Amazon Sales Scraper

Scrape Amazon product search results and product pages to compare prices, reviews, category rank, and visible purchase activity. Export Amazon product data to CSV or JSON for catalog research, price tracking, and a practical Amazon product data feed without maintaining your own collection infrastructure.

The actor accepts search keywords, product URLs, or ASINs. For products with a visible “bought in past month” badge, it reports a conservative monthly sales floor and a revenue floor calculated from the displayed price. Those values are lower bounds, not exact sales or a claim about a seller’s private order data. A run with the garlic press keyword collected 30 products with full detail pages; 29 had a purchase-count badge at the time of the run.

What data can you extract from Amazon?

Each dataset item represents one product. Field names use camelCase. Search results provide the core catalog and purchase signals; enabling full details adds the product page fields that are visible for that item. The actor leaves a field empty when Amazon does not show it or when a product page cannot be retrieved after retries.

Field	Description	Example
asin	Product identifier	B098WNQ44M
url	Canonical product link	https://www.amazon.com/dp/B098WNQ44M
marketplace	Marketplace suffix	com
title	Product title	Kitessensu High Effective Garlic Press...
brand	Displayed brand or store name	KITESSENSU
price	Current displayed price, including currency	$8.99
listPrice	Displayed list or crossed-out price	$12.99 when shown
discountPercent	Discount computed from those two prices	31 when both prices exist
coupon	Visible coupon message	A displayed coupon, when present
rating	Average rating out of five	4.6
reviewCount	Displayed review count	16105
reviewDistribution	Percentages for one through five stars	{"5StarPct":77}
bestSellersRank	Visible category rank entries	[{"rank":913,"category":"Kitchen & Dining"}]
boughtInPastMonth	Purchase-count bucket shown by the site	4K+
estimatedMonthlySales	Minimum monthly units implied by that bucket	4000
estimatedMonthlyRevenue	Sales floor multiplied by current price	$35,960.00
salesEstimateMethod	Source of the calculation	boughtBadgeFloor
categories	Breadcrumb category path	Home & Kitchen, Kitchen & Dining
features	Product bullet points	Built to Last...
description	Visible long description	Text when available
specs	Product detail label and value pairs	{"Material Type":"Zinc"}
variants	Visible option ASINs and labels	{"asin":"B098WFBBJZ","dimensions":{"Color":"..."}}
images	Product image links	Image URLs
availability	Visible availability message	In Stock
stockLeft	Quantity from an “Only N left” message	3 when shown
soldBy	Merchant shown on the detail page	Merchant name
shipsFrom	Shipping party shown on the detail page	Amazon
isPrime	Whether a visible product badge is found	true or false
isFba	Whether the merchant link identifies site fulfillment	true or false
badges	Visible product labels	Amazon's Choice
relatedAsins	Other product identifiers linked on the page	B098WFBBJZ
searchKeyword	Keyword that yielded the item	garlic press
searchPosition	One-based position in search collection	1
sponsored	Whether the search card has a sponsored label	false
detailStatus	full or listing	full
scrapedAt	UTC collection time	ISO 8601 timestamp

The estimatedMonthlySales name keeps sales data easy to compare with product feeds that use that field, while salesEstimateMethod records what it means here. If a badge says 4K+ bought in past month, the actor sets the value to 4,000. Actual purchases may be higher. The revenue field then multiplies 4,000 by the current displayed item price. It excludes tax, shipping, returns, fees, discounts applied at checkout, and variation-level differences. If the badge or price is absent, the corresponding calculation remains empty. Rank alone does not produce a sales number.

How to scrape Amazon products and sales signals

1. Choose one source: search keywords for discovery, product URLs for a known shortlist, or ASINs for a catalog you already have. The actor rejects an input that mixes those source modes, which keeps one run’s scope clear.
2. Select the marketplace. The default is the United States marketplace. For keyword and ASIN modes, the selection controls the domain and proxy country. Direct product URLs keep the marketplace in each URL.
3. Set maxResultsPerKeyword and maxItems. A small first run makes it easy to check the category and purchase signals before collecting a wider set. Turn fullDetails off if search-card fields are enough.
4. Start the run. Results appear as one dataset row per unique ASIN. Open the dataset to review the fields or download it in a structured format for further analysis.

The collection stops at your configured cap, at the end of reachable search pages, or when the dataset budget is reached. Search results are deduplicated across pages and keywords, so a repeated ASIN appears once per run. Direct product inputs are also deduplicated.

Input

Parameter	Type	Default	Description
keywords	string list	empty	Search phrases; choose this mode for product discovery.
startUrls	URL list	empty	Direct product links with /dp/ or /gp/product/.
asins	string list	empty	Ten-character product identifiers for direct lookup.
marketplace	string	com	Domain suffix for searches and ASIN lookups.
maxResultsPerKeyword	integer	50	Unique products to collect for each keyword, up to 500.
maxItems	integer	100	Total dataset cap; zero removes the total cap.
fullDetails	boolean	true	Fetch each search result’s product page for detail fields. Direct lookups always need a product page.
proxyConfiguration	object	automatic	Optional proxy settings; normal runs start with datacenter sessions.

