How we measure
Travel prices are the easiest thing in the world to make up, and almost nobody shows their working. Here is ours.
Every price carries three things
A price on this site is never a number on its own. It carries the date it was observed, the platform it came from, and the query that produced it — the same dates, the same occupancy, the same filters, every time. If we cannot give you those three, we do not publish the number.
We started a year early
Price recording began in August 2026, roughly a year before the eclipse. That is deliberate. A site launched in spring 2027 can show you today's price; it can never show you what the same trip cost twelve months earlier. The curve is the part that tells you whether you are looking at a good deal or at a panic.
Readings are taken on a fixed schedule rather than when a figure would be convenient, and the early manual readings are labelled as such, at lower confidence, so you can tell them from the automated ones.
Two adults, and the solo traveller measured separately
Lodging is quoted for two adults sharing, because that is what most of this stock is priced as. The cost for one person travelling alone is measured separately, never derived by halving. In this part of the world a whole unit costs the same whether one or two people sleep in it, so dividing would understate the solo cost by roughly a factor of two.
Never a conclusion from a single platform
An early analysis of one booking platform said a certain Egyptian city was priced out of reach entirely. Changing platform showed that was wrong. So a claim about availability now names the platform it is true on — "on this platform, this share of stock is gone" — and never asserts that a place is full in general. One inventory is an observation. It is not a conclusion.
Freshness, stated plainly
| Label | Age | What we allow ourselves to say |
|---|---|---|
| Fresh | 7 days or less | May be called a deal, with the exact date shown |
| Aging | 8 to 14 days | Usable, shown with a warning |
| Historical | Over 14 days | Kept as an observation, never called a deal |
Eclipse times come from calculation, not from the web
Durations of totality, contact times and whether a given spot falls inside the path are computed from published astronomical elements, never copied from another website and never taken from a travel company. Every figure is checked against NASA's own trajectory table before it is published. The calculation runs on our side for now; the calculator you can point at your own coordinates is being built.
Cloud: two reanalyses published side by side, a satellite to break ties
Cloud cover is the other number that decides this trip, and it is measured the same way as a price: from a source you can go and check, over a window we state, with the disagreement left visible instead of averaged away.
Every site is measured over 28 July - 7 August across 2001-2024 — 88 sites along the 2027 path, 264 days of daily means each. Two independent reanalyses are read: NASA POWER (MERRA-2) and Copernicus ERA5, from the Copernicus Climate Data Store.
They agree to within ten points on 66 sites, and both figures are published as a range with the source at each end named. On the 22 sites where they disagree — up to twenty points apart, which is the difference between a good bet and a bad one — a third and independent source arbitrates: the EUMETSAT CM SAF CLARA-A3 satellite cloud record, calibrated against the two reanalyses before it is allowed to vote.
The figure each climate section leads with is not the one a weather site would show you. A daily mean averages cloud over the night as well, and nobody watches an eclipse at 03:00. So each site also carries cloud over the hours totality actually falls in, which along this path ranges from 7.3 points below the 24 h mean of the same ERA5 data to 14.2 points above it. As far as we can tell nobody else publishes it.
Two caveats travel with these numbers because both of them moved a figure. Coastal sites are read on the nearest land cell, not on the cell containing their coordinates: 28 of them are affected, and on one Spanish town that single decision moved the answer from about 25 % to 10 %. You stand on land, so we read land. NASA POWER chooses its own grid cell server-side, so the same correction cannot be applied to that end of the range — where it matters, the page says so.
And what these are not: climatology is history, never a forecast. No forecast exists for August 2027, and anyone showing you one is selling something. A clear record does not promise a clear morning; it tells you the odds you are accepting.
- NASA POWER — MERRA-2 CLOUD_AMT, daily means 2001-2024
- Copernicus ERA5 — Generated using Copernicus Climate Change Service information 2026
- EUMETSAT CM SAF — copyright 2026 EUMETSAT (CM SAF CLARA-A3)
What money does not buy here
A site that tells you what a trip costs is worth exactly as much as its independence from the people selling the trip. So, the money, plainly: as of October 2026 this site earns nothing — no affiliate links, no sponsored placements, no advertising, no commercial relationship with any hotel, airline, booking platform or tour operator. A domain name and a free hosting plan, paid by the publisher.
That may change. If a paid link ever appears, these four hold:
- A ranking is never for sale. The cost scale is ordered by measured cost and by the risk you carry on the morning. A commission cannot move a rung, and no operator is added to a page because it pays.
- Every paid link is marked as one, at the link itself — not only in a policy page you would have to go looking for.
- This section says what is earned, and stops saying "nothing" the day that stops being true.
- No page is written because it can be monetised. The measurements decide what gets published, in the order they arrive.
And what we will not take: a free stay, a press trip or a discounted room in exchange for coverage. The whole point of this site is that the price you read is the price a stranger pays, measured from outside. A figure obtained as a favour is not a measurement.
When we are wrong
Prices move and inventories change. Where a figure has been superseded we replace it and keep the observation date visible, so you can always see how old the ground under a recommendation is.
The data itself
The figures described here are also published as files, with their dates and sources, on theopen data page.