How often does a price actually change?
Rarely. In 64 days of reading every shop twice a day, 418,682 listings changed at least once — 19.5% of everything we track — and only 126,287 moved more than once. The overwhelming majority of prices simply sit there.
Changes per day, the last two weeks
| Day | Changes recorded | |
|---|---|---|
| 2026-10-16 | 17,942 | |
| 2026-10-17 | 8,424 | |
| 2026-10-18 | 4,020 | |
| 2026-10-19 | 7,341 | |
| 2026-10-20 | 10,821 | |
| 2026-10-21 | 15,771 | |
| 2026-10-22 | 51,045 | |
| 2026-10-23 | 20,445 | |
| 2026-10-24 | 39,966 | |
| 2026-10-25 | 4,238 | |
| 2026-10-26 | 7,443 | |
| 2026-10-27 | 16,121 | |
| 2026-10-28 | 19,034 | |
| 2026-10-29 | 13,994 |
Lumpy, and that is the finding. Repricing is not a steady drizzle across the trade — it arrives in bursts when a large retailer runs a seasonal update, and on quiet days almost nothing moves at all. We publish the raw daily count rather than a smoothed average so those bursts stay visible.
What this record is, and why we started it
Every shop in the corpus is read twice a day, and any difference from the previous reading of the same listing is written down with its timestamp. The log begins on 2026-08-27 and now holds 2,786,744 records across 64 days.
It exists because the alternative is to trust a shop's account of its own history, and that account is always flattering. A struck-through price is a claim about the past that only the claimant can check. Once you have read the page yourself, twice a day, for a month, you no longer need the claim — and the first thing the record shows is how little most prices move, which is precisely what a permanent discount badge is designed to obscure.
The depth is still modest, and the page says so rather than dressing 64 days as a trend. For the long view we use dated third-party captures instead, and we never mix the two: see what a decade of archived snapshots says.
What we refuse to conclude from it
- 64 days is not a season. Nothing here supports a claim about annual cycles, and the page that tests one uses years of captures instead — is Black Friday actually cheaper.
- A change is not a discount. Up, down, and stock movements all count as changes. Direction is a separate question with its own page.
- Our first reading of a listing is not a market event. It is excluded, which is why the headline figure is far below the raw record count.
- A shop we could not read that day contributes nothing, and its silence is not stability. Sweeps do occasionally miss a host.
Why twice a day, and not once
The sweep runs at ten in the morning and ten at night, Montreal time, and the second reading is not redundant.
The corpus is global — shops across seventeen countries — and a single daily reading leaves whole time zones a full day behind. A European shop that reprices in its own morning would be read many hours later; by the time the figure appeared it would already be stale for the readers most likely to act on it. Twelve hours apart cuts that blind spot in half.
It also changes what can honestly be claimed. A drop is defined as a lower price than our previous reading of that same shop, so the meaning of “a drop” is directly tied to how long ago we last looked. At one reading a day, a price that fell and rose back would be invisible. At two, it is at least detectable — and the burst pattern in the daily column above only resolves at all because the window is short enough to separate one retailer's overnight update from the next day's trade.
The cost is the politeness budget: one request every 3.5 seconds to any host, four hosts at a time. Reading twice as often means the sweep takes as long as it takes, and it is the reason nothing on this site is ever real-time.
Questions this raises
Should I wait for the price to drop?
On most listings there is nothing to wait for. Over 64 days of twice-daily readings, 19.5% of listings moved at all. The exception is the product that has run out everywhere, where waiting has already failed — see sold out everywhere.
Why do some days show far more changes than others?
Because shops repricing are lumpy: a single large retailer running a seasonal update can move tens of thousands of listings in one night, and the corpus-wide count jumps. The daily column is deliberately raw so those events stay visible rather than being smoothed into a trend.
Is a change the same thing as a discount?
No. A change is any difference from our previous reading of that listing at that shop — up, down, or in and out of stock. Whether a move is a real bargain is a comparative question, answered on the deals and fake-discount pages.
How is this different from the archive pages?
This is our own log: every shop, twice a day, since 2026-08-27. The archive pages reach back years using dated third-party captures, which are far deeper in time and far sparser. This one is shallow and complete; that one is deep and patchy, and we never mix them.
Why does the first reading not count as a change?
Because it is us arriving, not the market moving. Every listing generates a record the first time we see it, so counting those would report our own coverage as price activity. Only listings seen changing at least twice are counted here.