Coffee is an immensely complex beverage, aromatically speaking - not only do we have the genetic precursors combining with the transformative step of fermentation, but it all gets transformed again by roasting. And it’s then transformed again by the final brewing, interacting with water chemistry, grinder & technique.
With our last set of releases, we had some feedback that we should probably try and give a bit more context for the way we frame our writing about coffee flavour.
We’ve always held the philosophy that everyone’s tasting experience is valid - that we all have different sense memories and associative recollections for any given aroma & taste.
We’ve always been very careful to write the words “we’re tasting” “we find” - etc, when talking about flavour - as opposed to the definitive “this coffee tastes like x, y, z” - and we put “tasting guide” on the bags as opposed to tasting notes, in the same vein as category of things we hope you might find. All of this to ensure we’re respecting the validity of your experience when you try these coffees - and if you find something different, that’s brilliant and we are so here for it - we love to hear feedback.
If we’ve ever come across a little… ‘over-enthused’ in our writeups, it’s come from a place of genuine passion & love for the coffees we are sharing - and a never-ending journey to try new fruits & foods, smell new flowers, etc - all to keep expanding our sense memory and more accurately describe our own subjective experience when we’re drinking these coffees.
Grant us permission to go nerdy in the next section…
How do we actually generate flavour descriptors?
When we smell and taste a coffee, we are generally working our way through a pathway of associative descriptors - starting with the most basic through to the most discrete. This is best visualised with the SCA flavour wheel - where you start from the centre and work outwards. We train anyone who is trying coffee cupping for the first time to start with the most basic premise - and to focus on that one thing alone on their first table - which is to say, “Do I Like This? Yes or No?” - because every other element follows from this basic affective judgement.
Fig 1: A Slice of the SCA Flavour Wheel
Fig 2: Mental model of tasting a coffee
A single cup of coffee is an incredibly dense packet of information. Containing somewhere in the region of 800 to 1000 distinct volatile aromatic compounds, which is more than wine, more than chocolate, more than almost anything else humans routinely consume. That's the effective layer, the actual chemistry arriving at the nose and tongue. But each of those compounds is also a trigger, capable of activating associative memory, emotional state, bodily response, cultural reference, all of which compounds (in both senses) with every other compound firing simultaneously
That cup is arriving with more information than the reader's conscious attention can process, the writer's vocabulary can name, or the language itself can carry. Every act of describing coffee is therefore an act of radical compression, and as we move through the rings of the SCA taster’s wheel, we’re narrowing the possibility space for the information we’re seeking to convey.
As we arrive at a final descriptor, it’s acting as a vector in a higher dimensional flavour space - steering the way towards what the writer hopes would be a hopefully shared destination. But everyone has their own reference frame and crucially, the difference between sense and reference means that your own personal references and memory will mean the pathway you take to construct and crystallise your own understanding of the experience will always be utterly unique to yourself. We can arrive at the same shared semantic anchor through entirely different conceptual routes.
The counterfactual version of this might be sharing descriptors that are entirely based around personal subjective experiences - those that are evoked alongside the common references. Describing a coffee as tasting like “the colour of the sky the first time we remember going to the park as a child” - might carry deep semantic meaning and richness for the writer but carry absolutely no shared reference frame in the same way as “a ripe strawberry” does to a general audience.
So when we seek to compress that higher dimensional flavour space down to something that has shared common references - we can look to compress further and further, till we reach the brute primitives that have the most likely chance of being seen as familiar, and evoking familiar conceptual directions for those that they are aimed at.
And this was what we originally wanted to do, when we set out to write descriptors for Scenery. But we hit an issue:
The Set of All Possible Grapefruits.
Fig 3: What follows from that description
When we compress down the information space - reducing the scope of possibilities to a final form, that coarse descriptor that has the widest shared reference point yet still narrows the infinite to the conceptually transmittable, we’re left with a problem. If we say a coffee tastes of grapefruit - that still contains the set of all possible grapefruits. Everything the word could legitimately refer to, every grapefruit experience that has ever happened or could ever happen, held together by the fact that we agree to call them all by the same name.
The bitter wedge, the heavenly one you once had on holiday from a country that grows them, the pink juice from a carton, the segmented half with a serrated silver spoon, the smell of the peel oil when someone scored the rind. All of it is grapefruit, and none of it is THE grapefruit. When you read "grapefruit" on a tasting guide, your mind doesn't retrieve a single grapefruit. It retrieves something more like a cloud, every grapefruit you've ever encountered, weighted by recency, vividness, and emotional charge. When we compress and transmit a flavour descriptor, we’re running up against the problem of which of our own subjective grapefruit experiences we are trying to evoke.
By definition, the set of all possible grapefruits must also contain Bad Grapefruit Experiences. If we are trying to convey a positive message about a coffee - because we do not want to have the read include “Bitter, Unsweet, Astringent, Unpleasant, Sad” - we will find ourselves reaching for one of two levers:
- Compound descriptors
- Specific descriptors
In compound descriptors, we’re adding a modifier that shifts that reference - eg “Grapefruit brûlée” - a grapefruit that’s had sugar blowtorched on the surface - now there’s warmth, caramels, a lessening of acidity, maybe a touch of smoke if it was taken too far, and a lessening of the space for acidity, bitterness, freshness. The set of all possible grapefruits has collapsed into the set of all possible grapefruit brûlées - a much smaller reference space.
