and switch to a custom site later?

Yes, and for most brands it is the safest path. Validate products and customers first, then move once revenue and workflow needs justify it.

Next Step

Bottom line: send us your last 12 months of platform invoices and we will calculate the 3-year cost gap for you. ScriptWalker (a Taiwan-based Laravel/Flutter custom development studio) offers e-commerce site builds and system integration: websites from NT$15,000, systems from NT$80,000, and a full custom e-commerce site typically NT$250,000–600,000. If you only need your platform connected to an ERP, we can do just the integration. Free 30-minute consultation:

• Email: [email protected]
• Phone: 0916-224-047
• LINE: @ufv9089p