Read 1688 wholesale offers, price ladders and suppliers as JSON

One offer returned three bands — 1 to 99 at ¥19, 100 to 9,999 at ¥18, 10,000 and up at ¥17 — with each band's upper bound derived from where the next one starts. That matters because the headline price on a wholesale listing is the single-unit price, which is the one quantity almost nobody buys.

Against us until you know the rule. A second offer returned price_tiers as an empty list, and that is correct for it: that offer does not discount by quantity, it prices per variant. Its spread lives in variants.skus[], where the same cable ran from ¥1.64 to ¥2.67 depending on the connector and the length. Read the ladder when it is there and the SKU rows when it is not.

Against us. On one offer every SKU carried its own stock figure — 1,877, 392 and so on per colour — and a null price, because that offer prices from the ladder rather than per variant. On another, every SKU carried both. Treat variants.skus[].price as optional and fall back to price_tiers; the stock number is the one that was present on everything we opened.

Against us, and it reached customers. The rate fields were published multiplied by a hundred, so a 52% repurchase rate went out as 5200. Fixed and re-measured for this page: across a full result set the highest value was 50.0 and nothing exceeded 100. offer_repurchase_rate was present on 15 of 20 rows and shop_repurchase_rate on 19 of 20 — a missing rate means 1688 published none, not that it is zero.

Keyword search answers with one page of twenty and says so: has_more comes back false and no total is published. The category surface is where the depth is — thirty offers a page, and pages one, two and three shared zero offer ids between them in a direct comparison. It also carries fields the keyword feed does not: certifications, fulfilment, the minimum order quantity, the listing date and a sales figure.

currency is CNY on every row and there is no country or currency parameter, because 1688 is a domestic Chinese wholesale market rather than an export storefront. The unit field carries the measure word the seller quoted in — 个 for a piece, 条 for a strand or cable — and price_display repeats it as ¥19.00/个. Converting the number without carrying the unit is how a per-cable price becomes a per-box price.

Against us, and it is the weak corner of this engine. Two offers reporting 1,100 and 600 reviews each returned exactly two review rows. The aggregate is solid and arrives in one call — the star rating out of five, the positive-feedback percentage and the total count — and each row that does come back carries the buyer, the text, the date, the quantity bought and the variant it refers to. If you need the review corpus rather than the score, this endpoint is not yet it.

Against us, and stated plainly because you should not plan around it today. supplier/search answered successfully but with zero rows for three separate keywords, in Chinese and in English. The response says why in its own note: 1688 publishes a curated shortlist of vetted factories per keyword rather than an exhaustive index, so an empty list is a real answer about the keyword. It is still an endpoint we would not build a workflow on right now.

54 top-level categories over three levels, and every node carries both the English name and the Chinese one, plus a child count so you know whether to descend. This is the part that makes 1688 approachable without reading Chinese, and it is one credit.

Search rows carry the company name, the login id, the member id and how many years the seller has been on 1688 — that last one is the cheapest trust signal on the platform. The detail block adds the member and user ids and the storefront URL, but city and province came back null on every offer we opened, so do not plan to filter suppliers by province from this field.

Search and product/detail are two credits; category/offers, categories, product/reviews and supplier/search are one. Keyword search answered in one to two seconds, a category page in about two, the category tree in under four, and a full offer detail — which is two upstream calls merged — in three to seven.