Scrape Taobao and Tmall into clean structured datasets — product details, keyword search, shop catalogs, customer reviews and buyer Q&A. Five operations in one actor, one clean dataset per run. No account, no API key, no setup.
All notable changes to Taobao & Tmall Product Scraper will be documented in this file.
[2026-08-28]
📊 One dataset view, and two fields that make every row traceable
The dataset now opens straight to one complete Overview — the separate "Key fields" and
"Alternate Source" tabs are gone. Every field from every operation lives in Overview now, so
there's one place to look instead of three.
Two new fields on every row: processedAt and scrapingDate — a timestamp and a plain
date for when the actor processed that row, independent of recordTime (when the source data
was actually captured). Use them for freshness audits and slowly-changing-dimension joins.
Product images and source URLs now render as clickable links in the dataset view instead
of raw text — images, picUrl, picUrlFull, picUrlList, detailUrl, and imageList all
open directly.
💎 User Benefits
One dataset view to learn, not three.
A clear answer to "when was this row processed," without cross-referencing the run log.
Faster QA — click a thumbnail instead of pasting a URL into a browser.
🎯 Use Cases
Data teams wiring processedAt into a nightly freshness check without touching recordTime.
Analysts scanning a shop catalog by clicking straight through to each product image.
[2.7.0] — 2026-08-26
⚡ Product Detail is reliable again
A new default detail version.Product Detail now leads with a faster, far more consistently
available profile (v3 ⭐ Recommended). Over 24 hours of availability sampling it was reachable in
21 of 24 hours, against 10 of 24 for the previous default — and it responds in roughly a quarter of
the time. If it cannot serve an item the actor still falls back to Standard and Full, exactly as before.
Stock and units sold are finally populated.totalStock and soldCount were empty on every
Product Detail row before this release — the previous default did not report them as numbers. The new
default fills both, as real integers.
The image gallery is populated too.pictUrls was empty on every Product Detail row, on every
version. It now carries the full gallery on Recommended, Standard and Full.
Prices are current, not expired. The previous default reported whatever promotion the listing last
carried — on a sample of 24 products, 13 had a promotion window that had already ended, some three
months earlier. The new default reports the live listed price. If you specifically want the
list-vs-sale spread, pick v1 Standard and check its startTime / endTime.
Wrong numbers are now left blank instead of guessed. Some versions report stock and sales as
rounded text ("1万+"). Converted naively that becomes 1 — a product that sold ten thousand units
reading as one. Those columns are now left empty on the affected versions and the raw text is passed
through instead.
💬 Clearer failure messages
"This item is not supported" is no longer reported as an error on our side — it means the item ID
could not be retrieved, and the message now says so and tells you what to check.
Plainer language on every failure, with the retry advice unchanged. Failed items are still never charged.
[2.6.1] — 2026-08-26
⭐ Fixed — review star ratings were the same number on every row
reviewRatingStars now carries the real 1-5 rating. It was reading an internal validity flag that reports 1 for every review, so the column looked filled in but told you nothing — you could not sort by rating or filter for negative reviews. It now reads the actual per-review score. If you have stored review data from before this release, that column needs re-fetching.
[2.6.0] — 2026-08-24
📄 A run report you can actually use
See your first result without leaving the report — the top of every run report now shows one complete row as copyable JSON, so you can check the field names your pipeline needs before you open the dataset.
A preview table built for the operation you ran — reviews show stars, text, the variant bought and the buyer; searches and shop catalogs show product, price, sales and shop; product detail shows the facts that version returned. No more guessing which columns apply.
Failed items come with a one-click retry — every failure now carries a plain-English reason, a concrete fix, and a Copy retry input button that hands you the exact JSON to re-run. Failed rows are never charged, and the report says so.
An itemized statement of what you paid — every charge on the run, at the price your plan actually pays, with the total. Shop Catalog duplicates are dropped before charging and appear as a separate number.
Run it again on a schedule, or from an AI agent — the report links straight to a schedule for this exact input, and ships a ready-made MCP config so Claude, ChatGPT or your own agent can run the Actor in plain language.
Chinese product names render correctly — the report is now served as UTF-8, so titles, shop names and review text no longer arrive as mojibake.
🖼️ Product images in the dataset view
Thumbnails instead of URLs — the Overview, Key fields and Alternate Source views now render the main product image as an actual picture, so you can scan a catalog visually instead of pasting URLs into a browser.
Readable field names — 36 raw column names (Discnt Price Yuan, Fahuo Dsr, Pic Url Full, …) now carry proper labels and icons in the dataset schema.
💎 User Benefits
Less time spent working out what a run produced, and what it cost.
A failed item is now a two-click fix instead of a support question.
Visual catalog scanning for sourcing and competitor work, straight in the Apify Console.
🎯 Use Cases
Dropshippers checking SKU pricing and stock before ordering, straight from the run report.
