Best web scraping API for ecommerce: choose by target

An Amazon request costs 5 credits on ScraperAPI’s recorded target meter, while Shopify has no recorded target-specific meter at all. That contrast is why there is no honest universal winner. In the dated source records, Amazon has 8 published target-meter lines across 6 vendors, Walmart has 4 lines across 3 vendors, eBay has 2 lines across 2 vendors, and Shopify has 0. Those counts do not rank quality. They show whether a target-specific bill can even be attempted before plan prices, output units, access rules, and unknowns are checked. All four target records were reviewed on 2026-07-26.

The workload builder above starts with the decision that generic scraping comparisons hide: what entity is needed, how often it must be refreshed, and how many calls or delivered records each entity requires. A catalogue search, a product record, an offer list, a review set, and a rendered page are different products. If a job needs three endpoint calls per product and four refreshes each month, 10,000 products x 3 calls x 4 refreshes = 120,000 target calls, before retries or pagination. A vendor allowance cannot be evaluated until that multiplication is explicit.

The short decision rule

Check the official interface first when it returns the required data and the account is eligible to use it. Amazon Creators API can serve supported affiliate-catalog jobs, eBay Browse API can serve supported catalogue discovery, Walmart Marketplace APIs manage an approved seller’s operation, and Shopify Storefront API is specific to a store and its access model. None of those names means unlimited public-site access. An interface with no separately published per-call fee is also not automatically free: account, programme, rate-limit, licence, and product-plan constraints still apply.

If the official interface does not return the needed output, move to the target-specific commercial record. A vendor that documents an Amazon endpoint has shown target support, but it has not necessarily published a usable price. A vendor that publishes a target meter has shown what one operation consumes, but that meter still has to be combined with the eligible plan, purchase floor, billing term, and output unit. Only then can two bills be compared.

If neither path is complete, stop at unknown. Do not borrow the one-credit generic HTML rate for a protected marketplace, silently add a residential-proxy price from a different product, or describe an endpoint as cheap because the plan card begins at a low monthly price.

Amazon has the deepest price record, not a universal winner

The Amazon record contains the broadest set of target meters. ScraperAPI records 5 credits for an Amazon request. ScrapingBee records 5 credits for its light Amazon mode and 15 credits for normal mode. Scrape.do records 1 credit for an Amazon-domain request routed through its Amazon Scraper API. Oxylabs Web Scraper API records a starting Amazon result rate of $0.0005 per result on the cited plan. SerpApi records 1 search for an Amazon Search or Product call. ScrapingDog records 1 credit for an Amazon product request and 5 credits for Amazon autocomplete.

Those six vendors still do not sell identical output. A delivered result, search, product call, autocomplete call, and page request cannot be added into one ranking. The application may also need separate search, product, offer, seller, or review calls. That is why the utility asks for calls per entity rather than assuming one URL equals one product observation.

Amazon Creators API is the first path to check for supported affiliate catalogue work. The retained access record requires an Amazon Associates account, identity verification, and qualifying sales. It does not turn arbitrary rendered pages, review pages, seller pages, or every retail interface into the same operation. The recorded Conditions of Use also restrict automated extraction without Amazon’s written consent. A commercial scraping subscription changes the access technology, not the target’s terms.

Walmart separates seller operations from public retail collection

Walmart’s record has target meters from ScrapingBee, SerpApi, and ScrapingDog. ScrapingBee records 10 credits for Walmart light mode and 15 credits for normal mode. SerpApi records 1 search for a Walmart Search, Product, or Reviews call. ScrapingDog records 5 credits for a Walmart product request. Oxylabs Web Scraper API, Crawlbase, and Decodo Web Scraping API also document Walmart-specific endpoints or templates, but the retained target record does not convert every endpoint into a public target-specific bill.

Walmart Marketplace APIs are a different product. They are for approved Marketplace sellers and solution providers managing their own items, inventory, orders, prices, promotions, reports, and fulfilment. That can be the correct first route for an authorised seller workload. It is not a public shopper-catalogue API by implication. The public-retail job must therefore stay separate from the own-catalogue job in both copy and calculations.

The recorded Walmart terms restrict automated access without prior written permission. That constraint belongs before vendor selection. A target-specific endpoint and a valid purchase path do not establish that a proposed collection or reuse is permitted.

eBay may be an official-API job before it is a scraping job

The eBay Browse API is recorded as available for supported search and product discovery through an eBay developer account, OAuth, and the applicable production-access review. No separate per-call fee is published in the dated source record; call limits and policy still apply. For public catalogue search or known product details that fit its response, this is the first route to check.

The commercial record is thinner than Amazon’s. SerpApi records 1 search for an eBay Search or Product call, while ScrapingDog records 5 credits for an eBay search request. Oxylabs Web Scraper API and Crawlbase document eBay-specific targets or scrapers. These facts establish possible product paths, not a complete cheapest-vendor result. Output equivalence, purchase floor, pagination, geography, and required fields still have to match.

eBay’s recorded agreement restricts robots, spiders, scrapers, and other automated access without prior permission. If Browse API does not expose a required field, the next step is not automatically a commercial scraper. It may be an access request, a licensed source, a narrower output, or no collection.

Shopify is store-specific and currently has no target price winner

Shopify storefronts differ from the three marketplaces because each merchant controls a separate store and its terms. The dated source record has 0 target-specific surcharge lines and 0 dedicated commercial endpoint records for the generic Shopify target. That is an evidence result, not a zero-dollar result. Generic scraping products may fetch a particular storefront, but this dataset does not publish a Shopify-specific winner.

Shopify Storefront API is shop-specific. Tokenless access covers a limited field set, while broader access requires a token issued for that store. It is the first route to check for an authorised store catalogue. It is not a cross-store discovery interface. A project spanning many unrelated merchants must evaluate each store’s access, terms, rendering, pagination, and output separately.

How to turn the profile into a bill

First, freeze the output: public catalogue search, known product details, an authorised store catalogue, offers or sellers, reviews, or a rendered page. Second, count entities and refreshes. Third, count calls or delivered records per entity, including search pages, pagination, and separate detail endpoints. Fourth, add the vendor’s published target meter and only the plan eligible for that meter. Fifth, model billable failures only where the vendor publishes the rule. Finally, keep terms, data rights, and official-interface restrictions visible beside the number.

The target pages linked in the matrix contain the full dated evidence and the unknowns for each marketplace. The web scraping API pricing table is useful after the workload is fixed. The URL-list crawl guide is useful when every row really is one ordinary URL fetch. Neither should replace a target-specific meter where the vendor publishes one.

A defensible choice may therefore be an official API, a target-specific scraper, a generic fetch API, a data licence, or no collection. The point of the matrix is to eliminate the false comparison before it becomes a false price.