Photo by Jake Olsen for World Coffee Events.

19 JANUARY 2018 | Life moves quickly - as anyone who has been in contact with me via email may have noticed, things have been a little bit hectic and overwhelming after Seoul in November! Since then, I’ve been asked a lot of questions about my routine, often in the form of a long list of questions. In each case, I’ve felt it was best to give short and clear answers, but sometimes these don’t quite communicate the whole story - or all of the important characters who played a role along the way.

The time limit of barista competition forces you to build a narrative that is logical, but full of allusions or references to things that deserve exploration in their own right. I hope I can better accomplish a little of that exploration here, as well as recognizing some of the friends that enabled me to deliver the presentation I envisioned and their vital contributions to what I presented in Seoul.

I want to begin by exploring the idea of the GcMS analysis of coffee - how this fed into, expanded, and went beyond the needs of my fifteen minutes on stage.

. . . . .

One of the goals of a competing barista can be to connect the flavours of the coffee you’re serving with the information we have about its origin: for me, it’s almost the holy grail of a routine. It’s also the hardest thing for a competition barista to do with any honesty - there’s still so much for us to learn.

We often associate attributes like acidity with a variety, an altitude, or a processing style, but the opportunity to taste coffees side by side with variables minimised to test these associations is so rare.

I’ve been lucky to have spent the last few years working as a part of a roastery that’s focussed on carrying different varieties and processes from a number of farms - something only possible because of the long-term partnerships in which Steve has invested over the years.

I chose to work with Ernesto’s Menendez’s Los Brumas lots - a story explored in more detail elsewhere - because of their brilliance on a blind cupping table. The fact that we had multiple lots with small variances between them gave me an opportunity to explore that coffee in a different (and exciting!) way.

. . . . .

As I’ve written elsewhere, the theme of my presentation - delivering a sensory experience based on flavour rather than facts about a coffee - was rooted in the experience of attending an exhibition on modern perfumery held at Somerset House earlier in the year.

This exhibition brought together a number of interesting ideas around how aroma/flavour can be communicated as well as how that information is subsequently interpreted by customers and professionals. As I played with this concept, I began to explore the idea of building a sensory playset based on the flavours within the coffee I was brewing. The question of how we could bridge the gap between my personal experience and the experiences of each judge became a key obstacle to overcome: traditionally, we accept that, whilst variations occur both in the physical characteristic of a brewed product as well as within different individuals’ subjective assessments, a professional’s ‘description’ of a coffee’s flavour is its quality fingerprint.

I wanted to see if there was a way to get one step closer to an objective measurement of the qualities of a given coffee. It was this hope that led me to the idea of trying to analyse the coffee I’d brew in Seoul through gas-chromatograph mass-spectrometry, or GcMS.   

This started as a very vague idea: would it be possible? What might it allow me to do? In all honesty, I expected this to be one of the many ideas we would brainstorm that wouldn’t make it all the way through development. Throughout our time together, Pete was very good at keeping me grounded in terms of what was (and wasn’t!) achievable during the limited time available to prepare. When he didn’t immediately veto the idea (!), we reached out to some friends who might be able to help.

The experimental kitchen at The Fat Duck is one of the most interesting teams with whom I’ve ever had the pleasure of working with for Hasbean. I knew that they had used similar research in building some aspects of their menus - our contact there was incredibly helpful, putting me in touch with Ian Fisk, Associate Professor in Food Chemistry at the University of Nottingham.

. . .

When you review Ian’s list of published research, you can’t help but be reminded of the scale of science conducted in the food industry - as you follow link after link, you’ll come across almost every product you can imagine.

We often talk about science and coffee - I think the visible nature of the chemistry of extraction and some of the equipment we use both at home and behind a bar makes it a very clear connection - but I think I’m almost more amazed at how much work happens in areas where the association between science and product seems much further away.

. . .

Ian kindly put me in touch with one of his PhD Students, Chujiao (Gloria) Liu. Gloria had not only done really interesting work on roast defect identification, collaborating with Morten Munchow amongst others, but was currently in the process of writing up her thesis, in which coffee aroma is playing a huge role.

