What we read

Only what a store publishes to every visitor: its public catalog, its public store details, and its homepage. Before reading anything we fetch the store’s robots.txt and follow it; a store that disallows our reader, or blocks automated visits, is not tracked. We identify ourselves as DTCSignalsBot, pause between requests, and honor a store’s requested crawl delay.

Four tracking depths

Shopify stores publish their catalog in a standard public format. We read up to 2,000 products and their published variants: listed price, the merchant’s compare-at price, availability, product type and tags; plus the store’s base currency, its published product count, its sale collections, and the homepage. Detailed comparisons retain up to 250 variants per product and 20,000 per store; only recorded matching variants support exact price changes.

WooCommerce stores publish the same kind of catalog through their public store interface: prices, sale prices and stock, up to 2,000 products.

Other stores are read from the product data they publish for search engines (schema.org structured data). We find product pages in the site’s sitemap and read the same sample of up to 250 pages each day, so a difference between two days is a change on the page and not a change in the sample. On these sites we see price and stock; compare-at discounts and removed products are not visible.

Homepage only. Where a site publishes no readable product data, we follow the sale message on its homepage and nothing else.

You are told which depth applies before a store uses one of your slots.

How often

Once a day per store, in batches through the day (UTC). We read one version of each store; prices shown to visitors in other countries may differ. A failed visit is retried, and an independent second path covers a store the first could not reach. A day on which neither succeeded is recorded as a gap, never as “no change”. During Cyber Week, tracked stores are read twice a day.

Market pulse: the offer over time

The Market pulse plots daily observations for public-index brands. “On sale” is the share of observed products priced at least 5% below their merchant-supplied compare-at price. “Listed available” is the share of observed products with at least one variant marked available; the rest can include unknown availability. Each date shows its sample size and coverage. Missing readings stay blank, and a changing product mix can change the share. This is not a count of units, orders, visitors or revenue. Traffic and orders are not connected.

What counts as a change

Each day’s reading is compared with the store’s previous successful reading.

Price up or down: the same recorded variant changed price, or a historical single-variant Shopify product did. When variant identity is unavailable, we label a changed product price range; a different variant mix can explain it. Older price observations without variant evidence are labeled as product-level readings. Discount started, deepened or ended: a product counts as discounted when its listed price is at least 5% below the merchant’s own compare-at price; “deepened” means at least five points more. New product: a product that was not in the previous reading. Removed: reported only after a product is missing from two complete readings in a row, so one short reading is not mistaken for a removal. Sold out and back in stock: by the store’s own availability flags, a product went from at least one variant available to none, or back. Sale collection: a collection named for a sale, clearance or Black Friday appeared. Homepage message: the promotion text in the announcement bar, header, title or first headings changed. We keep only text that carries a discount: a percentage off, an amount off, a code, or a named sale.

Store-level sales

“Sale started” compares the share of a store’s products that are discounted today with that store’s own usual level over the previous two weeks. A brand that normally has 40% of its catalog discounted is not “on sale” at 42%. The share must rise clearly above the usual level to start, and fall back close to it to end, so a store hovering near the line does not flip back and forth.

Guards against false alarms

When a catalog is larger than we read, or a reading is incomplete, we do not report removals and only count a product as new if its own publication date is recent. When almost every price moves by the same ratio and discounts are unchanged, we record one market or currency change instead of hundreds of price changes. A very busy day is capped at 200 listed changes per store; the rest are counted by type.

What we cannot see

Sales, orders, revenue, margins, inventory quantities and demand. Discount codes that apply only at checkout. Banners that an app adds after the page loads. Prices shown to visitors in other countries. Compare-at prices are chosen by the merchant and are not verified selling prices. A newly listed product is not necessarily a launch, and a product we no longer see is not necessarily discontinued.

Coverage and gaps

Some stores refuse automated reading, redirect to another address, or stop answering. A store with no successful reading for seven days is marked as not reachable and checked weekly; anyone tracking it is told. Of the 72 brands in the public index, 66 were readable when daily collection began; the other six are shown as gaps.

The public index and its baseline

The index pages and market map use a supplied scan of September 30, 2026: 107 target domains, 72 confirmed Shopify brands, 66 visible catalogs and 18,278 sampled product records, capped at 500 per store. In that baseline “recent” means a creation timestamp within 45 days and “deep discount” means at least 40% below compare-at. Kith, Ring Concierge and Burga met a possible bulk-change rule and are excluded from its activity comparisons. The baseline, and two test runs on October 1, used earlier methods and are never joined to the daily record.

Search research

The search pages are saved research: Google Ads search-volume estimates for 14 phrases in the US and UK, covering September 2025 through August 2026, retrieved September 30, 2026. They do not refresh. See sources and interpretation.

Commentary and DTC AI

Our commentary separates what was observed from what we think it may mean. DTC AI is part of the paid plans and answers on your own competitor pages, from the readings of the stores you track; its answers can still be wrong. Daily alerts and reports are computed from the data, not written by AI, except for one short interpretation paragraph in a report, which is labeled.

Corrections

Send the store, the date and what looks wrong to support@dtcsignals.com. If you operate a store and do not want it read, disallow DTCSignalsBot in your robots.txt and we will stop on the next visit, or write to us.