Sources
Everything the report runs on, what it costs a month, and the list of figures it will not produce. This page exists because the other nine make claims — 342 listings walked from 05:30 this morning, a band around a dozen of our products, a message on somebody's phone before seven — and each of those is worth exactly as much as the plumbing underneath it. One of the pipes did not answer today, which is the honest reason to read this page first rather than last.
What this report will and will not say
Seven terms the rest of the module is built on. Each one gives something up, which is the only way to tell they are meant.
Nothing it produces leaves the building
This is the only department on the platform with no outward surface at all. There is no approval queue on any page, because a queue is for things that go out, and nothing here does: no reply, no post, no listing, and above all no price. A threshold alert reaches your own pricing owner on your own Slack, which is a message between two people who work here. What it gives up is the obvious feature — the one where the tool sees a competitor undercut you at midnight and fixes it for you. That feature is somebody else's outage at two in the morning.
There is no official way to see a competitor's price
Not one marketplace publishes a seller API for anybody else's listings. Every competitor figure in this report came off a managed crawler walking a public page from behind Cloudflare and Akamai, which is why this module's feasibility note is amber rather than green. What it gives up is certainty of supply: a shelf can refuse, and one did this morning.
A shelf that did not answer is missing, not zero
N11 returned a challenge page to every request between 05:30 and 05:49, through 3 attempts and a second provider. The round did not stop and the gap was not filled in: those listings are yesterday's everywhere they appear and they carry the mark. What it gives up is a clean-looking morning. A silent hole is the one failure nobody catches.
A barcode is certainty; a likeness is not
Two grinders with the same barcode are one object in two shops, and a price beside ours means exactly what it says. Two bags of Ethiopia Yirgacheffe from two roasteries are the same kind of coffee and not the same coffee — different farm, different roast date, different hand on the drum. So how a match was established decides what it may do: an exact match may raise an alert, a likeness may sit in a band and inform a person, and anything weaker may only become a question. That is not a setting. No threshold on the Watchlist can be configured to let a likeness wake somebody at eleven at night.
Under 3 offers there is no band
A median of two offers is a sentence with a decimal point in it and nothing behind it. Where the watched shelves do not carry 3 comparable listings, the Price band tab prints the reason instead of a number — and it never averages an exact match together with a likeness, because the result would be true of neither. A bundle we assemble ourselves has no market price at all: the parts can be priced, the box cannot. What it gives up is coverage. Several products of ours have no band and are not going to get one.
Stock is a flag, and a discount is read rather than believed
No marketplace publishes how many units a seller is holding, so this module says "out of stock" and never "three left". It is just as careful in the other direction: a shelf claiming thirty per cent off is checked against its own history, and a badge that has been up for more than 21 days while the till never moved is reported as what it is — the price, with a label above it that keeps rising. What it gives up is the depth of anybody's warehouse, which cannot be had from a listing page at any price.
No reviewer identity, and a theme is never one review
No reviewer name, profile or picture is stored. The theme, the count and one quoted line, nothing else. A theme needs at least two reviews behind it. Anything found once is shown as one review and never called a pattern. A marketplace shows the newest and the most extreme reviews first, so this is a biased sample and the page says so. It is good enough to find what people complain about and useless for working out how often it happens. What it gives up is the number everybody wants — how often a problem actually happens. The sample can tell you what people complain about. It cannot tell you how many of them there were.
The first and the fourth are the expensive ones. Between them they rule out the feature everybody asks for first — the tool that sees a competitor undercut you at midnight and quietly fixes your price — and they rule it out on purpose, not for want of an endpoint.
The plumbing
9 connected, 3 short of itWhat it runs on
Four public shelves nobody gives us an API for, one set of official APIs for our own side, and the places the answer goes next.
| Connection | Has data through | What it does | A month |
|---|---|---|---|
| Trendyol · competitor shelfread this morningcrawler | this morning, 05:41 | 139 watched listings: price, list price, stock flag, rating, review count and the review text under it.Cloudflare, cleared with a residential exit. Finished 05:41. | $2.92 |
| Hepsiburada · competitor shelfread this morningcrawler | this morning, 05:38 | 103 watched listings: price, list price, stock flag, rating, review count and the review text under it.Akamai, cleared with a residential exit. Finished 05:38. | $2.16 |
| Amazon · competitor shelfpartly readcrawler | this morning, 05:36 | 57 watched listings: price, list price, stock flag, rating, review count and the review text under it.Rate limited; two listings came back on the second pass and two did not. 2 listings did not come back, and they are marked rather than carried over. | $1.54 |
| N11 · competitor shelfdid not answercrawler | 13 Sep — yesterday | 43 watched listings: price, list price, stock flag, rating, review count and the review text under it.Challenge page on every request, 05:30 to 05:49. 3 attempts, then second provider, also blocked. Cloudflare, not cleared this morning. | $6.45 |
| Seller APIs, the web store and the order hubofficialofficial API | this morning, 05:47 | Our own price, stock and listing state on 7 channels.Seller APIs and the order hub. This is the one leg of the round with a contract behind it, so it runs even when the crawlers do not. | — |
| Shelf scanread this morningcrawler | this morning, 05:50 | 10 watched words, 3 pages each, from a fixed location at the same minute every morning.Signed out, no history, no basket. A rank read at nine and set against one read at six is a comparison of two different pages, so the conditions are held still and printed on the tab. | $0.63 |
