01How Extralt works
Follow 1 shoe through Extralt.
Each rectangle is an online store. Follow 1 shoe from 3 of them: Extralt collects its product pages, turns them into clean records, matches the same shoe across stores and keeps every change, ready to explore.
The ecommerce web, drawn as a field of small browser windows. Each one is an online store and the blocks inside it are its product pages. 3 stores side by side are outlined.
02Extract
You pick the stores. Extralt builds the crawlers.
Extralt generates a crawler for each website and validates it on real pages before the first run. Then it finds the product pages whenever you collect. No scraper to write or fix.
The view zooms into the 3 stores you picked, north.example, field.example and tempo.example, each a storefront built differently with a grid of products. Above each, a crawler is generated, validated and running.
03Extract
3 stores sell the same shoe.
The Nike Air Max 90, black, EU 42. Each store writes the title, price, size and stock its own way. Most scraping stops here, and your cleanup starts.
3 product pages for the same shoe. north.example: “Nike Air Max 90 — Black, size 42”, €139.99, Size 42, In stock. field.example: “Air Max 90 by Nike / Black / EU42”, 139,99 EUR, EU42, available. tempo.example: “NIKE Air Max 90 Black • 42”, 134,99 €, Taille 42, En stock.
04Extract · 1 credit per page
Each page becomes 1 record.
A Capture keeps what the page said, as written, with its URL and the time it was seen. It is the first of the 2 things you pay for: 1 credit.
The title, price, size and stock leave each page unchanged and become rows of a Capture, with the page path and 1 Sep 2026, 09:00 UTC. Each Capture is marked 1 credit. The counter reads 3 credits.
05Enrich · 1 credit per page
Then every record uses one format.
Enrich turns each Capture into an Item: brand, model, colour and size as fields, one price format, stock in English and an industry-standard category. It is the second and last thing you pay for.
Each title splits into brand Nike, model Air Max 90 and colour Black. 139,99 EUR and 134,99 € become €139.99 and €134.99; Size 42, EU42 and Taille 42 become EU 42; available and En stock become In stock; the category Sneakers is added. Each Capture is kept behind its Item. Each Item adds 1 credit. The counter reads 6 credits.
06Enrich · included
3 pages. 1 exact variant.
Brand, model, colour and size agree, so the 3 Items become 1 matched variant with 3 offers side by side. A size 43 from the same store stays a separate variant. Every match can be inspected.
The matching fields merge into 1 variant, Nike Air Max 90, Black, EU 42, Sneakers, with 3 offers: north.example €139.99, field.example €139.99 and tempo.example €134.99, the lowest. A size EU 43 Item from north.example is kept apart. The counter stays at 6 credits.
07Enrich · included
The data model, built for you.
The variant links to its product, its other sizes and colours, a listing in each store and every offer, with reviews and changes alongside. Matching and the model use no credits.
A connected model: the product Nike Air Max 90, Sneakers, with variants White EU 42, Black EU 42 and Black EU 43; a listing and an offer in each of the 3 stores; reviews, stores and changes linked. The counter still reads 6 credits.
08Extract and Enrich · at scale
Every page your crawlers find.
We followed 1 shoe. The same path runs for each product page in the stores you choose: 1 Capture, 1 Item, then 1 catalog where the same variants line up across your stores.
Zoomed out: your stores sit lit in a row among stores you did not choose, which stay dark and send nothing. Each page passes a Captures line and an Items line, 1 credit each, into an example catalog of variants by store, collected 1 Sep 2026: Nike Air Max 90 Black EU 42 at €139.99, €139.99 and €134.99, with 2 more Air Max 90 variants kept separate; adidas Samba OG White EU 42 at €119.99, €109.99 and €119.99; New Balance 530 White EU 42 at €99.99, €119.99 and €104.99; ASICS GEL-1130 Silver EU 42 at north.example only, €109.99; Converse Chuck 70 Black EU 42 at field.example only, €89.99. Unnamed rows fill the rest of the catalog.
09Second collection · 7 Sep
Collect again, a week later.
You choose when to collect. On 7 Sep the same pages take the same path again, at the same 2 credits each, and the catalog fills with what they say now. What they said on 1 Sep is kept.
A second collection on 7 Sep 2026. The catalog as collected on 1 Sep steps back behind its tab and a 7 Sep tab comes in; the catalog fills again, row by row, as the same pages pass through the same Captures and Items. The counter reads 12 credits: 3 pages, collected twice.
10History · included
Every change is kept.
Against 1 Sep, 3 values changed in the example rows. Each is kept with its before, its after and when it was seen, for every page you collect, at no extra cost.
Compared with the 1 Sep catalog, 3 cells of the example rows changed: Nike Air Max 90 at north.example from €139.99 to €129.99; New Balance 530 at field.example from €119.99 to €109.99; adidas Samba OG at tempo.example from in stock to sold out. Each change is kept on a card with its before, its after and its date, 7 Sep. Other cells flash where they changed, without values. The counter still reads 12 credits.
11Explore · included
Answers that show their work.
4 Analyses answer questions from both collections of the example catalog. Each shows the values it was built from, and links back to their pages.
4 Analyses answer questions from the example catalog's 5 variants at 3 stores, each with the values it was built from. Price position: −5.5%, the Nike Air Max 90 at north.example against the median of field.example and tempo.example, from 3 offers. Price movements: 2 down, the Air Max 90 at north.example and the New Balance 530 at field.example, −€10.00 each, among 11 offers compared. Availability changes: 1 sold out, the adidas Samba OG at tempo.example. Assortment overlap: 3 exact variants shared by north.example and field.example, 1 unique to each, from 8 listings.
12Explore · included
Use it wherever you work.
In the dashboard, in plain language, through the Agent, or in your own code with the API and exports.
Your data reaches the dashboard, the Agent, the API and the MCP server. Plain-language queries read Captures, Items and the catalog, and Captures and Items export as JSONL or Parquet. The −5.5% price position appears in the dashboard, the Agent, the API and the MCP server.
13What it costs
2 credits per page. The rest is included.
Each collection of a page makes 1 Capture and 1 Item. Crawlers, retries, matching, history, Analyses and access come with them. The trial's 5,000 credits cover 2,500 page collections.
The cost: 2 credits per product page each time it is collected. Our example, 3 pages collected twice, used 12 credits. On the path, crawlers are included, Capture and Item cost 1 credit each, and the variant, history, Analyses and access are included. The free trial's 5,000 credits cover 2,500 page collections; Start at $29 a month covers 5,000 a month; Scale from $100 a month covers 50,000 a month, or 12,500 pages collected 4 times a month.
Start free trial7-day free trial · 5,000 credits · Plans from $29/month