Ecommerce PPC consultant who tests on our own stores first
Updated
TL;DR: We spend our own money on Meta, Google and ChatGPT ads for three stores we run, and we publish what happened, losses included. On our dried-peach store, Meta reported 65 purchases while our order data showed 34 that actually came from Meta. By our estimate that put real ROAS at about 0.78, not the much healthier number the platform implied. So we run every account the same way: reconcile with your real orders first, set a stop-loss before a test starts, then scale only what holds up.
What we do
Account audit against your real orders
Before touching budgets, we line up what Meta and Google report against paid orders in your store backend. If the platform claims twice the sales you can see, any scaling decision built on its numbers is wrong.
Meta ads
Campaign structure, creative testing, audiences and the Conversions API (CAPI) so purchases are reported from your server, not just the browser.
Google Ads
Search, Shopping (including your Merchant Center product feed) and, where it fits, Demand Gen. We match campaign types to how close the audience is to buying.
Small, bounded tests of new channels
New placements like ChatGPT ads get a fixed budget, a written stop rule, and a funnel we can actually see before the first dollar goes out.
Break-even math before scaling
We work out the cost per order you can afford from your margin, then judge campaigns against that number, not against platform averages.
Proof from our own stores
All three stores are ours and sell in Taiwan. Figures were pulled from our ad-spend sync and order database on 2026-10-07. NT$ amounts convert at the NT$31 to US$1 rate we use internally.
Taolilife (dried peaches): Meta, with the losses
| Metric | Number |
|---|---|
| Meta spend, Aug 15 – Sep 27, 2026 | US$1,309.76 |
| Purchases Meta reported | 65 |
| Paid orders in our backend attributed to Meta | 34 (NT$31,550) |
| Cost per Meta order (estimate) | about NT$1,194 |
| Average order value | about NT$928 |
| Meta ROAS on real orders (estimate) | about 0.78 |
| Break-even cost per order, from our financial model (estimate) | about NT$248–288 |
How we got there: US$1,309.76 × 31 ≈ NT$40,603 in spend. 40,603 ÷ 34 orders ≈ NT$1,194 per order. 31,550 ÷ 40,603 ≈ 0.78. Every Meta-driven order cost more than it brought in, and the cost per order sat roughly four times above our break-even estimate.
We're not presenting this as a success. It's the reason we reconcile first: if we had trusted the 65, we'd have raised the budget. The write-up of where the money went is in our Meta reported vs real orders post and the Taolilife case study.
Two new-channel tests, zero orders each
| Channel | Spend | Impressions | Clicks | Orders |
|---|---|---|---|---|
| ChatGPT ads (Sep 27–29) | US$171.98 | 14,815 | 207 | 0 |
| Google Demand Gen (paused Sep 30) | NT$1,899 (about US$61) | about 45,000 | 778 | 0 |
Demand Gen's on-site funnel: 497 product page views, 9 add-to-carts, 14 checkout starts, 0 purchases. For comparison, an early Meta window (Sep 3–8) produced 63 landing page views and 7 add-to-carts, about 11%, against Demand Gen's roughly 1.8% from product view to cart.
Both samples are small (one product, a few days, a few hundred clicks), so we don't treat them as verdicts on either platform. What they did teach us: Demand Gen told us where it failed because we had the funnel. For ChatGPT ads our record has no on-site funnel at all, so 207 clicks bought us a zero and not much else. Every test we run now gets its funnel set up first.
Maomuse (pet portraits): stopped in week two
US$369.04 on Meta between Jul 27 and Aug 8, 1,836 clicks, two confirmed paid orders worth NT$663. That works out to a ROAS of about 0.06 by our estimate. The plan had set a target of 1.3 over four months; we stopped in week two instead of spending the rest of the budget hoping it would turn.
