Why there is no score out of 100
Open a foraging app in Sweden and you will be shown a number. Eighty-two. Sixty-four. A little dial, a colour, a week of bars. It is a satisfying thing to look at, and it is the reason this product exists — because that number cannot mean what it appears to mean.
What the number would have to know
To tell you that today is an 82, a model has to have learned the relationship between conditions and finds. That needs a response variable: a record of searched here, found this much — and crucially, searched here, found nothing.
The public data for Swedish foraging has no such thing. Artportalen and GBIF hold presence-only records: somebody saw a Cantharellus cibarius and reported it. There is no matching record of the nine walks that turned up nothing, because nobody reports those.
Fit a model to presence-only data and it learns where people walk. Roads. Car parks. The edges of towns. Popular woods on a Sunday afternoon. That model will be confident, and it will be describing human behaviour with a mushroom's name on it.
What runs instead
A transparent rules engine. Two moisture clocks — a weeks-scale trigger and a days-scale expansion — composed with accumulated warmth and with terrain, as a necessary-AND rather than a weighted blend. Every threshold is a percentile against local climatology, because Sweden spans two climate regimes and a national constant misfires by construction.
The output is a conditions-favourable read on a 0–1 scale. It is called that everywhere it appears, because it is not a probability, and calling it one would be the same lie in a different costume.
The part that is actually yours
The regional engine is capped, honestly, at favourable within about ten kilometres — necessary, not sufficient. Nothing derived from public data does better, and saying so is cheaper than pretending.
The only thing that can sharpen a read to a single place is your own history: the spots you marked, the finds you logged, and the blanks. A day you searched and found nothing is the rarest and most valuable thing the model ever receives, because where they weren't is exactly what a regional forecast can never know.
That is why the collection is half the product rather than a retention feature. It is also why the app asks how long you were out and how hard you were looking — without that, a gentle family walk reads as bad ground.
The test
An instrument is worth believing on the good weeks only if it told you the truth on the quiet ones.
So when the stations stop answering, the number disappears rather than coasting on yesterday. When a species is genuinely unforecastable — cloudberry, truffle — it says so and explains why, instead of producing a read nobody can stand behind.
A missing number is more useful than an invented one. It is just harder to sell.