Example search input:

{
  "keywords": ["garlic press"],
  "marketplace": "com",
  "maxResultsPerKeyword": 30,
  "maxItems": 30,
  "fullDetails": true
}

For an existing catalog, replace keywords with an asins list such as ["B098WNQ44M", "B00HEZ888K"]. For product links, use startUrls with entries such as {"url":"https://www.amazon.com/dp/B098WNQ44M"}. These modes fetch product pages. A product that cannot be loaded or does not expose a title is skipped instead of becoming a blank dataset row.

Supported marketplace values are com, co.uk, de, fr, it, es, ca, com.mx, and in. The domains share the same input contract, but page labels, currency, catalog availability, and optional fields can differ. The United States marketplace is the one covered by the recorded 30-item golden run. Check a small sample before planning a large collection in another locale.

Output example

The following values came from one product in the successful full-detail run on 27 September 2026. The example keeps the fields most useful for a product comparison and omits long feature and image lists. The actual row contains the other schema fields too.

{
  "asin": "B098WNQ44M",
  "url": "https://www.amazon.com/dp/B098WNQ44M",
  "marketplace": "com",
  "title": "Kitessensu High Effective Garlic Press With Studs, Heavy Duty Garlic Mincer, Easy to Squeeze and Clean, Rust Proof, Ginger Crusher- Imperial Silver",
  "brand": "KITESSENSU",
  "price": "$8.99",
  "rating": 4.6,
  "reviewCount": 16105,
  "boughtInPastMonth": "4K+",
  "estimatedMonthlySales": 4000,
  "estimatedMonthlyRevenue": "$35,960.00",
  "salesEstimateMethod": "boughtBadgeFloor",
  "bestSellersRank": [
    {"rank": 913, "category": "Kitchen & Dining"},
    {"rank": 4, "category": "Garlic Presses"}
  ],
  "availability": "In Stock",
  "searchKeyword": "garlic press",
  "searchPosition": 1,
  "detailStatus": "full"
}

The site’s 4K+ bucket means at least 4,000 purchases in the past month. The revenue floor in this example is 4,000 × $8.99. It is a product-level gross merchandise calculation, not seller income or profit. Price and badge are snapshots; rerunning later can produce different values.

Use cases

• Seller research: Search a niche, then sort products by purchase bucket, rank, price, and review volume to find items that deserve deeper investigation. Compare several related phrases because search ordering and sponsored placements change what appears first.
• Competitive tracking: Supply an ASIN list on a schedule to watch current prices, availability, visible seller names, and category ranks. Retain each run’s timestamp so changes can be plotted over time.
• Catalog audits: Fetch known product URLs to extract descriptions, feature bullets, dimensions, materials, and related identifiers. The specs object keeps the labels as shown on each product page.
• Retail assortment planning: Compare purchase floors and ratings for candidate categories. Treat a missing purchase badge as missing information; it does not mean zero demand.
• Market reporting: Use the lower-bound revenue calculation to form a conservative visible-signal segment report. Aggregate with care because product variants, duplicated listings, and changing prices can affect totals.
• Price monitoring: Collect a consistent keyword set repeatedly and compare the displayed current price with a displayed list price when both exist. The actor computes discountPercent only from those two visible prices.

How much does it cost to scrape Amazon products?

The configured dataset-item price is $4.44 per 1,000 delivered products, or $0.00444 per item. At that rate, 100 delivered products cost about $0.44, 1,000 cost $4.44, and 5,000 cost $22.20. A 30-product check corresponds to about $0.13 of dataset-item charges. Charges follow successfully delivered rows; an unavailable direct product that creates no row does not add a dataset item.

The price examples describe the actor’s per-result charge. Large full-detail jobs make more page requests than listing-only jobs and may use extra proxy resources if the target starts refusing pages. Platform account terms and any separately metered resources should be checked in your run details. The actor stops pushing items if the platform reports that the run’s charge limit has been reached. Set maxItems to bound output count before running a broad set of keywords.

Tips for best results

Use a specific phrase when you need a coherent product set. A broad query such as kitchen tools can mix unrelated items, while stainless garlic press gives a more focused comparison. Search placement is a discovery signal, not a permanent rank: the same product can move when time, location, availability, or sponsored inventory changes. Record the keyword and collection time together with the price and purchase badge.

Use fullDetails: true for best-seller rank, specifications, review distribution, seller, product images, and variant identifiers. Listing-only mode makes fewer requests and is better for quick scans of title, price, rating, review count, image, and the visible purchase bucket. If a product page is refused during keyword mode, the actor can still retain the listing row and mark detailStatus as listing. This preserves an otherwise useful search result while clearly showing which detail fields were unavailable.

