Why Sparkfare doesn't predict future prices

Some flight tools will tell you a price is likely to drop, or advise you to wait. Sparkfare never does either. Every number on the board describes what already happened — today's price against this route's own real trailing history — never a forecast of what happens next.

How this compares to what's actually out there

We looked at this directly rather than assuming our approach was unique. Hopper frames its claim as an explicit percentile — "better than 95% of prices we've seen in this market" — a real historical comparison, not a prediction, but still a broader market-wide benchmark than a single route's own history. Expedia flags fares 20%+ below an ML-estimated "typical price," computed across roughly 2 million flights analyzed daily — genuinely a modeled estimate, not a straight historical average. Both are legitimate approaches. Neither is what we do.

Why we stopped short of a model

A prediction is a claim about the future, and future claims can be wrong in ways a straight historical comparison can't. "This route's cheapest fare today is 20% below its own 30-day average" is a statement about the past — checkable, and either true or false the moment it's made. "This price is likely to drop" is a bet, however well-informed. We'd rather ship the first kind of claim than the second, even though the second is arguably more directly useful for deciding whether to wait.

What this means in practice

A "deal" badge on Sparkfare has never meant "this price is as low as it will ever get" — it means "this price is genuinely below what this route has actually cost over the past month." Whether it goes lower still is a real, open question we simply don't claim to answer. Full mechanics in how we rank deals.

See today's board