One reading a week, and the date it was taken.
Volatilo is a weekly record, not a live feed. Every figure on this site is a single observation taken on a stated date, and every claim made from those figures is a claim about our own record — not about what Amazon is doing right now. This page describes how that record is built, what is derived from it, and the things it is deliberately not allowed to say.
Collection
Every tracked product is read once a week, on a fixed schedule, Saturday at 02:00 UTC. That reading is stored with its date and is never overwritten by a later one. The site is rebuilt from the stored record, so a page published on Monday describes Saturday's observation and says so on its face.
Coverage is the top 100 of each of the 23 Amazon leaf categories above, plus every product that has recently left one of those lists and is still being watched.
Two different spans sit behind this site and it is worth being plain about which is which. Volatilo's own weekly collection began in August 2026, and everything about category composition — what entered a list, what left it, how much of a top 100 turned over — is measured only from then. Separately, a subset of products carries a longer price and rank record retrieved from the source at the point we began following them. Where a page speaks about a span longer than our own collection, that is the retrieved record it is describing, and the page says how long it is.
What a number on this site means
It means: on the date printed beside it, this is what we read. It does not mean the value holds today, and a figure that no longer matches Amazon is usually not an error — it is last week's reading doing its job. For the current price and availability of anything, go to the listing itself.
A week can also carry no reading at all. Amazon may return no offer, or the listing may fail to resolve. Those weeks are recorded as what they are and excluded from averages, minima and change counts. They are never silently treated as a price of zero.
What Volatilo does not publish
This is the shortest section and the one that constrains the most. Volatilo does not publish live Amazon prices, discounts or promotions. We observe a price once a week and we record the date. We do not observe Amazon's list price, deal events, coupons, promotional flags or stock status, and nothing on this site may be written as though we do.
So there are no discounts, no deals, no sale claims and no percentage-off framing anywhere on this site. Where a week returned nothing, the site says no offer was recorded in that observation — never that a product is out of stock, which is a thing we would have to know continuously in order to say.
The rule is not a style note. A check scans every generated page at build time and fails the entire publish rather than let one sentence through, so the constraint holds even on the weeks nobody is looking.
How rank is written
Amazon's sales rank counts downward: a smaller number is a stronger position. Rank movement here is always written so that improvement reads as improvement — a product moves toward the front of its category or away from it, and the size of the move is stated without a sign. You should never have to work out whether a negative number is good news.
When a signal appears
A signal is shown only once the window behind it has actually filled. A four-week trend needs five observations; an eight-week slope and its goodness of fit need six; a twelve-week price comparison needs twelve; a year-long range needs a year. Until then the signal is absent rather than estimated.
A product page therefore carries fewer statements in its early weeks than it will later, and that thinness is the method working rather than the page failing. Each page shows how far into that horizon it is, so the difference between nothing to report and not enough record yet stays visible.
Products that leave a list
Dropping out of a top 100 is not the same as declining, and neither is the same as being unavailable. A product that leaves a list enters an eight-week watchlist and keeps being read, so that a temporary disappearance can be told apart from a lasting one before anything is claimed about it.
Each week a product is recorded as listed, as ranked but below the cutoff, as carrying no offer, or as unreadable. Only the second of those supports a statement that a product has lost ground. The others describe our view of it, not its performance.
Price rhythm
Where a listing has enough history, we look at whether its price moves with any regularity — how far below its usual shelf price it tends to go, how long those stretches last, and whether they arrive on anything resembling a schedule. Where a rhythm is legible we say so and say what it implies about waiting. Where it is not, we say that instead.
These are descriptions of a past record, not predictions and not alerts. Volatilo does not track prices for you, does not notify anyone when a price moves, and has nothing to sell that would depend on it.
Who writes the sentences
The sentences on this site are written and approved by a person, not generated. The figures inside them are computed from the stored record, and which wording a given product gets is decided by a fixed rule from its own identifier — so a page reads the same way from one week to the next, and a change in what it says means the record changed rather than the page did.
No language model runs in the pipeline that builds these pages. There is no such step: nothing is generated when a page is rendered, and no model decides what a page is allowed to claim.
The software itself is a separate matter, and an honest one. AI coding assistants helped write the collectors, the queries and the templates that build this site, which is a different thing from generating what it says. That is set out in full on the about page.
Source and attribution
The underlying observations are retrieved through the Keepa API, and are published here as derived weekly signals with Keepa's written consent. Volatilo is not affiliated with, endorsed by, or operated by Amazon or Keepa.
What this site publishes is one reading a week. If you want the continuous, high-resolution history behind it — every price change rather than every seventh day — that is Keepa's product and Keepa is where to get it.
Corrections
A figure that was wrong when it was read is worth telling us about: a rank attributed to the wrong product, a price that never applied, a product matched to the wrong category. Write to us with the page URL and the observation date printed on it, and the reading can be traced to the row it came from.