When comparing marketplaces, keep currency in view. estimatedMonthlyRevenue uses the currency and amount displayed for that item, and values from different currencies should not be added without conversion. Product selection also varies by marketplace. Run each market with its matching suffix, and use a small sample to confirm that the needed labels are present in that locale.

For more than one keyword, maxResultsPerKeyword applies separately while maxItems applies to the whole run. The actor avoids duplicate ASINs across keyword pages. A product found under two phrases keeps the first searchKeyword and searchPosition that yielded it. If keyword attribution for every phrase matters, use separate runs per phrase and join the exports afterward.

Limits

Amazon can return a refusal page, a challenge, an error page with a successful status code, or a different product layout. The actor checks for actual search-card or title markers, rotates the datacenter session, and can escalate proxy type after repeated failures. A refused search page can end pagination early. A refused product page in keyword mode leaves a listing-only row; a refused direct lookup is skipped. No collection process can guarantee that every visible item or every optional field will be available on every run.

Purchase badges are buckets. 4K+ gives a minimum of 4,000, not an exact count, and the true value could be materially higher. Some products show no badge at all. The actor does not fill those gaps with an unsupported best-seller-rank formula. Best-seller rank is returned separately when present; it varies by category and should not be interpreted as a direct sales count.

Review counts on search cards can be abbreviated, so listing-only values may be rounded from a K suffix. Full product pages can expose an exact displayed count. Review text is outside this product dataset; reviewDistribution contains percentages when the product page exposes them. Variants are limited to identifiers and visible option labels in the page. Variant-specific price, stock, and seller require fetching each variant as its own ASIN.

Search pages can reorder results or stop exposing further pages. The actor visits at most 20 pages per keyword and caps each keyword at 500 unique products. A large category can therefore require narrower keywords to cover distinct slices. Some product pages omit rank, description, coupon, stock count, shipping party, or seller details. Empty values represent unavailable source data, not guessed values.

Automated workflows

The input and output schemas make recurring product collection straightforward. A workflow can submit a keyword list each morning, download the dataset, and compare purchase buckets or prices with the prior run. A second workflow can submit a fixed ASIN list and flag changes in availability or seller. Always keep the scrapedAt value with exported records so the observations remain tied to their collection time.

Example request: “Collect 30 garlic presses in the United States marketplace with product details, then show the products with a visible monthly purchase floor and their category ranks.” Another useful request is: “Check this ASIN list for price and availability changes and return only records whose displayed values changed since yesterday.” The actor supplies observations; any alert threshold or downstream decision belongs to the consuming workflow.

FAQ

Is scraping these products allowed? This actor reads publicly visible product information. Review the site’s terms and applicable law for your use, respect rate limits and privacy rules, and avoid using collected data for personal-data misuse. The dataset is designed around products, not private customer records.

Do I need to configure a proxy? Usually no. The default starts with a datacenter proxy in the selected marketplace country, rotates the session when a page fails, and can escalate after repeated blocks. A custom proxy can be supplied through proxyConfiguration when needed.

How fast is a run? It depends on the input and the number of product pages. Listing-only keyword mode needs far fewer requests. Full-detail mode requests a product page per result, and retries or challenges can make a run longer. The recorded full-detail check collected 30 products; use a small run to measure your current marketplace conditions.

Can I schedule or monitor repeated collection? Yes. Save the same input for recurring runs and compare dataset rows using asin, marketplace, and scrapedAt. Watch the run status and output count; a sudden drop can indicate a source layout change or increased refusals.

Can I export the results to a spreadsheet? Yes. Download the dataset in a spreadsheet-friendly format or use the dataset endpoint in your own processing pipeline. Nested fields such as ranks, specs, and variants are easiest to preserve in JSON; flatten them after export if your spreadsheet needs one value per column.

Why is a field empty? The site may not display it for that product, the field may be available only on a detail page, or the detail page may have been refused. Check detailStatus first. A listing row normally has empty specifications and rank. A full row can still have empty coupon, description, stock quantity, or sales fields when the site does not show them.

Is the monthly revenue figure actual seller revenue? No. It is a conservative multiplication of the minimum purchase bucket and the displayed item price. It does not account for fees, returns, taxes, shipping, variation mix, or whether every purchase used the displayed price. Use it as a visible activity floor.

Integrations

The default dataset can be consumed through an endpoint, exported to a file, or passed to scheduled workflows. Use callbacks to react when a run finishes, then combine the results with your own product catalog, reporting database, alerting process, or spreadsheet. The stable ASIN, marketplace, source keyword, and timestamp fields support repeat collection and comparisons.

Support

Open an issue on the Issues tab; we reply within 24h and add fields on request.