Price monitoring software, field by field

This is the collection engine. On every check it records the shelf price, the crossed-out was-price, stock status, shipping cost, promotional text, the seller behind the listing and the currency, then stores that observation with a timestamp and the exact source URL. Observations are appended, never overwritten, so a number you saw in March is still there in September, traceable to the page it came from.

What price monitoring software captures on every check

One number is not enough to price against. A listing at $179.00 means something very different when it is in stock, ships free and carries no promotion than when it is a backordered marketplace offer with $14.00 shipping and a 10 percent coupon banner. Every refresh stores the whole listing.

How collection works, from URL to stored row

1. 1

   Resolve the source

   Competitor URLs are attached by hand, imported in bulk by CSV, or discovered by identifier from your product feed. Each domain carries its own fetch profile and extraction recipe.

2. 2

   Fetch the page

   Checks are spread across the refresh window rather than fired in one burst, and pages that build their price in the browser are rendered before reading, so a client-side price is not stored as null.

3. 3

   Read structured data first

   Where a retailer publishes schema.org Product markup, price, currency and availability come from it. That is the same feed the site hands to search engines, so it survives most redesigns. A per-domain extraction recipe is the fallback, regenerated automatically when a layout changes.

4. 4

   Validate before storing

   Every candidate price passes plausibility checks. A $0.00, a shipping fee read as a price, or a fortyfold overnight jump is quarantined rather than published, and the source is flagged in data quality.

5. 5

   Attach it to a product

   Matching runs by GTIN, UPC, EAN or MPN first, then title and attribute embeddings, then image. Every match carries a confidence score, and low-confidence matches never reach a repricing rule.

6. 6

   Append the observation

   The row is written with its timestamp, source URL and extraction method. Nothing is overwritten, which is what makes history real rather than a rolling snapshot.

How bot mitigation and moved URLs are handled is covered in price scraping. The matching waterfall and its harder cases are in the product matching guide.

Refresh cadence by plan

Refresh rate and product count are the only two things that change between plans. Every plan includes matching, MAP monitoring with screenshot evidence, repricing, analytics, all integrations, the API and unlimited users.

Pick the cadence from how fast your category moves, not from ambition. Here is the arithmetic that settles it: a competitor runs a six-hour flash sale, and a once-daily check sees it only if the check lands inside those six hours, which is 6 in 24, a 25 percent chance. A four-times-daily check catches it nearly every time. In categories with frequent short promotions, daily collection is sampling rather than monitoring. In furniture or industrial supply, daily is plenty.

How variants are handled

Each purchasable variant is monitored as its own row. Size, color, capacity and pack count each get their own price, stock and history, because averaging them hides the one that is losing.

• Variant to variant. A 5Ah battery pack is compared against a 5Ah battery pack, not against the 2Ah version on the same product page.
• Pack count is normalized for comparison. A 12-pack at $47.88 is $3.99 a unit. A single at $4.49 is 12.5 percent more per unit, because (4.49 - 3.99) / 3.99 = 0.125. Both the pack price and the unit price are stored, so you can compare on either.
• Bundles are not folded in. A drill kit with two batteries and a bare tool are different products. Matching flags the difference instead of treating a $129.00 bare tool as a 31.7 percent undercut on your $189.00 kit.
• Parent-level rollups are derived, not collected. A product family shows the range and the worst-positioned variant, both computed from the variant rows rather than scraped from a category page.

How history is stored, and what it answers

History is an append-only series of observations. Each one holds the price, was-price, stock, shipping, promotion text, seller, currency, source URL, extraction method and the time of the read. Change events are derived from that series rather than stored separately, which means a correction to a bad reading does not rewrite the past.

• When a competitor's price drop started, and whether it has ended.
• Whether a low price is an event or a cadence. A price that falls every second Thursday is somebody's promotional calendar, and chasing it costs margin twice a month.
• How long a listing sat below MAP, which is the number that matters in MAP protection.
• How many times a seller changed price in a quarter, which tells you whether they are repricing automatically.
• What a page said on a given date, since the source URL and timestamp travel with the observation.

Raw observations export as CSV over SFTP or through the REST API, not just as charts. If you want the series in your own warehouse, take it.

How alerts are configured

An alert is four choices: a trigger, a scope, a threshold and a destination. Set them on a boundary you would actually act on, or the channel becomes noise inside two weeks and everyone mutes it.

1. 1Trigger. An in-stock competitor crossing your price in either direction; a move larger than a set percent or dollar amount; every in-stock competitor now sitting below you, which is a different signal from one seller going low; a listing below MAP; a stock flip; or a source that has gone quiet for two refresh cycles.
2. 2Scope. All products, a category, a brand, a saved product group, or a single SKU. Thresholds usually need to differ by category: 3 percent is a sensible starting point for general merchandise, 1 percent for high-velocity electronics.
3. 3Threshold. Percent, dollars, or a crossing of your own price. Alerts are deduplicated within a refresh cycle, so a price oscillating around your number does not fire six times before lunch.
4. 4Destination. Email, a Slack channel through the Slack integration, a webhook, or the weekly digest for people who want the summary rather than the feed.

Every alert carries enough to decide without opening anything: your product and price, the competitor and their new price, the move in dollars and percent, their stock status, and a link to the source page with its capture time.