
Search pages
Ranked results
The fetch layer behind RoundupForge
Live since 1 October 2026
Fetchwright reads public product and search pages and hands back what a product roundup needs — titles, details, prices, ratings, the latest reviews. A page only counts when real data comes out of it.
What goes in

Ranked results

What a shopper sees

Reused, kept fresh
Read · check · deliver
Where it goes
Every product slot filled.
Kept current, never stale.
Ratings and the latest review dates.
What moved since last time.
Each page at a steady, polite pace — and only until it has what it needs.
A page counts only when real data comes out. Wrong place or empty? Fetched again.
Titles, details, prices, ratings and review dates go straight into the article.
Since the 5 October update Fetchwright delivers almost every page by itself. For StrongMocha, scraping is projected to cost 88 % less than at the start — each month, and over a full year if usage stays as it is.
Production record since the 5 October update: more than 60,000 pages in the first 38 hours, no hour below 97 %. Savings are projected from current usage, in percent against StrongMocha’s starting point: two rented plans a month, before pages were reused.
Fetchwright runs through the same frozen test as the rented service it replaces. Same pages, same rules, no curve.
| Metric | Gate | Fetchwright | Rented service |
|---|---|---|---|
| Search pages usable | ≥ 85 % | 97.1–99.3 % | 37.5 % |
| Product pages usable | ≥ 92 % | 92.0 % | 88.3 % |
| 95 % of pages delivered within | ≤ 30 s | under 6 s | — |
| Sustained capacity, 4 hours | ≥ 250k / day | partly | 130–250k |
| Cost per usable page | below incumbent | a fraction | baseline |
Fetchwright: frozen 300-page test in three different setups, 30 Sep – 1 Oct 2026; every setup passed. Capacity: holds RoundupForge’s real load in production; a four-hour test at eleven times that load showed recurring slowdowns. Rented service: its own production record over 30 days.
A product roundup is only as good as the pages behind it. Every miss is a product that never makes it into an article — Fetchwright exists so that doesn’t happen.
Every page arrives the way a shopper in the right country sees it. A page for the wrong delivery location never reaches an article.
An answer from the server is not a result. A page counts only when real data comes out of it — otherwise it is fetched again.
95 % of pages arrive in under six seconds. The slow ones no longer hold up an article.
Pages already read are kept and reused, with prices refreshed from the search result. More than half of all product fetches disappeared.
Public pages only, read at a steady pace. No logins, no personal data, nothing a visitor could not see.
Fetchwright takes over only where it costs less per usable page than the service it replaces. Quality alone does not get it the job.

Every attempt counted · no quotas · no guesswork
Results land here as they happen — pass or fail.
The first day and a half. More than 60,000 pages since the update: 99.5 % delivered by Fetchwright itself, about three seconds each, and not a single hour below 97 %. The rented backup took the other 0.5 %.
Scraping costs down 88 %. StrongMocha started with two rented plans a month. Reusing pages halved that; on current usage Fetchwright cuts the rest by about three quarters — 88 % less each month, and over a full year if usage stays as it is.
Nearly every page by itself. Since the 5 October update Fetchwright adapts by itself when the shop starts pushing back. In the first hours it delivered 99.7 % of pages itself, in about three seconds each; 0.3 % still went to the rented backup.
Three setups, three passes. The same 300-page test in three different setups: 97–99 % of search pages and 92 % of product pages usable on every one.
The right place, always. Some addresses that every map puts in the US are placed abroad by the shop itself. Fetchwright now spots that on the first page and stops using them.
Fetchwright is live. Every page RoundupForge reads now goes through Fetchwright first; the rented service only steps in as a backup. First hours: 99.7 % of pages delivered, the backup was needed six times.
As complete as the incumbent — or more. Each page is checked against the rented service in the background: titles identical, prices and review counts missing less often than there.
Cost rule met. Fetchwright runs at a fraction of the rented service’s cost per usable page.
Both quality gates passed in a paid setup. On 300 real pages, 99.3 % of search pages and 92.0 % of product pages were usable — confirmed on a full re-run.
Right place, faster, lighter. Every page arrives for the right delivery location, the slowest pages arrive more than twice as fast, and each page needs less data — with nothing lost.
Pages are reused. Product fetches down 55 %; prices stay fresh from the search result.
First live pages. 42 pages, zero blocks, all on the first attempt; on 30 identical pages it matched the rented service with more complete data.
The rented service retires. It stays on as a backup until its remaining credit runs out around mid-October and is not renewed; then Fetchwright runs entirely on its own.

Fetchwright fills RoundupForge’s articles; it is not offered as a service. Write to us — a person reads every message.