In specific descriptors - we might say “Ruby Grapefruit” - but the secret is that a specific descriptor can be read as another form of compound descriptor - Ruby grapefruit are sweeter, less bitter and searingly acidic than the average grapefruit, so the set of possible ruby grapefruits is smaller and generally more likely to convey the semantic meaning that the writer intended to convey - but inside, it’s just the basic building blocks of “GRAPEFRUIT + SWEET”
Fig 4: Using a compound or specific descriptor to constrain the reference.
We roast a lot of single origin coffees - over 110 different lots in 2025, on track for perhaps 150+ in 2026. When you have multiple coffees that evoke the flavour of Grapefruit - but they do not taste identical to each other, in fact all have unique differences from the subtle to the most striking - you find that compound and specific descriptors are necessary to elucidate and communicate what those differences are, in your experience.
Communicating intent - you do have to market the coffees.
When we originally set out with the intention of putting the most bare bones, strikingly simple flavour descriptors - with the goal of finding the most reductive form of approachability we could, we found that we were underselling the value of the coffees we were trying to put forward.
We did not have the cachet as a roaster to pull off that approach when we launched, we weren’t communicating how much we loved the coffees, and we weren’t ascribing the value and work that the producers’ lots deserved.
The problem you run into is that if all you use is simple descriptors, then every Ethiopian washed coffee is “Citrus, Stonefruit, Floral” - and yet despite their similarities they all clearly taste different. And that level of compression destroys the signal to the point where the difference between a low grade G2 washed and an exceptional G1 is clearly apparent in the cup but is completely absent on the front of the bag.
Choosing to allow specificity back in on the tasting guide was a necessary decision - both to handle the cadence and sheer number of coffees - but also to provide that crucial distinction between them, We still try to include one “category” descriptor, and to shy away from any truly niche or specific calls unless we can really, truly stand behind them as something we hope can be found. And even then, the compound & specific descriptors are just ways to pack more information density into a short number of words.
A clear example of this progression can be seen below, in 3 crop years of El Encanto Gesha:
Fig 5: 23/24 Crop, first time featuring.
Fig 6: 24/25 Crop, second time featuring.
Fig 7: 25/26 Crop, third time featuring.
There were qualitative differences between the three seasons of Encanto we’re highlighting above - the third, most recent feature was emphatically one of the best we’ve bought from Daisy & Fredy yet. And having spent 3 years featuring their coffee, we’re also better at roasting it - knowing how to get the best possible rendition of it, and with an understanding of what the baseline is. The first season we undersold the quality - it’s too generic, but it’s understandable. The second, it’s a good level of specificity - but actually missed the sweet melon note which we’ve come to love with Encanto. The third, we likely went too far with the calls - but we had to communicate that:
- The classic Gesha Bergamot/Earl Grey is there - it was very obvious/loud
- The florals are more complex
- The Melon note is more prominent but it’s very much mingled with the sweet florals
- It’s overall better than last year
We narrowed the set of all possible descriptors down to specifics to try and communicate the above. The Charentais Melon (a very sweet & incredibly floral cantaloupe) call is accurate - we highly recommend trying one, had you had one and a cup of this coffee side by side you’d see what we meant. But it can also be seen as communicating the depth of sweetness and aromatic complexity whilst also reducing the set of all possible melons to a much narrower band. Simply saying “Melon” or even “Cantaloupe” would have left too many possible versions that didn’t communicate this coffee correctly.
Why every effort to describe a coffee is ultimately flawed:
What we're trying to transmit, ultimately, is qualia, the subjective character of experience itself, the redness of red, the bitterness of bitter, what it is like to taste this particular coffee on this particular morning. And qualia are famously the one thing language cannot actually carry. You cannot transmit a taste any more than you can transmit the experience of seeing a colour to someone who has never seen it. We had our experience, you’ll have yours drinking this coffee - and what we have between us is language. It’s not a pipeline for experience, but it can be a set of instructions that might help you construct your own understanding.
There's a version of coffee writing that operates like the emperor's new clothes, the descriptors are presented as definitive, the reader is positioned as the one whose palate is being tested, and the social contract makes it uncomfortable to say "I didn't find any of that." The reader who can't “verify the bergamot” (or might never have tried bergamot, or donut peach, or smelled mexican orange blossoms) has two options that both feel bad, doubting the writer or doubting themselves, and most people quietly pick the second. That's the moment a tasting note tips over into performance, not when it gets specific, but when it gets prescriptive - and that prescription comes from a performative appeal to authority.
This is why we write the way we do: "We're tasting" and "we find" aren't hedges or false modesty, they're the only honest way for us to communicate once you accept that the experience of drinking a coffee can only ever happen inside the person drinking it. If our descriptors land for you, brilliant, we've found common ground. If they don't, you haven't failed a test, you've had your own encounter with the coffee, and that's the whole thing.
The emperor isn't naked, but he's also not wearing what the tailors said he was, he's wearing whatever you actually see when you look.