#51: Using Your Coffee Data as Business Intelligence to Deliver Quality, Consistency, and Control | Expo 2018 Lectures

Specialty coffee’s artisanal and handcrafted customer face is real and supported by a complex supply chain and highly specialized production. These two things create and collect lots of information. Understanding what can be collected and how it can help coffee businesses is simpler than people think and is also the first step towards gaining real benefits. What’s more, “business intelligence” is already accessible to businesses of all sizes and is easy to use and inexpensive. It helps remove guesswork for beginners and delivers new insights for experts. In today’s lecture, Andreas Idl presents how this can be done with a focus on roastery information.

Andreas Idl a socially minded entrepreneur, is the founder and CEO of Cropster GmbH. As a software developer with a degree in Information Systems he has focussed on using his technological experience in socially positive ways. This focus led him to Cali, Colombia and the Research Center CIAT. His work there focussed on helping small farmers in developing countries through research and development projects, specifically in coffee. As the project came to an end, Cropster was created to continue the work in 2007.

Related Links
Presentation Slides
Table of Contents

0:00 Introduction
2:00 Introduction to the presentation and Andrea Idl’s background
4:40 What is data and why it’s important to delivering a great cup of coffee
10:30 The importance of data collection when roasting
13:30 What a business intelligence solution can look like for a roasting department
22:30 An explanation of the Wisdom Pyramid and how it underpins business intelligence systems
28:10 Applying the Wisdom Pyramid to a coffee roasting department
34:20 Applying the Wisdom Pyramid when buying green coffee
42:20 How a business intelligence system helps make sense of coffee business data

Audience questions

46:50 What is the “rate of rise” in the observable data and who owns this data – Cropster or the roaster?
48:30 Outro

Full Episode Transcript

0:00 Introduction

Heather Ward: Hello everybody, I’m Heather Ward, SCA’s Senior Manager of Content Strategy and you’re listening to the SCA Podcast. Today’s episode is a part of our SCA Lectures series, dedicated to showcasing a curated selection of the extensive live lectures offered at SCA’s Specialty Coffee Expo and World of Coffee events. Check out the show notes for relevant links and a full transcript of today’s lecture.

As we’re taking some time to work through our 2019 lecture recordings from Expo, we thought we’d take this time to share some great content from 2018 that hasn’t been released yet. If you want to see some live lectures in person, you still have time! Visit visitworldofcoffee.org for a full schedule of your lecture series in Berlin this June.

Specialty coffee’s artisanal and handcrafted customer face is supported by a complex supply chain and highly specialized production. These two things create and collect lots of information. Understanding what can be collected and how it can help coffee businesses is simpler than people think and is also the first step towards gaining real benefits. What’s more, “business intelligence” is already accessible to businesses of all sizes and is easy to use and inexpensive. It helps remove guesswork for beginners and delivers new insights for experts. In today’s lecture, Andreas Idl presents how this can be done with a focus on roastery information.

Andreas Idl a socially minded entrepreneur, is the founder and CEO of Cropster GmbH. As a software developer with a degree in Information Systems he has focussed on using his technological experience in socially positive ways. This focus led him to Cali, Colombia and the Research Center CIAT. His work there focussed on helping small farmers in developing countries through research and development projects, specifically in coffee. As the project came to an end, Cropster was created to continue the work in 2007.

Andreas’ presentation is quite visual, so we recommend listening to his presentation while looking at his presentation slides. There’s a link to Andrea’s presentation in the episode notes.


2:00 Introduction to the presentation and Andrea Idl’s background

Andreas Idl: Thanks everybody for being here.  My name is Andreas Idl. I’m one of the co-founders of Cropster. I’m Austrian so my accent is not really accurate. I hope you can hear that, and I hope the message is kind of clear. For me actually, I think it’s the first time I have a lecture at SCA. I know we had actually one before three years ago. It’s the first one I hold alone, and the topic actually got elected. It is about data and it’s about using data as business intelligence in the coffee business. And how do I get to data actually. So, my background is as universities IT so I’m working with data for many, many years or thinking about data and we nowadays use data as a word and as a thing more often and I think that came with computers and the more data we share their might be privacy things or what Facebooks doing with what data set.

