Data sources and data quality: where each price comes from

01Every price names its source

A price row in Pricebeeper holds the price, currency, seller, URL and time, the stock, condition, shipping and total where the source reports them, and the source it came from. These are the source labels:

The Data Quality page counts observations per source, MAP evidence names the source of the offer it recorded, and the API and the price-history CSV export carry the source and kind of every row. Marketplaces and channels describes each channel in more detail.

02Offers, ads and listings are kept apart

A search result isn't always one seller's price for your product. Pricebeeper gives every row one of three kinds, and each kind counts in different places:

• Offer: one seller's offer, from a search, a store page or a marketplace account. Only offers raise price and stock alerts, set the lowest market price, count as a seller's price and feed repricing.
• Placement: an ad or merchandising slot, such as a Google Shopping ad, an Amazon Sponsored Product or Amazon's own-brand section. It is often another product than yours, so placements never raise price alerts or enter market averages. They feed competitor ad monitoring.
• Listing: a search result that names no seller, recorded under the platform's name. Every Amazon keyword result is one. Listings count in market statistics, which describe what shoppers see, but are never a seller's price.

03What the exact flag means

An offer marked Exact match comes from the seller list of a product Pricebeeper has matched to one Google Shopping product ID or Amazon ASIN, or from a re-check of one of those sellers' pages. The seller lists that exact item.

Offers from a keyword search carry no flag, because they may be your product or a close relative in another size or bundle. Pinned pages carry none either: you chose the page, and it isn't checked against a product ID.

MAP monitoring raises violations only for exact offers or sellers on your list, and repricing rules can be limited to exact matches. Exact product matching explains how matches are made and reviewed.

04Store reads are seller history, not market statistics

Re-checks and pinned pages read one store's page from our servers. That makes them good seller history and poor market statistics: 5 stores read 4 times a day would outweigh a monthly search of 20 results. So they are left out of market averages, the category index, price forecasts and regional arbitrage in pricing analytics, and used for seller lines, alerts and stock.

• A store read can lower a product's lowest market price, never raise it, and never with a sold-out offer. The next search sets it again.
• A sold-out reading from a store page raises a stock alert only when the next reading agrees. Unknown stock never alerts.
• The first read of each offer is compared with the region's search price. When the store shows our servers a different price than shoppers in that region see, the offer stays on the regional search data.

05The data-quality dashboard

The Data Quality page, included with Growth, Pro, Business and Custom, reviews the last 30 days of your account:

• Source mix: observations per source and their share, with how many were offers, placements and listings.
• Stale products: monitored slots whose last check is older than your plan's cadence plus a day. Paused slots aren't counted.
• Dead products: active products with no price at all in 60 days, for example after a URL broke or the product stopped being listed.
• Source disagreement: a product, region, currency and day where 2 or more sources recorded offers that spread by more than 5% of their average (as a standard deviation).
• Thin coverage: product and region pairs with fewer than about one observation a week.
• Stock coverage: the share of offers with unknown stock, per source. Keyword search results never state stock, so stock alerts rely on seller data and store pages.

The page shows the source mix and the lists of stale and dead products. The full overview, with disagreement, coverage and stock figures, is also available through the API. The dashboard reports problems; it never hides or changes a price.

06How Pricebeeper collects: robots.txt and marketplace terms

Google Shopping and Amazon results are bought from DataForSEO, so Pricebeeper doesn't scrape either site itself. One search result for a product and place serves every account that monitors it, which keeps the number of searches down. Sources you connect with your own credentials are never shared with other accounts.

Store pages are read by our own crawler, PricebeeperBot:

• robots.txt is checked for every page and every redirect. If robots.txt can't be loaded, the store is treated as closed to us.
• Each store gets one request at a time, at least 5 seconds apart, across all accounts.
• A store that refuses access or answers with a bot challenge is marked Store blocks checks and left on the monthly search data.
• Amazon pages are never fetched.

Marketplace APIs are used only where our reading of their terms allows it. Each source records the terms page and the date it was read, and the Amazon Product Advertising API, Etsy and Best Buy are switched off because their terms don't allow competitor price monitoring. Barcodes read from a store through the Shopify Admin API are never shared with other accounts, as Shopify's API terms require.

07Labels next to each offer

In a product's offers table, each offer shows its seller, the shop domain when known, the price with shipping, stock, condition and when it was read. Badges say how it is kept current: Checked daily, Checked every few hours, Monthly (regional price), Monthly (store blocks checks), or Store's online price for a pinned page. Search results that name no seller sit in their own panel below.

Shop domains for sellers known only by name are learned from Google Shopping data. A domain is used once the name and the domain were seen together on 2 or more days and in at least 80% of that name's observations; until then the offer shows the name alone.

08Limits of the data

• Searches run monthly on standard plans, and known sellers are re-read weekly on Starter and Plus and daily on Growth and up. A price that changes and changes back between two reads can be missed, and offers that aren't re-checked, such as Amazon's, change only at the monthly search.
• Keyword search results don't state stock, so stock is unknown for many offers of unmatched products.
• Some Google results don't link the seller's page, and those offers can't be re-read directly.
• Individual price rows, with their source, URL and time, are kept for 90 days. Older history is kept as daily summaries per seller and kind (lowest, average and highest price, and stock), without the source.
• History starts with your first check. Pricebeeper has no prices from before you signed up.
• The dashboard flags problems. Fixing them, with a new URL, a confirmed match or a pinned page, is up to you.

Questions people ask