Sourcing teams scanning a Tmall shop catalog by product image rather than by ID.
Data teams wiring the Actor into an AI agent with the MCP config the report hands them.
[2.5.0] — 2026-08-15
🔄 New operation: Product Detail — Alternate Source
New: productDetailAlt — the same product lookup served from a different source, for items the standard detail versions cannot return. In testing it resolved products that returned "This item is not supported" on v1. Returns a compact flat row: price and CNY price, total stock, per-SKU variants, images, description, brand and tags. Comes with an Alternate Source dataset view.
Billed at the same per-result rate as Product Detail — it is a coverage option, not an upsell. No pricing change in this release.
It is a separate operation, not a new detailVersion. It comes from a different source whose response shares no fields with the Product Detail versions, so folding it in would have changed Product Detail's columns depending on which version answered. Filter on _operation to tell them apart. Product Detail is completely unchanged by this release.
itemId is stamped from your input and detailUrl is rebuilt, because this source returns neither.
It also passes through import-oriented fields (exchangeRate, expressFee, startQuantity, batchPrice). These were empty or trivial on every item tested, so they are documented as best-effort rather than advertised.
🐛 Fixed
Telemetry recorded the wrong detail version. Runs that did not set detailVersion were fingerprinted as v9 while the actor actually requested v1 — skewing the per-version data used to decide which version to lead with.
[2.4.1] — 2026-08-14
⚡ detailVersion: v5 (Cached — Lite) is retired
The Lite version is no longer available. Every request to it came back rejected, so a run that asked for Lite could only ever produce an error row.
If you have a saved task set to Lite, it keeps working — it is now served by Fresh — Standard (v1) instead, and the option stays in the list purely so those tasks do not break. ⚠️ Standard returns different field names from Lite, so check the field matrix in the README if your pipeline pinned a Lite-only field (shop_id, sku_props, desc_url, …). Do not pick Lite for new work.
Product Detail stopped dead-ending on unavailable items. Lite sat at the end of the automatic fallback chain, so any item the other versions could not serve finished on a cryptic rejection message for what is really "no version covers this item". The chain is now Standard → Full, and an exhausted item returns the honest "This item is not supported."
A version that reports the item is unavailable to it now falls through to the next version instead of ending the attempt.
🐛 Fixed
The run log named the wrong detail version. Runs that did not set detailVersion were logged as "Full (fresh)" while the actor actually requested Standard.
[2.4.0] — 2026-07-25
🎯 Product Detail reliability — and honest per-version documentation
⚠️ Behaviour change — the default detailVersion is now v1 (Standard), was v9 (Full).v9 has been returning effectively no data for weeks, so leading with it meant most Product Detail runs spent their whole retry budget on a version that could not answer. If your pipeline depends on v9's flat field shape, set detailVersion: "v9" explicitly — but note it is currently unreliable. Runs that never set detailVersion will now receive v1-shaped rows (different field names — see the field matrix in the README).
Product Detail now recovers instead of giving up. When a detail version cannot serve an item, the actor falls through to the other standard versions rather than failing the row. Previously only "item not supported" triggered the fallback; a timeout ended the run empty. Recovered rows are complete rows, and failed attempts are never billed.
Fixed: Standard (v1) rows were missing most of their content. Specifications, description HTML and the whole seller/shop block were silently dropped, and the price aliases were empty. All are now populated — a v1 row went from ~20 usable fields to ~50.
📚 New: a field × version matrix in the README. Each detail version returns a genuinely different field set — only itemId, title and images are common to all of them, and v5 uses snake_case where v1 uses camelCase. The matrix is generated from live live payloads, and the previous output field list (which advertised fields the actor never emitted) has been replaced.
📚 Documented the automatic fallback, so it is clear that detailVersion is a preference and the version that answers determines the field names. The 💎 Premium version is never substituted in either direction — a standard request can never be upgraded into premium billing.
📚 Removed the v2 "async premium" option from the docs. It was never selectable: it is offered only through its own dashboard task page, and calling it over the API returns no data. Use v4 for premium final pricing.
[2.3.0] — 2026-06-19
🚀 Five operations — premium final pricing and buyer Q&A
💎 Premium Detail — Final Payable Price — a new premium product-detail version returns the exact after-coupon final price per product and per SKU, so you can calculate true margins and re-price with confidence. Priced at $0.20/result, effective 2026-07-02.
❓ Product Q&A — pull real pre-purchase buyer questions and answers for any product, plus the top buyer-concern tags — ideal for listing optimization and customer-intent research. Priced at $0.008/result, effective 2026-07-02.
🏪 Bigger Shop Catalogs — an additional catalog mode for the latest, largest seller listings.
🔍 Keyword Search up front — keyword search is now the default operation: the fastest way to discover products by query, price, and rating.
💎 User Benefits
Know your real cost — final after-coupon pricing replaces guesswork for sourcing and re-pricing decisions.
