How we read price history - and why "buy" isn't guessing

A price drops, a green badge appears, and the instinct is to click. But "cheap" alone says nothing - cheap compared to what? For the word BUY to mean anything, it needs a reference: the actual price history of that exact route, not a gut feeling, and certainly not someone else's "−40%" in an ad.

Median, percentile, trend

For every route or destination we track, Voyscout keeps its own running series of prices over time - that's the baseline everything else gets compared against. From it we compute the median: the price below which exactly half of what we've observed has fallen, and above which the other half sits. Today's offer then gets ranked inside that series as a percentile - say, "cheaper than 92% of everything we've ever seen on this route." And because one price alone says nothing about direction, we also track a moving trend: a recent-weeks average against an older one. The bar for calling that trend "rising" or "falling" scales with how much the route naturally swings - a calm route needs only a small move to count, a volatile one needs a bigger shift, or every wiggle would look like a trend. For flights we build that series over a whole grid of dates and airports - exactly how is shown in How we watch flight prices.

An illustrative example: say a weekly package has shown prices between ~22,000 and ~31,000 CZK/pp over recent months, with a median around ~25,900 CZK/pp. An offer at ~23,000 CZK/pp sits in the cheapest quarter of that route's own history - a different kind of statement than a percentage in an ad, because it's measured against real data, not a guess.

When BUY actually fires

We only let ourselves say BUY when several things line up at once: there's enough history behind the series for it to mean anything (a handful of days is never enough), the price sits at or near its own historical floor, and the trend isn't actively rising against you. Miss even one condition and the verdict softens:

• BUY - deep history, price at its own floor, trend not rising.
• WAIT - price sits in the pricier part of its own history, or is trending up.
• WATCH - the honest default: either too little history, or the signal isn't clean enough to be confident.

Even a seemingly clear BUY can fall back to WATCH when we're not confident in the data underneath it - the price is older than our freshness window, it came from a search result never verified directly with the seller, or it's being compared against a different travelling party or a different week than the history behind it. We never let shaky data produce a confident BUY.

A verdict is statistics over observed price history - not a guarantee the price won't drop further.

That's the actual difference between guessing and a verdict: real numbers, a reference series, and clear conditions for when we dare to make a call - and when we say we're still just watching. The full overview of the methodology lives on How Voyscout works.