Testing Methodology & Dataset
Every performance number on this site — the 34.6% broken-image rate, the 19.2% variant mismatch rate, the 91.2% time saving — comes from one documented protocol. This page publishes that protocol in full, including its limitations.
Study lead: Zi Dong Liang · EasyCatch Architecture Lab · August 2026
Bottom Line Up Front: In a controlled August 2026 comparison across 500 real 1688 products, the manual 1688-to-Shopify workflow produced a 34.6% broken-image link rate and a 19.2% variant-name mismatch rate at 118.2 seconds per SKU. EasyCatch's local DOM sniffing cut variant mismatch to 0.2% and export time to 10.4 seconds (a 91.2% reduction).
1. Sampling Frame
The dataset is 500 distinct 1688 product listings, drawn across three category clusters that together cover the overwhelming majority of cross-border dropshipping demand. Listings were selected from 1688 search and category pages without regard to whether EasyCatch performs well on them, and each listing had to satisfy three inclusion criteria: a live product page at capture time, at least one structured variant axis (Color, Size, Material, or Configuration), and at least four distinct SKU rows.
| Category cluster | SKUs | Why it is in the frame |
|---|---|---|
| Apparel & Accessories | 200 | Deepest variant matrices (Color × Size, often 30+ rows) and the highest rate of Chinese option labels. |
| 3C / Consumer Electronics | 150 | Configuration axes (plug type, memory, bundle) plus the heaviest CDN thumbnail-suffix usage. |
| Home & Kitchen Goods | 150 | Often a single variant axis but large image counts (5–12 images per SKU), stressing image handling. |
2. Metric Definitions
Each headline figure measures exactly one thing. They are reported separately because they have different causes, different fixes, and are frequently — and incorrectly — quoted interchangeably.
34.6% — Manual broken-image link rate
Share of manually imported SKUs where at least one image URL failed to resolve to a live image on the target storefront after import. This is an image-link failure, not a variant-mapping error.
Observed causes
- Alibaba CDN hotlink protection returning 403 when a copied URL carries a foreign referrer.
- Thumbnail-suffix downgrades: grabbing the
_50x50.jpgor.summ.jpgvariant instead of the original asset. - Expired or session-bound
cbu01.alicdn.com/img.alicdn.compath tokens captured mid-session.
19.2% — Manual variant-name mismatch rate
Share of manually prepared SKU rows where the mapped variant label (Option1 / Option2) did not match the variant image or the source page's authoritative mapping — wrong colour paired with the wrong photo, duplicated rows, or missing SKUs.
Root cause is multi-axis cross-referencing: translating 颜色分类 and 尺码 labels by hand across dozens of rows, then re-pairing each row to the correct image, is where human transcription degrades.
0.2% — EasyCatch variant mismatch rate
The same row-level measurement applied to EasyCatch exports produced by local DOM sniffing plus automatic variant-to-image pairing. The residual 0.2% is dominated by source pages whose own HTML carries an inconsistent mapping — the extension reproduces the page faithfully rather than guessing.
91.2% — Time reduction per 20-variant SKU
Capture-to-import time for a 20-variant listing fell from 118.2 seconds manually to 10.4 seconds with EasyCatch — a 91.2% reduction. The manual figure covers saving each image, renaming files to match variants, uploading to Shopify Files, and copying URLs into the CSV.
100% — Alibaba CDN original-image hit rate
Every image resolved by the auto-sniffing pipeline pointed at the original-resolution Alibaba CDN asset, with no downscaled thumbnail substitutions. This is the metric that drives the broken-image rate toward zero.
3. Control Protocol & Test Environment
The comparison is a within-subject design: the same 500 listings were processed twice, once entirely by hand and once with EasyCatch, and the resulting CSVs were imported into the same Shopify stores through Matrixify. Holding the source listings and the destination stores constant isolates the variable under test.
The "manual" arm is defined as:
- A competent operator — not a novice — working without automation.
- Opening each 1688 product page in a browser, saving images one at a time.
- Renaming files to match the intended variant rows by hand.
- Uploading to Shopify Files and pasting URLs into a Matrixify CSV template.
- Translating and mapping Chinese option labels manually.
A measurement is counted as a broken image if the imported row's image URL does not serve a valid image on the storefront after import. A row is counted as a variant mismatch if its (Option1, Option2) pair does not agree with the variant image assignment on the authoritative 1688 source page.
| Collection window | August 2026 |
|---|---|
| Browser baseline | Chrome stable channel, desktop |
| Destination stores | 12 Shopify development stores, imported via Matrixify |
| Extension build | EasyCatch production build, local-first (no server transit, no uploaded content) |
| Timing method | Wall-clock, capture to completed CSV export, per SKU, averaged across the set |
| Study lead | Zi Dong Liang, EasyCatch Architecture Lab |
4. Limitations & How to Read These Numbers
- 500 SKUs is a sample, not a census. It is large enough to be directional and to separate a 34.6% failure mode from a 0.2% one, but it does not describe 1688's entire catalogue.
- The category mix is deliberate and skewed. Apparel is over-weighted (200 of 500) because its variant matrices are the deepest — the hardest case, not the average case.
- Timing is hardware- and network-dependent. The 118.2s and 10.4s figures come from one fixed desktop baseline; absolute seconds will differ on your machine even though the ratio is stable.
- The "manual" arm reflects a competent operator. A novice would be slower and less accurate, so 34.6% and 19.2% are a floor for human error, not a ceiling.
- The web is not static. 1688's DOM structure and CDN policy change over time. Figures describe August 2026 behaviour and are re-measured when the pipeline changes.
- This is a vendor-run study, published for refutation. We are the authors of the tool under test. That is exactly why the raw protocol — sampling frame, control definition, and measurement rule — is published here rather than only the conclusion. If you can reproduce or contradict these results, we want to know.
5. Where These Numbers Appear
Every benchmark figure quoted across this site resolves back to this page. If you see a metric without a link here, treat it as unverified.
- • 1688 to Shopify Sourcing Hub — benchmark data block and comparison table
- • Import 1688 to Shopify — The Complete Guide
- • Shopify CDN Image Auto-Sniffing Guide — image-failure rate
- • Matrixify CSV Variant Matching — variant mismatch rate
- • Verify 1688 Suppliers & Negotiate MOQ
- • EasyCatch vs DSers vs CJ Dropshipping
- • Multi-Store Shopify Sourcing Automation
- • Manual CSV vs Extension vs Bulk Tool — per-SKU import time and broken-link rate
- • 1688 Payment Methods for Foreigners — payment rail comparison
Reproduce it on your own catalogue
The protocol above is the one the extension runs internally: local DOM sniffing, automatic variant-to-image pairing, and a Matrixify-ready CSV — with no data leaving your browser.
Install EasyCatch — FreePrefer to run the numbers first? Open the free sourcing tools →