Understand demand before you buy — buyer questions reveal what shoppers actually care about before they purchase.
More coverage, one actor — products, prices, shops, search, reviews, and Q&A in a single tool.
🎯 Use Cases
Dropshippers validating true landed cost with after-coupon final pricing before listing a product.
Sourcing teams mining buyer Q&A to pre-empt customer questions and sharpen product descriptions.
Listing optimizers using top buyer-concern tags to rewrite titles and descriptions that convert.
[2.2.0] — 2026-06-17
🕒 Data Freshness — capture timestamps + clearer detail versions
🆕 New recordTime field on every row — the exact capture timestamp (when the data was actually scraped), separate from _fetchedAt (when your run executed). Use recordTime to verify how fresh each row is. Surfaced on Product Detail (all versions), Keyword Search, Shop Catalog, and Reviews.
🔁 Product Detail versions relabelled by freshness — the detailVersion option now states clearly which versions are fresh vs cached:
v9 Full — live, non-cached. New default.
v1 Standard — live, non-cached. Slightly different coverage; combine with v9 to maximise coverage.
v5 Lite — ⚠️ cached (data can be days to months old, no reliable capture time). No longer recommended when freshness matters; use only when stale data is acceptable.
📖 Docs — README adds a "Data freshness & detail versions" section, corrects the freshness FAQ, and documents the optional async detail versions (fuller campaign pricing, available on request).
📊 Dataset views — added an Overview view (all fields) plus a curated "Key fields" view that includes recordTime.
💎 User Benefits
Provable freshness — every row tells you exactly when it was captured.
No more surprise stale data — the cached version is clearly marked; the default is now a fresh version.
[2.1.0] — 2026-05-07
🔧 Bug Fixes + Enhanced Pagination Metadata
🗑️ Removed Shop Catalog v3 — the v3 option is not available. The catalogVersion option now only supports v1 (Standard) and v2 (Extended).
📊 New Pagination Metadata — all paginated operations now include total count and page metadata in every row:
📖 New REFERENCES.md — added example inputs and outputs for every operation, including sample fields and pagination metadata
💎 User Benefits
More predictable pagination — see total results upfront without extra API calls
Cleaner input options — v3 option removed to prevent confusion and errors
Better documentation — concrete examples for every operation help users understand data structure before running
[2.0.0] — 2026-05-06
🚀 Major Upgrade — Direct API + v9 Detail + v2 Shop Catalog
🆕 Product Detail v9 — flat response with coupon-adjusted pricing (afterCouponAmountPrice), full product attributes, specs map, SKU images, seller type, and promotion model data. The most complete product detail available.
🆕 Shop Catalog v2 — 60 items per page (2x v1's 30), five sort options (comprehensive, hot selling, new arrivals, price ascending, price descending), and sell-point highlights per item. Requires shopId in addition to userId.
🔀 Sort Order Support — keyword search and shop catalog results can now be sorted by sales, price, comprehensive ranking, hot selling, or new arrivals via the new sort input field.
💬 Review Sort — product reviews can be sorted by date (feedbackdate) or relevance (general) via the new orderType input field.
⚡ Faster responses — roughly 2s per page, down from ~5s. Simpler and more reliable.
🛡️ Clearer Error Messages — when a row can't be retrieved, it now says what happened and what to try next instead of showing a raw error code.
💎 User Benefits
60 items/page with v2 — scrape full shop catalogs in half the time and cost. A 1,000-item shop now needs ~17 pages instead of ~34.
Coupon pricing in v9 — see the actual price after coupons and promotions, not just list price. Critical for accurate margin calculations.
5 sort options for shops — find bestsellers, newest arrivals, or cheapest products directly. No need to sort 3,000 rows client-side.
Faster responses — direct API calls cut ~3 seconds of latency per page compared to the previous architecture.
🎯 Use Cases
Dropshipping operators using v9 detail for coupon-adjusted margin calculations and full SKU attribute analysis
E-commerce teams switching to v2 shop catalog for 2x faster competitor catalog sweeps with price sorting
Market researchers using sort options to find hot-selling and new-arrival products without post-processing
Data teams leveraging v9's structured attributes and specs for automated product categorization pipelines
[1.0.0] — 2026-04-27
🎉 Taobao & Tmall Product Scraper — Launch!
🛍️ Four-in-One Scraper — Product detail lookup, keyword search, shop catalog sweep, and customer review extraction — all in one actor, one operation per run
🎨 SKU Variant Intelligence — Per-SKU prices, stock, property paths, and swatch images for every color/size combo — perfect for dropshipping decisions
🔍 Multi-Language Keyword Search — Search Taobao and Tmall with Chinese, English, or mixed queries, plus price range and Tmall-only filtering
🏪 Full Shop Catalog Sweep — Pull a seller's entire product list by shop ID — ideal for brand mapping and competitor intelligence