In email over the following few weeks, Gloria and I refined the idea we wanted to explore and the methodology we would use to analyse the coffee and results. What became evident very early on was that the demands of the competition timetable would mean that this would be an indicative analysis only, allowing us to test an idea and a methodology rather than a truly scientific experiment proving a connection between growing conditions and compounds.

The number of samples we could compare was further limited by time in other ways: the availability of the GcMS equipment, the processing and analysis of each sample, and our own schedules.

. . . . .

The following feels both overly complicated/specific and short of many of the relevant details. Rather than giving a full overview where the experiment could be repeated, I hope to better illustrate the approach we took and the complexity involved in the choice of samples alone.

. . .

We set a specific goal for our work with Gloria: to highlight compounds with direct links to flavour/aroma attributes specific to this coffee when prepared to the two extraction recipes used in the routine.

In order to be able to directly link results from the GcMS with sensory experience of the drinks I would be serving the judges, we chose to conduct headspace analysis on brewed product rather than ground coffee. This necessitated establishing brewing recipes early, locking down as many variables as possible to attempt to replicate the extraction that would occur in Seoul. We also had to accept that the quality of the results would be tied to the shots produced.

We used similarly profiled roasts, roasting times, and rest periods for each coffee: we also acknowledged that roast would be a significant variable and that each coffee could potentially have ‘better’ quality characteristics with a more nuanced roast applied.

With more time, a more rigorous approach would include analysis of multiple shots of each coffee to each recipe, to identify and separate variance in compounds from extraction from those variances occurring directly from variety/altitude/shade, as well as extractions from multiple roasts of each coffee.

. . . 

As with every attempt to develop understanding through the scientific method, you’re left with the realisation that with more technicians, money, and time, you could do something really interesting! You’re also left with an increased understanding of the need for very small precise questions for any experiment to be truly definitive.

. . .

The coffees analysed were:

• Coffee 1 El Salvador - Los Brumas - Washed SL28, wild forest microlot

• Coffee 2 El Salvador - Los Brumas - Washed SL28, standard lot

• Coffee 3 El Salvador - Los Brumas - Washed bourbon, standard lot

We attempted to minimise variances between the lab environment and the competition environment as much as possible:

• Each coffee was brewed to the final competition recipe for analysis: 19.8g dose, 44g yield, 30-32 second brewing time.

• A sample was immediately drawn from the extraction and sealed in a vessel to retain temperature; the remainder of the extraction was then tasted to confirm it matched ‘balanced’ extraction expectations for the coffee (not tasting significantly sour/bitter/over extracted).

• As space was limited in the lab, shots were brewed on a Linea Mini set to mimic the profile we would use on the Black Eagle: 93.5º, 8.8bar and fitted with 20g baskets.

• The machine was filled with water to a recipe that matched the expected WBC water conditions.

• Coffee was ground on a k30 air with new burrs to match the grinder we would be using at the competition.

All of these parameters would potentially affect the results, as would roast and brewing methodology (OCD, pergtamp, etc).

An additional shot was brewed of Coffee 1, to the slightly different recipe we had established for the milk beverage: 19.5g,40g yield at the same grind setting. The differences between analysis of this and the espresso extraction would inform the presentation for the milk beverage.

. . . . .

On lab day - October 9, 2017 - we set up our espresso station and, one by one, dialled in each coffee to the parameters required. Once we were achieving our desired extraction results consistently, we submitted a sample for analysis. Submission needed to happen quickly in order to maximise the volatile aromatics we sought to measure: prepare the sample vessel, brew the shot, seal the vessel, taste the remainder of the sample to confirm extraction parameters had been met... at which point we would run the sample across to Gloria at the GcMS, where she would begin the analysis.

Each GcMS analysis took around 40 minutes, during which the next coffee would be dialled in.

We also spent some time of this time blind cupping samples of coffees that Gloria had prepared as part of her thesis: robustas treated with a specific post-harvest/process treatment in an attempt to remove negative flavour attributes.  It’s a super interesting project with lots of future interesting potential!

. . .

