We don’t.
Your published phenol figures set the bounds on smoke. Your cask declarations set the direction on fruit and spice. Tasting prose only positions intensity inside them. Every number in your profile traces back to the evidence that produced it. What we can show you depends on whose evidence it is: where it is yours, or our own tasting, we can put the record in front of you. Where it is a third party’s writing, held privately as working material, we can tell you what it is and where it came from and walk you through the derivation — but we will not reproduce someone else’s words to do it. Either way the answer is a record, not an opinion.
Bring us your lab data and the platform treats it as first-class.
Where this stands today: declared data — cask, mashbill, age, strength, phenol — is captured with its provenance and shown to guests now. Two things are deliberately not yet true of it. The calibration that lets those figures shape a score is unwritten, because a prior guessed to look finished moves real numbers invisibly. And the founding catalogue was seeded from a curated export rather than derived, so those bottles carry profiles without an evidence record behind them. Everything mapped from here forward goes through the derivation path, and the seeded profiles are being re-derived against real evidence rather than left as they are.
Read the method →An event is a time-boxed instance you create in minutes: name, dates, venue, and a hand-picked subset of the catalogue, so the compass matches only the drams you’re actually pouring.
A guest steered to the right expression tastes happier and buys the bottle. The compass runs as a kiosk in your tasting room, scoped to your own range, in your branding.
The tasting-room surface is in development and does not ship today.
Almost no distillery measures what visitors actually wanted at the point of tasting. Point of sale tells you what sold, which is a function of what you poured and what your host recommended. This is expressed intent, before the pour.
Sixteen flavour characters, each anchored to a congener class rather than to a style or a region. Matching is deterministic geometry over those characters, and the formula is published in full — you can check a score by hand.
There is no model in the guest path. The same taps give the same answer every time.
Producer work begins with your data and what you want to learn from it, so it starts with a call rather than a sign-up form.