| Category and bestseller scanread Mondaycrawler | this morning, 06:00 | 5 watched aisles: seller counts, the first page's price spread and the bestseller list.Weekly, because these lists do not move enough to be worth reading daily. It publishes no volume of any kind, and the tab says so in the first block rather than the last. | $0.04 |
| Pazarama and PttAVM · competitor shelvesnot writtencrawler | never | Nothing. Our own price and stock are read there through the seller API; the competitor side is not.No managed crawler exists for either shelf and nobody has written one. They are listed here so the report never implies coverage it does not have — the briefing says four shelves because it watches four. | — |
| Sales Cockpitread acrossinside Omni | continuous | What each product of ours brought in over the last 28 days.This module holds no sales figure of its own and should not. It is what the morning list is ranked by: a rival cutting a shape you barely sell is not the same news as one cutting the shape that pays the rent. | — |
| LLM gatewayMonday passinside Omni | this morning, 06:30 | Groups the low-star reviews already collected into complaint themes.Only the one- and two-star reviews on the listings already being watched, and only the ones the shelf shows without a login. No new connection and no new crawl: the text was already in this morning's snapshot. The model groups; it does not decide, and every theme carries the count it came from. | — |
| Company brainreads and writesinside Omni | continuous | The noise marks, the closed gap decisions and the threshold history.3 marks a reader has left on this list so far. They sink a shape of line rather than deleting it, and a gap candidate somebody closed is never raised again by the machine. Two customers with the same watchlist get different mornings, because their histories differ. | — |
| Automation Builderconnectedinside Omni | on threshold | Carries a threshold alert to WhatsApp, Slack or e-mail.Quiet hours and who gets which rule live in the backbone's notification settings, not here. This is the one thing that moves off the page, and it moves to your own team. | — |
The two rows that are not green are the module in miniature. One shelf refused this morning and its listings are Sunday's wherever they appear; two more have no competitor-side crawler at all, and are listed here so no page quietly implies coverage that does not exist.
What it costs
Computed from this watchlist, not quoted from a price list: listings × the crawler's per-thousand rate × a round every morning.
| Line | A month | What is behind it |
|---|---|---|
| Crawling the watched listings | $13 | 342 listings across four shelves, once every morning, at each crawler's per-thousand rate. N11 is $6.45 of it — 49% of the bill for 13% of the listings, because its actor costs seven times what the others do. It is also the shelf that failed this morning. |
| Reading the shelf for each watched word | $0.63 | 10 words × 3 pages × 30 mornings. Every word added to the Watchlist adds three page reads a day forever, which is why the list is short on purpose. |
| The Monday aisle scan | $0.04 | 5 aisles, four times a month. The gap list and the review themes are drawn from data already collected and cost nothing extra to produce. |
| Residential exits | $35 | About 4 GB a month at $8 a gigabyte, assuming 0.4 MB a page. It is larger than the crawler fee, which surprises everybody: getting through the bot protection costs more than reading what is behind it. |
| The whole watchlist, a month | $49 | Usage-based and metered to the credit account. The listing fee is linear in the size of the watchlist, so a watchlist fifteen times this one costs about fifteen times this much — which is the argument for a small list rather than a policy about one. |
Usage-based, metered to the credit account, and linear in the size of the watchlist. Adding a seller costs their whole catalogue every morning; adding a word costs three page reads a day for as long as it stays on the list.
What it does not measure
Eight figures a competitor report is expected to produce and this one does not. Stated once, here, so no tab quietly invents one of them later.
| Not claimed | Why not |
|---|---|
| How much anybody sells | No marketplace publishes units or revenue for a listing. Nothing in this report should be read as a guess at a competitor's sales, and no page here converts a bestseller position into one. |
| How deep a rival's stock is | In stock or not is the whole of what a listing page shows. “Out of stock” is a fact; “running low” would be an invention. |
| What anything costs them | Their margin is not on a listing page and never will be. It is why the gap list carries no profit figure: what a candidate would earn depends on an invoice this module cannot see. |
| How often a word is typed inside a marketplace | None of them publish in-site search volume. This module measures a position, not the traffic behind it. Search volume for the open web is a different question and it belongs to the SEO & GEO Radar, which buys it from somebody who measures it. |
| Why a rank moved | A marketplace never says. The Shelf tab can tell you that you fell from nineteen to twenty-six overnight and who is standing above you now; the reason is a person's job. |
| What the others charge on their own sites | Every one of the 5 watched sellers runs a shop of its own, and each would be a separate crawler with its own protection and its own breakage. The marketplaces are where a buyer actually compares, so that is where the money goes. |
| Anything before 16 Jun 2026 | The snapshot store starts there and ninety mornings is all it holds. A year-on-year comparison is a promise this store cannot keep yet, and no tab offers one — which is also why the period pills stop at ninety days. |
| Whether a move under 1% means anything | It is treated as the shelf rounding itself and never reaches a person. It is counted on the Changes tab so the ladder can show it falling away, and that is the whole of its career. |
Three of these eight could be produced by dividing two numbers this module already holds. They are missing because the division would answer a different question from the one the label promises, and a plausible answer to the wrong question is the most expensive thing on a competitor screen.
Sample data throughout: the shelf, the five sellers and ninety mornings of snapshots are generated from one seed, and the crawler names and per-thousand rates are the shape of the real ones rounded off. What is not simulated is the arithmetic. The matcher reads titles and barcodes, the bands refuse to draw under 3 offers, the ranker weighs a change by the revenue behind it, an alert fires only where an exact match lands more than 8% under our price, and the sale-that-never-ends is read out of a price history rather than written into a fixture. All of it would run unchanged against a live shelf; only the provider behind it changes.