Platforms we support
Our own stores are custom builds. For hosted platforms, the notes below come from each platform's official documentation as of 2026-10-07, not from our own hands-on work on them. We check your admin before we scope anything.
| Platform | Ads-relevant setup per official docs | What to know first |
|---|---|---|
| Custom build / WooCommerce | Anything: Meta's official WooCommerce plugin includes CAPI; Google's official plugin syncs to Merchant Center | Someone has to maintain plugins and hosting |
| Shopify | CAPI through the Facebook & Instagram app; Merchant Center through the Google & YouTube app | Google Tag Manager only runs in a sandboxed Custom Pixel, which Shopify doesn't support and Tag Assistant can't detect. In Taiwan, a 0.2–2% third-party payment fee applies to every order |
| SHOPLINE (Taiwan) | Meta Pixel only through Facebook Business Extension, which turns CAPI on automatically; Merchant Center supported | 2% of sales (1.5% on the top plan) goes to the platform (pricing page), so ROAS needs to be calculated after that cut |
| CYBERBIZ (Taiwan) | CAPI through Facebook Business Extension; Merchant Center; Google Ads enhanced conversions guide | "+" plans take 1% of sales and Enterprise 5%, per the official comparison table (confirm on your quote) |
| EasyStore | CAPI turns on automatically when Facebook is connected; one-click Merchant Center sync | No sales commission, so platform costs don't rise as ads scale |
| WACA (Taiwan) | Pixel + Conversions API, Merchant Center, LINE ads tags listed | No sales commission |
| meepShop, SHOPSTORE (Taiwan) | Pixel and shopping feeds listed | No CAPI on their feature lists; server-side tracking has to be added separately, with its own hosting cost |
| 91APP (Taiwan) | Not publicly documented | The platform sells its own ad services, so check your contract first |
What we won't promise
- No guaranteed ROAS or revenue. Our own dried-peach store is at about 0.78 on Meta. We'd be lying if we promised you a number.
- No scaling on platform-reported numbers. Budgets move on reconciled orders.
- No open-ended tests. Every new channel gets a budget cap and a stop rule written down before launch.
- No store operations. Listings, customer service and promotions stay with you.
FAQ
- What does an ecommerce PPC consultant do differently from an agency running my ads?
The way we work, the first job is checking whether the numbers you're optimizing toward are real. Ours weren't: Meta reported 65 purchases where our backend showed 34. After that it's campaign structure, creative, feeds and break-even math.
- Why would Meta report more purchases than my store received?
Meta counts purchases that happened within its attribution window after someone saw or clicked an ad, while your store records one source per order. Duplicate events between the browser pixel and server-side CAPI can inflate it further. The gap is normal; how big it is decides whether you can trust the platform's ROAS.
- Are ChatGPT ads worth testing for ecommerce?
We spent US$171.98 over three days for 207 clicks and no orders, but the sample was tiny and we didn't have on-site funnel data, so we can't call it either way. If you test, set up the funnel first and cap the budget.
- Is Google Demand Gen good for online stores?
For our single-product store it wasn't: 497 product views, 9 add-to-carts, 0 orders for NT$1,899. Demand Gen runs on browsing placements like YouTube, Discover and Gmail (Google's overview), so it reaches people who weren't searching for you. That can work for some products. It didn't for ours at that stage.
- What's a good ROAS for ecommerce?
It depends on your margin. We calculate the cost per order you can afford from your own numbers first. For our dried-peach store the estimate was about NT$248–288 per order, which is why NT$1,194 was a clear loss.
- Do you work with Shopify stores?
Yes. The main catch is tracking: Google Tag Manager only runs inside Shopify's Custom Pixel sandbox (Shopify's GTM guide), so we set up tracking with that limitation in mind.
- Do you only work with stores in Taiwan?
No. Our stores are in Taiwan, but Meta and Google work the same way for stores selling elsewhere.
- How is pricing set?
Each engagement is quoted on your store's situation. Tell us what you sell, your monthly ad spend and which platforms you're on.
Want to know what your ads are really bringing in?
Send us your store, platform and ad accounts on Telegram. We'll start by putting your platform numbers next to your real orders.
Related: All ecommerce services · SEO consultant · Analytics consultant · Our three stores