How old are the fares in flight deal alerts? What 17,712 cached prices showed us

Many flight deal alerts don't search the airlines when they email you. They read a fare cache: a record of prices that travellers recently found on a flight-search site, kept for a while and served to anyone who asks. It's cheap, fast and usually close enough. It's also how an alert can quote a fare that was found days ago and may already be gone.

We read from a cache like that too, so we measured it. For eleven nights, October 19 to 29, 2026, we asked it about two dozen popular routes for every departure day out to six months, one-way and round trip, and kept everything it told us. Here is what was in it.

Most of what a cache shows you tonight, it showed you yesterday

Over the eleven nights we recorded 17,712 fares. Counted once each, by the day the fare was originally found, they come to 5,343. The rest were the same fares coming back night after night.

After the first night, the pattern was steady: about 1,600 fares a night, of which about 370 were new. Roughly three in four fares the cache returned on any given night were ones it had already returned before, unchanged, with the same date found. The entries we saw had been found anywhere from that day to about a week earlier.

None of that is the cache doing anything wrong. Holding a fare for a few days is what a cache is for. But it changes what an alert built on one can honestly tell you:

• The price may be days old. A fare found last Tuesday can still be sitting in the cache on Saturday, long after the seats behind it sold. That is why our alert says it is the lowest fare we've seen, and why its link opens a fresh search rather than a booking page insisting on a price that may have gone.
• Repeats are not evidence. A service that counts each night's copy as a new data point thinks it has three or four times more history than it really does, and will call a fare "unusually low" against a much thinner record than it claims. We count each fare once, on the day it was found, which is a large part of why we refuse to score some routes for weeks before we say anything.

About one answer in eight was for a different airport

A fare cache usually answers by city, not by airport. Ask it for New York JFK to Los Angeles and it may answer with a flight into Ontario, still labelled with the city code LAX, because Ontario is in the Los Angeles area. For a traveller browsing options that's helpful. For a price history it's a mistake: a fare into a different airport isn't evidence about the one you asked for.

On the night of October 29 the cache returned 1,707 fares at the airports we asked for and 247 at other airports in the same city areas: about one in eight. It wasn't spread evenly. Routes touching Los Angeles were the worst, with around a third of JFK–LAX answers in each direction at another airport and about half of LAX–Las Vegas. Our transatlantic routes, and the New York–Chicago ones, had none.

We refuse every one of those rather than record it, because a wrong row poisons every score computed from it later. If you get a deal alert from anyone, check the airport in the itinerary itself, not just the city in the headline.

Four departure days in five had no fare at all

On a typical night we asked about 7,940 route-and-departure-day combinations and the cache had a fare for about one in five. Popular routes did better, quieter ones much worse, and dates far ahead were the thinnest of all.

So "no fare" from a cache means nobody has searched it lately, not there are no seats. An alert service that goes quiet on your route may simply have nothing to read. That's one more reason we say nothing rather than guess: an alert built on a gap is worse than no alert.

What this means when you read a deal alert

• Treat the price as a lead, not a quote. Open the link and check the live fare before you plan around it. The fare in an alert, ours included, is the lowest one seen, and bags and seat selection are extra, as we explain in is basic economy actually cheaper.
• Check the airport, not the city. About one cached answer in eight was somewhere else.
• Be wary of confident claims built on short histories. Ask how much history a "lowest price in 90 days" claim rests on, and whether repeats were counted once. Our answer to the first is on how the Price Hash Score works; it's also why we explain what "26% below typical" means before we ever say it.

We'll update these numbers as the record grows. If you'd like a fare judged against a history counted this carefully, watch a route — it's free while we're in beta.