Why we refuse to score 40% of the routes people ask us to watch
A thin price history produces a confident-looking number that is worse than silence. Here is exactly what our confidence gate rejects, and why we set it where we did.
Someone asks us to watch Bristol to Ljubljana. We start recording the fare that day. Three weeks later they have had no email from us, and the honest reason is that we still do not know enough about that route to tell them anything useful.
We could send something anyway. Most price tools would. We think that is the single most damaging thing a product like this can do.
What the gate actually checks
Every fare we record is filed against three things: the route, the departure month, and how far ahead of departure it was observed. That combination is the cell we compare a new price against — because a fare 90 days out in March is not comparable to one 5 days out in November, and pretending otherwise is how you get a confident wrong answer.
Before we will score anything, that cell has to contain enough observations:
| Observations in the cell | What happens |
|---|---|
| Under 20 | No score is produced at all |
| 20 to 59 | A score exists, but no alert is ever sent |
| 60 or more | Eligible for an email |
The two-step shape is deliberate. A thin baseline is good enough to show someone a number they have explicitly asked to see. It is not good enough to spend an unrequested email on.
Why silence is the cheaper mistake
The costs here are wildly asymmetric.
A missed alert costs one opportunity. You book at a fair price instead of a good one, and you never know the difference.
A bad alert costs the entire relationship. You open it, act on it, and discover the price was ordinary. You have now learned that our emails are not worth opening — and that lesson is permanent. Every genuinely good alert we send afterwards arrives to someone who has already decided to ignore us.
One misleading alert does not cost one email. It costs every future email.
The number is provisional, and we will say so when it changes
Sixty is a starting guess. It comes from wanting roughly two months of daily observations behind any claim we make about a route, not from anything more principled than that.
We are watching what it does to our open rates. If the data says the threshold should be higher, we will raise it and write about why. If it says we are being too cautious and leaving people uninformed for no reason, we will lower it and write about that instead.
What we will not do is quietly ship a number we do not believe because silence felt unproductive that week.