How we score whisky flavour

The eight axes

Two worked examples

The same eight numbers, on two whiskies that sit at opposite ends of the smoke axis.

Ardbeg scores 9 on Peaty and 1 on Sherried; the Macallan is the mirror image at 0 and 7. Both score 4 on Cereal, which is why they still taste like the same drink.

How a score is assigned

1. Tasted neat at room temperature, in a nosing glass, with water available but added separately.
2. Each axis scored independently, against the anchors in the table above.
3. Scored against the anchors, not against the flight. A whisky does not become smokier because the one before it was gentle.
4. Re-tasted on a second occasion before publication. Where the two sessions disagree by more than two points on any axis, we taste a third time.
5. Price is never an input. A £30 bottle and a £300 bottle are scored on the same scale.

Where the axis names come from

Whisky's existing flavour systems are almost all proprietary, so the naming had to be done carefully rather than by instinct.

• The structure follows the Scotch Whisky Research Institute's flavour wheel — single-word categories like Peaty, Fruity, Woody, Cereal — which the Scotch Whisky Association reproduces for consumer education in itsTasting Toolkit. That arrangement is the industry's neutral vocabulary. We reproduce neither SWRI's artwork nor its roughly 130-term detailed descriptor list — copyright in both remains theirs, and publication for education is not a licence for commercial reuse.
• What we deliberately avoided. The Flavour Map — Delicate to Smoky vertically, Light to Rich horizontally — is jointly copyright Dave Broom and Diageo Scotland Limited, as are its nine flavour camps. The Scotch Malt Whisky Society's profile names are its own, both the current eight and the twelve they replaced in 2025. The Wishart/Strathclyde dataset has no traceable licence.
• Why every axis here is one plain word. Both proprietary systems name their categories as pairs — "Fragrant & Floral", "Smoky & Peaty", "Toasted Oak & Vanilla". That compound form is their family resemblance, and "Fragrant & Floral" is in fact claimed by both of them. Single words in the SWRI arrangement stay clear of it.

Take it and use it — CC BY 4.0

The whole system is published openly: the vocabulary, the axis definitions, the scoring anchors, the rubric, our scores for every whisky, and the similarity function in full.

That last part is unusual. Every comparable system keeps its matching maths private — one major flavour-discovery site describes its own as "a little magic" and puts it behind a paywall. Ours is thirteen lines of arithmetic and hiding it would only make it harder to check:

const keys = ['smoke','sweetness','fruit','spice','oak','sherry','malt','body'];

function similarity(a, b) {
  let dot = 0, na = 0, nb = 0;
  for (const k of keys) {
    dot += a[k] * b[k];
    na  += a[k] ** 2;
    nb  += b[k] ** 2;
  }
  return dot / (Math.sqrt(na) * Math.sqrt(nb) + 1e-9);
}

It is cosine similarity, which compares the shape of a profile rather than its intensity — so a gentle sherried malt and a big sherried malt read as similar. That is deliberate, because it matches how people actually ask the question. If you want intensity to count, use Euclidean distance instead.

Download the full taxonomy (JSON)

Where this method is weak

Worth stating plainly, since almost nobody publishing flavour scores says any of it:

• These are judgements, not measurements. Two competent tasters will differ by a point or two on most axes. Treat one point as noise and three as meaningful.
• Batch variation is real and we usually score one batch. Non-age-statement and small-batch bottlings vary most.
• Eight axes cannot capture everything. Texture beyond weight, finish length and balance all matter and none has its own axis. That is what the tasting note is for.
• Palate fatigue is unavoidable. We cap a session at six whiskies, which helps but does not eliminate it.
• One score covers the whole dram. Version 2 will score the same eight axes separately at nose, palate and finish. Nobody currently does per-stage scoring on a fixed axis set, and it would make "long finish" something the data shows rather than something we assert.

Frequently asked questions

What is the WhiskeyCask flavour scale?

An eight-axis scale scoring Peaty, Sweet, Fruity, Spicy, Woody, Sherried, Cereal and Oily from 0 to 10. Each axis is independent — a whisky can score high on several — and higher means more pronounced, not better.

Why not use a flavour wheel or a spider chart?

A radar or spider chart distorts area, implies an ordering between adjacent axes that does not exist, and is very hard to read comparatively. Bars encode the same eight numbers accurately and let you compare two whiskies at a glance. The flavour wheel is a vocabulary tool for tasters, not a comparison device.

Is this the same as the Wishart flavour profiles or the Flavour Map?

No. The Flavour Map runs Delicate to Smoky vertically and Light to Rich horizontally, and it is jointly copyright Dave Broom and Diageo Scotland Limited, along with its nine flavour camps. The Wishart/Strathclyde dataset has no licence we could trace and derives from a copyrighted book. We use neither. Our axis names are single common words in the arrangement used by the Scotch Whisky Research Institute's wheel, which the Scotch Whisky Association publishes for consumer education.

Can I use the WhiskeyCask flavour scale?

Yes. The vocabulary, axis definitions, scoring anchors, rubric and the similarity function are published under CC BY 4.0 at /data/flavour-taxonomy.json, including the scores for every whisky we have rated. Attribute it and you can build on it commercially.

How reliable are the scores?

They are editorial judgements on a consistent scale, not laboratory measurements. Two competent tasters will differ by a point or two on most axes. Treat a difference of one point as noise and three or more as meaningful.