IKEA Products Scraper — Prices, Ratings & Images

Reads IKEA products from the JSON endpoint ikea.com's own search calls. No key, no auth, no HTML parsing.

What you get

recordType	One per	Carries
PRODUCT	product	range name, product type and a joined full name, price with currency, rating and review count, measurements, main and contextual images, colours, category path, business structure, availability, and the product-page URL
SEARCH_SUMMARY	search term	IKEA's own match count, how much of it is actually reachable, and IKEA's related searches
ERROR	failed input	a named reason

There is no pagination, and the match count is a false promise

offset is not an optional extra — every value returns HTTP 400. The only lever is size, and it is capped:

size=1000  ->  1000 products  (7.5 MB of JSON)
size=2000  ->  HTTP 400

So a term whose count reads 9,231 will yield 1,000 products, ever. The honest total and the unreachable remainder sit in the same object, and nothing marks the difference.

Every summary reports totalMatching, estimatedReachable, estimatedUnreachable and hitReachabilityCeiling. Because narrowing the query is the only route to the rest, IKEA's own related searches are returned on the summary row — that's what they're for here.

The price is not where it looks like it is

The product object carries priceNumeral and currencyCode at its top level — exactly where a reader would go — and both are null on every row. The real values live under salesPrice:

product.priceNumeral           null      <- decoy
product.currencyCode           null      <- decoy
product.salesPrice.numeral     99.99     <- the price
product.salesPrice.currencyCode "USD"    <- the currency

A scraper reading the obvious fields produces a dataset where every price is null and nothing errors. This actor reads the real ones and also emits the decoys as priceNumeralDecoy / currencyCodeDecoy, so you can see for yourself they're empty rather than wonder whether something was dropped.

name is the range, not the product

IKEA names ranges, not products: name is "MICKE" and typeName is "Desk". Neither alone is a usable title — "MICKE" means nothing to a reader. Both are emitted, plus fullName ("MICKE Desk") for the common case.

The market selects a catalogue, not a translation

Same query, same day:

/us/en/  ->  196 products, USD
/gb/en/  ->  259 products, GBP
/de/de/  ->  528 products, EUR

So a run against another market is a different product set, not the same one relabelled. Market and language are recorded on every row, and product URLs come back market-correct. A market or language that doesn't exist is an honest HTTP 404 — /us/zz/ fails rather than silently falling back to English.

Honest, for the record

• a nonsense term → zero products with a count of 0, not an unfiltered baseline
• an empty term → a genuine browse-everything result (9,231 matches), not an error, exposed as the Browse the whole catalogue toggle
• an unknown types value → HTTP 400

Notes

• One request per search term. A size=1000 response is roughly 7.5 MB, so concurrency defaults to 2.
• sik.search.blue.cdtapps.com/robots.txt names no AI bot and carries no rule matching this path; www.ikea.com allows /search/. No WAF observed.