In discussing the methodology used to prepare those samples, particularly around how the coffees had been sourced and roasted, Pete, Gloria, and I saw clear areas where some of the knowledge and technology the specialty industry now uses on a daily basis - the repeatable, measured roast profiles and small samples possible with tools like the Ikawa roaster, extraction testing by refractometer, etc -  could further refine work like Gloria and Ian’s. Think of how much we could all achieve if the specialty movement can find ways to better engage with industry and academia!

. . . . .

After winning the UKBC in late August, we had a limited amount of time to prepare for Seoul. I further complicated this by taking a previously-booked holiday after Tamper Tantrum in San Francisco. Once I returned, Pete and I were working full-steam-ahead on competition but, as this idea about running the samples through GcMS took shape, we realised lots of important aspects of the routine - most notably the finalisation of signature beverage and playset aroma development - would need to wait until after we’d received the GcMS results. With a lab date of October 9th and London “final” run-throughs with UK judges planned on October 20th, we were cutting it pretty fine.

Following the sample analysis, we were able to briefly explore some of the data uncovered through the equipment, ensuring the testing program had worked as expected and there were no large gaps due to mechanical or sample failure, etc. The hefty work work of compiling the data into something we, as outsiders to the field, could interpret and understand, fell on Gloria.

Whilst we’d discussed schedules in trying to get to lab day, we’d never really discussed the expected turn-around time of the data compilation with Gloria. Excitedly, we asked when we might expect to have readable results so that we could plan out the remainder of our preparation time.

When she saw our faces after suggesting it might take a few weeks, she put aside a number of urgent projects to help us move forwards.

. . .

I can’t thank Gloria enough for her role in this journey: it’s a reminder that all competition routines are built off the backs of extended teams beyond those you see and many kindnesses. We could not have created the routine we presented, let alone achieved the final placement, without Gloria's and Ian’s help and support.

. . .

On the 10th October, just a month before we would take the whole routine to Seoul, Gloria delivered the following graphs and summaries, plus the huge body of data detailing compounds and concentrations.

. . . . .

From this point on, beyond some very loose but interesting suggestions we could make from the data about the impact of brewing recipe, variety, and maturation rate on the presence and volume of some compounds, our timelines insisted that we moved the focus from interesting experiments to drawing on the analysis purely for the competition routine.

The data we received was interestingly complex; there was a real risk that each member of my competition team could get lost forever in a black hole of analysis. To ensure we could achieve our goals in a tight timeline, we were forced to separate the many tasks and each apply a little headspace to a different problem.

Pete focused on managing the elements that needed to go into production to enable meaningful, deliverable results on the WBC stage and sig drink methodologies. Whilst I attempted to parse out what questions needed to be asked of the data to present each drink course specifically, Jenn attempted to answer those questions.

Using Gloria's data, Flaments’ coffee chemistry, and a lot of internet searching to sort the results into meaningful lists - first by concentration aligned with perception thresholds, then with links to the flavour descriptors we’d selected for espresso and milk courses through sensory analysis - Jenn was able to leave us with perceivable compounds at their highest concentration in each beverage.

She then shifted her focus to finding a way to meaningfully communicate these results visually on my station tables, whilst Pete and I focused on tying everything together for the competition: the flavours we were experiencing in drinks, the sensory playset we had built to communicate aroma compounds in the espresso, and the methods we would use to create a signature drink rooted in the espresso’s other characteristics.

The performance felt complete. It was now a matter of refining words and actions to deliver it well.

. . .

It’s easy to define or describe the final routine as scientific, bearing in mind the connections we built between this research and other areas we explored. I believe it’s more honest to say that we used a scientific method of analysis to explore this coffee further - a truly scientific approach would ask smaller questions, make smaller statements, and require less drink making and more talking than is suitable for a barista competition.

. . .

It’s my hope that in the future, we - both my team and the greater specialty industry - will find ways and opportunities to use some of the information we gathered for this routine to drill deeper into the conditions are responsible for exceptional flavour attributes as well as just generally using GcMS to better understand and consistently deliver improved taste experiences.

. . . . .

If you've made it all the way down here, thanks for reading!
With more stories to tell and many more people to thank, expect to see more here soon. 

x Dale Harris, 2017 World Barista Champion