Finding Irish whiskey by flavour, not by label

Most people do not start with a flavour wheel. They start with a bottle.

Maybe it is Redbreast 12. Maybe it is Green Spot. Maybe it is something you tried once in a bar and have been trying to find your way back to ever since. You know roughly what you liked: honeyed malt, orchard fruit, soft oak, a little spice, not much smoke. The awkward bit is what comes next.

Irish whiskey is good at giving you names, ages, cask finishes and brands. It is less good at answering the more useful question: if I like this, where should I go next?

That is what the Whiskey Wanderers flavour map is for.

It is a very nerdy, maths-based approach for sourcing your favourite golden liquid going on past experience and association. Using distillery descriptions you can move through Irish whiskey by flavour instead of by label. Pick a bottle, look at its neighbours, read why they sit nearby, and start building a path through the category that follows your palate rather than the shelf order in a shop.

That sounds simple. Getting there is the hard part.

The problem with tasting notes

Whiskey notes are beautifully messy. One producer writes green apple. A retailer says orchard fruit. A reviewer reaches for fresh pear. Someone else says Christmas cake, which is useful to a human and deeply irritating to a system trying to compare bottles.

Those phrases may be circling the same territory, but they are not the same words. If you take them literally, the map gets stupid very quickly. It treats near-neighbours as strangers and lets vague words such as smooth, rich or complex throw their elbows around.

So the first job is not glamour. It is tidying.

The flavour map starts by taking tasting-note language and making it less slippery. It strips filler. It normalises obvious variants. It recognises that cloves and clove belong together, that oloroso is a sherry signal, and that Christmas cake is not a single flavour so much as a bundle of dried fruit, spice, orange peel and sherry richness.

Just as importantly, it tries not to bluff. If a phrase cannot be resolved cleanly, it stays unresolved rather than being forced into a confident-looking mistake. That restraint matters. A map like this is only useful if it admits what it does not know.

What you see on the map

The live flavour map currently plots 782 bottles. The companion Words view is built from 376 recurring descriptor terms across the tasting-note corpus.

Those numbers are not the point by themselves. The point is what they let you do.

On the map, each bottle becomes a flavour profile. Bottles with similar profiles pull towards one another. If two whiskeys share honeyed malt, green fruit and toasted oak, they should feel closer than a whiskey driven by treacle, dried fruit and clove. If a bottle has an unusual smoky or coastal edge, that should not be drowned out just because vanilla and oak are everywhere in Irish whiskey. Still simple?

That last detail is important. Irish whiskey has a common backbone. Vanilla appears 560 times in the current descriptor set. Oak appears 390 times. Honey appears 358 times. If the map simply rewarded the most common words, almost everything would slide into the same sweet, oaky middle and sit there looking pleased with itself.

The better question is not just what a bottle has. It is what makes it distinctive.

That is why the map colours bottles by their most distinctive flavour axis relative to the wider catalogue. Sweetness is common, but common is not the same as defining. Once you ask what makes a bottle stand out, the more revealing edges appear: malty bottles, floral bottles, citrus-led bottles, smoky outliers, heavier sherry pockets, drier oak-driven drams.

That is where the map starts to become useful.

How to use it

Start with a bottle you know.

Search for it, open its profile, then look at the nearby matches. Do not treat them as identical. Treat them as suggestions with reasons attached. One neighbour might share the same honey-and-malt centre. Another might keep the orchard fruit but bring more spice. A third might sit nearby because the oak structure is similar even though the finish feels drier.

That is the sweet spot. The map is not telling you that all neighbours are substitutes. It is giving you a shortlist of bottles worth reading next.

The Words view works from the other direction. You can start with a bottle you love and find nearest neighbours or start with language. If you are drawn to green apple, citrus, honey, malt, sherry, smoke or dried fruit, the Words view shows how those descriptors behave across the catalogue. Big words show the category's familiar ground. Smaller words point towards the quirks.

Together, the two views answer different versions of the same question:

• Bottle first: I like this. What else is nearby?
• Flavour first: I like these notes. Where do they show up?

So at a glance you can learn a lot about whiskeys you have yet to taste.

Why the machinery matters, quietly

There is plenty of maths under the bonnet, but the useful idea is plain enough.

The system counts recurring descriptor words, watches which words appear together, resolves related phrases into controlled families, then turns those resolved descriptors into weighted bottle profiles. Nose, palate and finish are kept separate before being blended, because aroma, taste and aftertaste do not always tell the same story.

The palate carries the most weight by default, but nose and finish still count. That is a practical choice, not a philosophical one. People drink whiskey through all three stages, and a good similarity system should respect that.

The map also gives rarer traits more voice. If every Irish whiskey has some vanilla, vanilla should not be allowed to make everything neighbours with everything else. Smoke, coastal salt, unusual fruit, floral notes or particular cask-driven signals can matter more precisely because they are less common.

This is the bit that makes the map feel less like a pretty chart and more like a discovery tool. It is not just arranging bottles by the loudest words. It is trying to notice the words that actually separate one bottle from another.

What Irish whiskey looks like from here

Seen through the map, Irish whiskey still looks like itself.

The centre of gravity is sweet, woody, fruity and spice-led. Vanilla, honey, oak, malt, orchard fruit, citrus peel, baking spice and dried fruit come up again and again. Smoke is present, but it is not the heart of the category. That fits the plain truth that most Irish whiskey is still unpeated and built around approachability, polish and balance.

But the interesting thing is not the centre. It is the pull away from it.

Some bottles move towards bright citrus and green fruit. Some sink into sherry richness. Some lean dry and malty. Some push smoke or maritime notes harder than Irish whiskey normally does. The map makes those departures easier to spot, because it gives the common ground a shape first.

That is why a flavour map can be more honest than a shelf label. A label can tell you age, brand and cask. The map can show whether the bottle is actually behaving like a honeyed malt, a sherry-rich dram, a citrus-led single malt, a spicy pot still whiskey, or something sitting awkwardly between categories.

Awkward is often where the fun begins.

Where it can still fool you

The map is useful, but it is not an oracle.

It is built from language, and language varies wildly in quality. Some bottles have careful nose, palate and finish notes. Others arrive with thin retailer copy. Some have cask information doing more work than actual sensory description. A map built from notes will always inherit the strengths and weaknesses of those notes.

There is also the usual problem of flattening. Any two-dimensional map loses detail. Two bottles can land close together because they share fruit, oak and spice structure, while still differing in texture, weight, sweetness or finish. Nearby does not mean interchangeable.

And some words are simply weak data. Smooth, balanced, rich and complex can be fair human reactions, but they are poor coordinates. The system tries to route that kind of language away from flavour scoring, but no taxonomy can rescue vague copy entirely.

That is why the map should be read as a guide to exploration, not as a final judgement.

The house view

The flavour map works best when it helps you ask better drinking questions.

Not: what is the best Irish whiskey?

Better: what else lives near the bottle I already like? What changes if I follow the fruit? Where does sherry richness pull away from vanilla sweetness? Which bottles sit at the smoky edge of a mostly unpeated category? Where does the map confirm what I expected, and where does it surprise me?

That is the point of the project. Not to make whiskey look scientific, and certainly not to replace the glass in your hand. The point is to make tasting language more useful.

Irish whiskey is full of familiar words: vanilla, honey, oak, malt, apple, spice. The trick is seeing how those words combine, where they repeat, and where a bottle breaks the pattern.

That is what the map gives you. A way to stop browsing by label alone, and start wandering by taste. So grab a dram, see who lives nearest and see for yourself!