Categories and storefronts

Four surfaces describe who sells what and where it sits in the tree.

Categories (browse nodes)

amz category resolves a browse node by id or URL into its name, the other nodes it links, its merchandised shelves, and the ASINs on the landing page:

amz category 172282                   # by node id
amz category "https://www.amazon.com/b?node=172282"
amz category 172282 --related         # the nodes it links, with their names
amz category 172282 --top -o url      # just the top ASINs as URLs

Each Category carries node_id, canonical_node, name, slug, related, shelves, top_asins, and item_count. Walk outward by feeding the ids in related back in.

There is no parent and no breadcrumb, because a /b page publishes neither. Amazon links a browse page to its children and its siblings with the same markup and gives no marker for which is which, so the field is called related and carries references rather than pretending to be a tree. Header and footer links are excluded: eight of the twenty four node links on the Electronics page are Gift Cards, Sell, Registry and the rest of the site chrome.

The edge upwards comes from a product instead. Each entry in a product's ranks links the node it is ranked in, which is an identifier Amazon stated rather than a name lifted from a breadcrumb.

Brand storefronts

amz brand takes a bare name, a slug, or a /stores/ URL:

amz brand anker
amz brand "https://www.amazon.com/stores/Anker/page/D24FDA17-..."
amz brand anker --featured            # the featured ASINs

A bare name costs three extra requests, and it is worth knowing why. Amazon puts a brand store at /stores/<name>/page/<uuid>, nothing derives that uuid from the name, and /stores/anker is a hard 404 of 1,147 bytes. There is no lookup endpoint and no redirect from the short path.

The only public page that states the uuid is the byline link on a product the brand sells. So amz resolves a name the way a person does: search the name, open up to three organic results, and follow the byline on the first one whose brand folds to the brand that was asked for. Sponsored cards are skipped and the name has to match exactly, because handing back a competitor's storefront under the name you typed is worse than handing back nothing.

The literal path is tried first, so a slug or a URL that already carries the uuid costs one request and resolves nothing.

A Brand carries slug, page_id, name, description, logo_url, banner_url, canonical_url, featured_asins, nav, and widgets. The slug is the identity and the page id is not: a storefront is seven pages and each one carries its own uuid, so a record keyed on the page id would give one brand seven identities.

nav is the field a crawl wants. A storefront landing page is a navigation page rather than a catalogue, and nav is the seven sub-pages the products are actually on.

follower_count is declared and comes back empty on every storefront measured. Amazon renders a Follow button with the word Follow on it and no figure, so the number is not published, and the envelope says as much rather than the record reporting a brand nobody follows.

Seller profiles

amz seller reads a third-party seller's profile and feedback breakdown:

amz seller A1XYZSELLER22
amz seller "https://www.amazon.com/sp?seller=A1XYZSELLER22" -o json

A Seller carries seller_id, name, business_name, business_address, about, storefront_url, shipping_policy, return_policy, rating, rating_count, positive_pct, feedback, rating_histogram, and reviews.

rating and rating_count are the twelve month window, which is the one Amazon puts in the header. feedback keeps all four windows, because a seller at 5.0 over 44 ratings in twelve months and 4.8 over 6,365 for its lifetime is two facts and one number cannot carry them. business_name is the legal entity behind the display name, and the two are usually different: the storefront trading as SIKAI CASE is Hangzhou Hang Kai Technology Co.,Ltd.

reviews is the first page only. The rest come from an AJAX call to /sp/ajax/feedback and are not in the HTML at any depth, so five reviews beside a rating count of 6,365 is a first page rather than a quiet seller.

Author pages

amz author reads an Author Central page:

amz author jrr-tolkien
amz author jrr-tolkien --books -o url  # the author's book ASINs

An Author carries slug, page_id, name, bio, photo_url, website, about_url, books, book_asins, total_books, sort_options, and languages.

books holds whole product records rather than bare ASINs, because the grid publishes the title, the price and the rating for each one and dropping them would cost a request each to get back. total_books is what Amazon states, which is larger than the length of books whenever the grid paginates: the measured page reads 70 and 135. sort_options and languages are the grid's own filter vocabulary and are the parameters that page through the rest.

Compose

Walk a brand's featured items into full records:

amz brand anker --featured --fields asin -o csv --no-header \
  | xargs -I{} amz product {} -o jsonl