Method

How we pick

Last updated:

The best trips from each city, found around the clock from live prices, ranked by an editorial scoring with AI assistance, every number checked by code. This page says exactly what that means.

Where the numbers come from

  • Flights. We screen dozens of destinations per city against live round-trip fares for the exact dates (one adult, taxes and fees included), then price the finalists again with full totals for two adults from the airlines and the big flight search engines. The price on a page is the cheapest total we found for two, divided by two, at the time shown.
  • Hotels. For each pick we pull up to eight hotels for the dates and keep three: the best value (lowest price among well-reviewed places), the top rated and a budget pick. Prices are the total for the stay as the booking site showed it.
  • Weather. A 16-day forecast when the trip is within it; otherwise the average of the same dates over the last five years. Weather data by Open-Meteo.com.
  • Price history. Every fare we check is kept, so each route has a typical price (the median of the last 60 days) and a lowest-seen price. A page says "Lowest in 30 days" or "n% below typical" only when at least five checks support it.
  • Destinations. A catalogue of about 190 places with their airports, best months, areas, signature experiences and price level, written and maintained by the team.

How a trip becomes a pick

  1. For each horizon (next weekend, the weekend after, a week away) we keep the destinations that fit the trip length, the distance and the season.
  2. Each candidate gets a value score: the route's typical price divided by today's price, a bonus for a nonstop flight, a penalty when the forecast is wet or the temperatures miss what the destination is for, and a penalty for expensive weekends.
  3. The top four per horizon are priced in full. An AI model reads their numbers (price, typical price, nonstop, duration, weather, the kind of place, the last two picks) and chooses one per horizon with a one-sentence reason. Code enforces the rules: three different destinations, the pick must be among the finalists, and when the model fails the best value score wins.
  4. The page is published with the facts first. The text (why go now, where to stay, what to do, the questions) is drafted afterwards by an AI model from the data on the page only: every number in the text must exist in the data or the draft is rejected, and a draft that fails twice is dropped for the plain facts. The team reviews pages and fixes or removes what readers report.

Cadence

The search runs around the clock: every few minutes it takes one city and publishes one new find from live fares, a city's picks are refreshed when readers visit and the prices have gone stale, and cities with recent visitors are refreshed again every night. Readers can also ask for another trip from their city, tweak a pick (longer, shorter, cheaper, sunnier, closer, another destination) or ask a question about it; these run one at a time for the whole site and build a page the same way. A pick stays until its departure date, then moves to the archive with the price it had. Pages nobody visited for four months after departure are retired.

What we never do

  • No made-up reviews or testimonials; the review scores are the booking sites' own.
  • No urgency: no countdowns, no "only two left".
  • No badge without data behind it.
  • No pick is ever sold. Links pay us a referral fee when you book; that fee changes neither your price nor the ranking. Details.

Corrections

Found a wrong number or a stale price? Email hello@vacaymatch.co with the page link. We check within a business day and fix the page or take it down.