Home / Cost to scrape / Buy a dataset or scrape it

Buy the dataset or scrape it?

100,000 Amazon records cost up to $250, once, from Bright Data's dataset marketplace; scraping 100,000 Amazon pages costs from $29 for a month on Scrape.do. Scraping is cheaper for 17 of the 19 sites Bright Data sells datasets for; the dataset is cheaper for Facebook and Instagram, where few APIs publish a rate.

Bright Data's dataset prices read Oct 11, 2026; each scraping bill is worked out from its vendor's own pricing page, dated on that vendor's page here.

100,000 records bought, or 100,000 pages scraped

Bright Data's dataset price is its published ceiling of $0.0025 a record with a $250 minimum order; the scraping bill is the cheapest complete monthly bill for that many of the site's pages. 19 sites, cheapest to scrape first.
Site	Dataset, 100,000 records, once	Scrape 100,000 pages in a month	Dataset, 1,000,000 records	Scrape 1,000,000 pages in a month
GitHub	up to $250	$19 BrowserCloud, Hobby	at most $2,500	$90 ScrapingDog, Standard
Amazon	up to $250	$29 Scrape.do, Hobby	at most $2,500	$90 ScrapingDog, product page, Standard
Indeed	up to $250	$40 ScrapingDog, Lite	at most $2,500	$90 ScrapingDog, Standard
Zillow	up to $250	$40 ScrapingDog, Lite	at most $2,500	$200 ScrapingDog, Pro
eBay	up to $250	$90 ScrapingDog, Standard	at most $2,500	$350 ScrapingDog, Premium
Google Maps	up to $250	$90 ScrapingDog, Standard	at most $2,500	$350 ScrapingDog, Premium
LinkedIn	up to $250	$90 ScrapingDog, post, Standard	at most $2,500	$350 ScrapingDog, post, Premium
TikTok	up to $250	$90 ScrapingDog, Standard	at most $2,500	$350 ScrapingDog, Premium
Walmart	up to $250	$90 ScrapingDog, Standard	at most $2,500	$350 ScrapingDog, Premium
X, formerly Twitter	up to $250	$90 ScrapingDog, Standard	at most $2,500	$350 ScrapingDog, Premium
Yelp	up to $250	$90 ScrapingDog, Standard	at most $2,500	$350 ScrapingDog, Premium
YouTube	up to $250	$90 ScrapingDog, search, Standard	at most $2,500	$350 ScrapingDog, search, Premium
Airbnb	up to $250	$99 Scrape.do, Pro	at most $2,500	$699 ScrapeBadger, Scale
Booking.com	up to $250	$99 Scrape.do, Pro	at most $2,500	$699 ScrapeBadger, Scale
Crunchbase	up to $250	$99 Scrape.do, Pro	at most $2,500	$699 ScrapeBadger, Scale
Etsy	up to $250	$99 Decodo Web Scraping API, $99 plan	at most $2,500	$415 HasData, Growth 5M
Glassdoor	up to $250	$99 Scrape.do, Pro	at most $2,500	$699 ScrapeBadger, Scale
Facebook	up to $250	$725 SerpApi, Searcher	at most $2,500	$3,750 SerpApi, Cloud 1M
Instagram	up to $250	$725 SerpApi, Searcher	at most $2,500	$3,750 SerpApi, Cloud 1M

A dataset record is one product, listing, place, company, profile or post with its fields already pulled out; a scraped page is the raw page, or the parsed result where the API offers one for that site. A mode named after the vendor (a "light request", a "post") is the cheapest endpoint that vendor prices for the site, as the site's own page lists them. At 1,000,000, scraping costs less than the dataset's ceiling for 17 of the 19 sites; Bright Data's volume discounts, which its page does not price in writing, would narrow that gap.

When buying the dataset is the better deal

You need the data once. A one-off pull of 100,000 records costs up to $250 and needs no code. Scraping the same pages means a month of a plan, the code that finds every page, the code that pulls the fields out, and re-runs for the pages that fail.
The site is one few APIs price. For Facebook and Instagram, the cheapest published scraping bill for 100,000 pages is above the dataset's $250.
You want history. A dataset can hold records collected before you started; scraping only sees the pages as they are today.

When scraping is the better deal

You need it every month. Bright Data says monthly refresh plans take up to 80% off, which puts 100,000 records at $50 to $250 a month, depending on the discount you are given. Scraping 100,000 Amazon pages costs from $29 a month.
You need specific pages or fields. A scrape fetches exactly the pages you choose, as they are on the day you fetch them. Bright Data's datasets can be filtered before purchase, and you pay only for the records you keep; a field the dataset does not collect is not there to filter.
You need it fresh. Bright Data's ready-made data dates "back from a few days to a couple of months", in its own words; freshly collected data is a separate option at checkout.
The site is not in the catalogue. Bright Data lists more than 1,000 datasets; anything else has to be scraped.

What neither price includes

Finding the pages. To scrape 100,000 product pages you first need their addresses, which usually means scraping search or category pages too. Those are billed like any other page; the crawl budget adds them up.
Failures. Some APIs bill a blocked request and some do not; the failed requests tool shows what a lower success rate does to each bill.
Pulling out the fields. A dataset arrives ready to load, in JSON, NDJSON, CSV, XLSX or Parquet. A scraped page arrives as HTML unless the API has a parser for that site, which is often a higher rate; see each site's page for the endpoints.
Your own time. Writing and keeping a scraper running is the cost neither vendor bills, and the one most worth counting.

Where these figures come from

Bright Data's datasets page publishes "up to $0.0025 per record" and a minimum order of $250, and it says what a record is: "one record equals one entity: one person profile, one product listing, one job posting, one company, one social media post, and so on". Its volume tiers (100K, 500K, 1M, 5M and 20M records) are drawn on its page without prices in the text, so this page uses the published ceiling at every volume and never a guessed discount. Read on Oct 11, 2026 from brightdata.com/pricing/datasets and each dataset's own page, linked from its price above.

The scraping bills are the same figures each site's own page on this site opens on: the cheapest complete monthly bill for that many pages, from each vendor's published rate for the site where one exists, and otherwise from the general price for a page that hard to reach. A bill sold only by quote, or that needs a rate the vendor does not publish, is left out rather than guessed.