Skip to main content

ONAÏ case study: a Korean pre-order fashion store built to review-ready in 4 days

Published · Updated

TL;DR: ONAÏ is a Korean-style pre-order store that RedClaw owns and runs for Taiwanese shoppers. It sells womenswear, kidswear and mother-daughter outfits (WOMAN, MINI and TOGETHER). We went from proposal on 1 October 2026 to a review-ready site on 4 October, four days in total, with Lighthouse mobile scores of 97–100 and 107 end-to-end checks passing. Checkout stayed closed on purpose during the payment review. The site opened to search engines on 18 October and has taken real payments through ECPay since 27 October 2026.

Some Taiwan context first

  • ECPay (綠界) is a local payment gateway that small Taiwanese web shops apply to for card, bank-transfer and convenience-store payments. Before it lets you collect money, it reviews your website and your business details. That review decided when this store could open.
  • Pre-order here means shoppers can order at any time and the item arrives in about 14 to 30 days. If a supplier runs out, the order is refunded in full.
  • LINE is Taiwan's default messaging app. Links shared there, and on Facebook and Instagram, open in an in-app browser, which turned out to matter for checkout (see pitfall 5).

Store facts

ItemItem
ProductKorean-style pre-order: womenswear, kidswear, mother-daughter outfits
ModelPre-order at any time; delivery in about 14–30 days; full refund if out of stock
TimelineReview-ready 4 Oct 2026 (work started 1 Oct); taking real payments since 27 Oct 2026
StackNext.js 16 (App Router), Firebase App Hosting (asia-east1, ISR 60 s), Firestore, Cloudflare DNS
PaymentsECPay, live since 27 Oct 2026 (ECPay fees, Chinese)
ShippingECPay logistics; free shipping from NT$3,000
IndexingOpen to search engines since 18 Oct 2026; verified in Google Search Console

What did four days of building look like?

Date (2026)What happened
1 OctSEO competition research on Korean-style fashion and pre-order shops in Taiwan; 5 UI directions. We chose our own site over a hosted platform and a "strawberry milk" visual style. Build started
2 OctAdmin, checkout, ECPay on staging, technical SEO, first articles. Zero failed checks
2 OctChinese heading font cut down to a single subset; Lighthouse mobile back to 97–100
3 Oct8-week content plan written; groundwork for a virtual try-on shipped with the feature switched off
4 OctBrand named ONAÏ, domain bought, Little Her girls' line added, visual score raised to 86, first 9 items listed. 88 end-to-end checks passed, then a 107-check regression run passed

What did we measure when the review-ready build was done?

These are the numbers from before the store started taking payments, as of 7 October 2026.

MetricValue
Proposal to review-ready site4 days
Items in first listing9
Lighthouse mobile97–100 (dropped to 85 before the font fix)
Visual design score55 → 78 → 86
Full-site SEO audit70 (weakest: images 25, content 56)
Article SEO acceptance checks6 of 6 passed
End-to-end testsRound 1: 62 checks (53 passed, 7 failed, 2 not built) → round 2: 88 all passed → regression: 107 all passed

Why did checkout stay closed during the review?

Checkout has three modes: review, stage and live. During the review the site ran in review mode, which shows "payment opening soon" on the checkout page. The server also refused to create orders, so even submitting the form with the Enter key was blocked. Since 27 October 2026 it has been in live mode and orders go through normally.

The reason: a payment gateway wants to see a complete site it can browse, but until it approves us, we don't want orders that can't be paid. Our dried-peach store, Taolilife, ran into exactly this. While its payment applications were stuck, its ads kept spending.

Which problems came up in four days?

1. Without product photos, layout can't save you

  • What happened: when the build was done there were no images anywhere, and the visual score was 55.
  • Cost: empty product cards made the whole site look unfinished, no matter how much we polished the layout.
  • Fix: added illustrations and fixed details, which took the score to 86. For a fashion store, product imagery has to come first.

2. Using an environment variable to decide "is this a test?"

  • What happened: the original code treated anything not explicitly set to production as a test environment. If the production site ever lost that variable, a simulated payment would be recorded as a real one.
  • Cost: no real loss, since we aren't taking orders. But it was a design that fails open.
  • Fix: we now check the merchant ID in ECPay's callback. Only ECPay's public test merchant can trigger a simulated payment; everything else is blocked.

3. The email you register with ECPay must match the one on your site

  • What happened: ECPay's application guides (documents, review criteria, both in Chinese) require your registration details to match the contact info published on the website, and our domain mailbox didn't exist yet.
  • Cost: the application had to wait until the mailbox was set up.
  • Fix: setting up the domain mailbox now comes before the application. Taolilife was rejected by ECPay for a similar mismatch, with a phone number.

4. A Chinese heading font split into 110 files

  • What happened: loading the Chinese font through next/font/google split it into about 110 files. Each page downloaded 8 of them, 329 KB in total.
  • Cost: Lighthouse fell to 85.
  • Fix: built a single subset containing only the characters used in headings. Back to 97–100. The subset has to be rebuilt whenever heading text changes.

5. LINE, Facebook and Instagram in-app browsers can't complete payment

  • What happened: when the site is opened inside those apps, the in-app browser can't POST the payment form to ECPay.
  • Cost: shoppers who arrive from a LINE or Instagram link would get stuck right at payment.
  • Fix: checkout detects in-app browsers and asks the shopper to open the page in their regular browser.

Where is the store now?

The store is open for business:

  1. It opened to search engines on 18 October 2026, with Google Search Console verified and the sitemap submitted.
  2. ECPay went live on 27 October 2026.
  3. The focus now is real product photos, styling content and search visibility.

Once the order and ad numbers are large enough to mean something, we'll add them to this page.

What this means for your store

  1. A fast build doesn't mean you can take money fast. We built the site in four days. What holds it up is business details, shipping fees and photo permissions, things only the owner can provide. Our coaching starts from that application checklist. See our e-commerce coaching service.
  2. Don't take orders before the payment gateway approves you. A review mode lets the gateway see a complete site without collecting orders nobody can pay for. Taolilife didn't do this; the cost is in the Taolilife case.
  3. For a fashion store, sort out product images and fonts first. Those two things decided our visual score and mobile speed. Our other stores, including the MaoMuse case, are collected in our e-commerce hub.

FAQ

Can a store built in four days be any good?

We measure it. Lighthouse mobile is 97–100. The first end-to-end round of 62 checks had 7 failures; after fixes, a second round of 88 checks and a 107-check regression run both passed. The full-site SEO audit scored 70, with images and content the weakest areas, so there's work left.

Can people order now?

Yes. ONAÏ has taken real payments through ECPay since 27 October 2026. Before that, checkout deliberately took no orders during the review.

Is ONAÏ running ads?

Meta ads haven't started yet, though the tracking events are ready. Google Search Console was verified when the site opened to indexing on 18 October 2026.

Planning your own web store in Taiwan and want to know what the payment application needs? Message us on Telegram: https://t.me/eric16888999