Taolilife case study: a one-product dried peach store in Taiwan, 38 paid orders and still losing money on ads

If you have never sold online in Taiwan, three things on this page need a quick explanation.

• ECPay (綠界) and NewebPay (藍新) are local payment gateways. A small independent shop applies to one of them to take credit cards, bank (ATM) transfers and payments at convenience-store counters. Each gateway reviews your website before it lets you collect money, and that review is where we lost the most time.
• Convenience-store pickup is a normal way to ship small parcels here. The buyer picks a 7-Eleven or FamilyMart near them at checkout, and the parcel waits there. ECPay covers the four big chains; a 7-Eleven parcel costs us NT$65.
• LINE is the default messaging app in Taiwan, so our order notifications, order lookup and reviews run through a LINE Official Account rather than email.

1. The supplier's spec didn't match the actual label

• What happened: we only saw the physical pack the day after launch. "No added sugar, 100% peach, 100 g, 12-month shelf life" was all wrong. The label said added sugar (74.1 g per 100 g), 150 g, 6 months.
• Cost: 40 passages to rewrite, one of them in an article about blood sugar. Left as it was, that would have been a false claim.
• Fix: rewrote everything from the label. Rule since then: no ingredient or spec copy until we've seen the real label.

2. Two brands on one payment account fail together

• What happened: to move fast, Taolilife took payments through our other store's NewebPay account. NewebPay audited it and told us to split the brands.
• Cost: from mid to late August Taolilife had no stable way to take payment, while the first campaign kept spending until 21 Aug.
• Fix: one gateway account per brand. No sharing.

3. Every payment rejection pointed back at the website

• What happened: both gateways said the site was missing the owner's name, phone and business address. ECPay also required the phone or email on the site to match the one used for seller verification, and ours didn't. (ECPay's own review criteria say the contact details on your sales page must match your registration.) NewebPay wanted proof of purchase from our supplier, a small farm that isn't registered and can't issue receipts.
• Cost: on 22 August all four applications were stuck at once.
• Fix: added the missing details and made the contact info match the gateway records word for word. For supplier proof we used facts that can be checked, such as the address printed on the pack matching the supplier's own website.

4. "Approved" did not mean we could take money

• What happened: ECPay kept returning "merchant key not found". From the first report to the fix took more than a month.
• Cost: card payments were approved on paper but checkout still didn't work.
• Fix: ECPay's own hosted payment link failed the same way. We used that as proof the problem was on their side, and it was fixed on 29 August.

5. Switching gateways broke checkout for four days

• What happened: when we moved to ECPay, the convenience-store map parameters on the front end weren't updated.
• Cost: 21 checkout starts and zero orders between 31 Aug and 4 Sep, with Meta ads already running again.
• Fix: patched on 4 September, together with a shipping-type field that the same missed change had cut down to 2 characters.

6. Meta reported 65 purchases; we could match 34

• What happened: Meta's dashboard counted 65 purchases. Our database had 34 paid orders with a Meta source.
• Cost: going by the dashboard, cost per order looks half of what it is, and you scale the wrong campaigns.
• Fix: we judge ads only on our own paid orders. Early orders didn't store UTM parameters, so we couldn't tell which ad brought in the very first sale; we added that later.

7. Discovery traffic browses but doesn't buy

• What happened: Google Demand Gen sent 778 clicks in a little over a day.
• Cost: NT$1,899 and zero orders. The real funnel was 497 product views, 9 add-to-carts, 14 checkout starts, 0 orders. Add-to-cart rate was 1.8%, against 10–20% during our Meta campaigns.
• Fix: paused on 30 September. Google now runs only Search and Shopping.

8. Chinese web fonts slowed the site, and the content attracted the wrong searches

• What happened: the Chinese web font made up 76% of homepage transfer, and mobile LCP was 13.1 s. Separately, organic impressions grew on pages like a fruit calendar and shipping costs, while the main product keyword got zero impressions in two months.
• Cost: a slow homepage, and search visitors who mostly weren't looking to buy dried peaches.
• Fix: switched to system fonts. Homepage transfer went from 2,133 KB to 445 KB and LCP to 2.2 s. The content problem isn't solved yet (see below).

1. Fix your site's disclosures before you apply to a payment gateway. Owner, phone, business address and refund policy, matching the gateway records word for word. This alone cost us two weeks. It's the first thing we check for clients, see our e-commerce coaching service.
2. Work out your cost per order before you leave a marketplace. On our numbers, a NT$928 order on Shopee Taiwan would pay about NT$130 in platform fees (our estimate, standard seller, free-shipping program option 1, card payment). The same order on our own site cost about NT$1,194 in Meta ads. A marketplace is often the cheaper way to get a first order; your own site pays off once you have repeat buyers, a list or organic search. Fee rates are on Shopee's seller help page.
3. Judge ads on your own orders, not the ad platform's dashboard. Ours over-reported by almost double. Our other two stores are in the MaoMuse case and the ONAÏ case, and everything is collected in our e-commerce hub.

Want to know which stage your store is stuck at? Send us your URL and last month's order count on Telegram: https://t.me/eric16888999