There is an enormous amount of confident advice about when to book flights. Tuesday afternoons. Exactly fifty four days ahead. Clear your cookies. Use a private browsing window.
I spent a year checking six routes daily, at the same time each day, recording the cheapest available fare. It is a small sample and it is my own data rather than someone else's assertion.
The private browsing thing is not real
The most persistent claim, and the one I most wanted to be true.
I checked the same routes in a normal window and a private window, on the same machine, within a minute of each other, repeatedly. The prices matched every time.
Prices did change between one day and the next, sometimes substantially, and they changed identically in both windows.
What I think is happening is that people search, see a price, come back later, see a higher price, and attribute it to being tracked. Fares genuinely do move, frequently, for reasons that have nothing to do with the individual.
There is no magic day of the week
For booking, that is. I could find no consistent pattern in which day of the week was cheapest to buy on across six routes over a year.
The day of the week you fly is a different matter and there the effect is large and consistent. Midweek departures were reliably cheaper than Friday or Sunday on every route I tracked, sometimes dramatically on the business-heavy ones, which behaved in the opposite direction.
Which suggests the advice got garbled somewhere. Flight day matters. Purchase day does not, at least not detectably.
The booking window is real and wide
The clearest finding. Prices were high a long way out, fell into a broad trough, and rose steeply close to departure.
The trough was wide — roughly two to four months ahead on the long-haul routes, and about one to three months on the short ones. Within that window there was no reliable optimum, just noise of maybe ten or fifteen percent.
Which means the precise-number advice is meaningless. Book somewhere in the broad trough and you have captured almost all of the available saving.
The steep rise close to departure was the most consistent thing in the whole dataset. Prices in the final three weeks were dramatically higher on every route.
Last-minute bargains barely exist any more
I checked specifically for this because the folklore persists.
Across a year of daily observation on six routes, I saw prices fall in the final fortnight on a handful of occasions, and never to a level below the trough price.
Airlines manage inventory well enough now that empty seats near departure are priced for the people who have no choice, not discounted for the flexible.
Charter and package holidays are a different market with different economics and the old advice may still hold there. For scheduled flights it does not.
Flexibility on dates is worth more than anything else
The single largest saving available, by a wide margin.
On several routes the difference between the cheapest and most expensive day within the same week exceeded thirty percent. Across a month it was often more than double.
Which means an hour spent looking at a flexible date grid saves considerably more than any amount of cleverness about when to press buy.
Nearby airports too. On two of the routes I tracked, an airport an hour further away was consistently cheaper by enough to pay for the transfer several times over.
Error fares and sales
The genuine bargains I saw during the year were of two kinds.
Scheduled sales, announced, lasting a few days, generally for travel in low season. These are real and they are predictable enough to plan around if you are flexible about where you go.
And a small number of what looked like pricing errors, which appeared and disappeared within hours. Catching these requires monitoring, which is what the alert services do.
Neither of these is a booking strategy so much as an argument for having somewhere in mind and being ready.
What I do now
Decide roughly when, not exactly when.
Set alerts on the route two to five months out and let them run.
Look at a flexible date view before fixing anything, and be willing to shift by a few days.
Book when the price is somewhere reasonable within the trough rather than waiting for a bottom that I now know I cannot identify in advance.
And stop clearing cookies, which achieved nothing and made me log into everything again.
The caveat
Six routes, one year, one person checking. Airlines run pricing systems of considerable complexity and my data would not survive contact with anyone who does this properly.
What it did do was let me stop believing several things I had believed for years on no evidence at all, which was the point.