So, we talk about data all the time, and we use it wrong all the time and that’s pretty interesting how that came. So, today I will talk about what’s the idea behind data, what data is, where it ends. What information is and knowledge is, how that works. We will talk about was what a business intelligence solution is. Why it is based on data? How we can use that in coffee. What kind of data sets do we have in coffee for quality control and steering a company and besides that maybe theoretical background. We will have a look at for real-time examples to kind of underline how that actually works.  So, one thing more, so after studying IT, I worked in software engineering and through a job offering in research in Colombia.

I moved to Colombia where I met, I have been working with the two other co-founders of Cropster and at that time we had been working on data collection on the farmers side, on the production side of specialty green coffee in Colombia, and that’s actually how I made my way into coffee. So, I’ve never been working as a barista as a roaster before that. I came from the green coffee side and the IT side are 30 side and from there we actually made our way into the industry and thinking about all those processes.


4:40 What is data and why it’s important to delivering a great cup of coffee

Andreas Idl: So, what is actually data? You might be surprised, it’s basically everything. So, if you ask a physicist today that person would actually tell you that data is everything like every atom, everything they describe, all the description of matter and energy and whatever else it’s data and that’s actually true also. So, we could define data in a way that everything we see, everything we taste, hear, sense or any way of perception is actually data. So, a wavelength which we see as a color that’s data actually and so data is actually a very broad term. Data is not a text we can read so if you talk about data sharing in online systems, that’s not data. Below that there is data, but actually we share information and the distinguish between those two things is actually quite critical and important for us that we know the difference.

So, if you think that data is everything we could also say that everything we do in coffee is based on data. A temperature is data. When we roast at that very specific moment in time is data. The color of the coffee, the smell of coffee, a specific flavor. That’s all based on our perception and that’s all data and that means data is the thing that the describes our reality and coffee and how we can actually understand what’s happening in coffee and we all know that coffee is super-duper complex, and I think that’s actually the real interesting thing about our industry. Coffee is chemically so complex we can deliver so many things out of that, out of the product. That there are so many ways that we can continue learning and learning for so many years and we still deliver new products and that’s actually, for me, myself that’s an awesome thing and really interesting.

So, if you think about that all these parts in coffee are actually data sets. We can actually think about how to use that, but we need to consider what is actually important when we produce coffee, when we roast coffee or sell coffee or brew coffee. It’s actually not to have a development time in roasting of whatever percentage or degree. It’s not a specific roast color. It’s not even a flavor. I think actually our goal as an industry and as a company is to deliver great coffee. I think that’s the baseline and great coffee is not a specific thing. It’s different from one location to another. A great espresso might be something different than a great filter brew. You even might have a cupping description of a coffee telling you this is great, but does an 89 points on SCH really tell us this is a great coffee for a specific use? No, it doesn’t. You might need to have a different coffee for your espresso blend with different flavor sets.

So, great coffee is very specific in the context where we use it, but I think it’s the goal of what we need to do and that’s when we buy green coffee, when we roast coffee and brew it and etc. That’s our goal as much as great coffee as possible and guess what that was the slogan of SCA before the merger. So, the slogan at that time, “Because great coffee doesn’t just happen.” And it’s actually pretty accurate. I thought about that phrase and I think it should actually tell great coffee might happen, might just happen but not again and again and again.  So, you might have a lucky shot, but then you didn’t understand what’s going on. Everybody can have a lucky shot, but it doesn’t help us, so our goal is to deliver great coffee again and again, and again. That’s a part of consistency and that actually implies we need to understand what we are doing otherwise; we don’t get there and SCA when they started out, they attacked that problem of not knowing, of not understanding how to deliver the coffee great again and again and again.

So, what they did they created standards, best practices how to handle coffee. They introduced cupping forms, they introduced green grading forms, roasting protocols and a lot of exchange of information, but basically establishing standards and if you look at those standards what they developed like the cupping form, it’s actually a data collection form. If you think about what it is SCA targeted the thing to collect important coffee that affect the quality of the coffee and that helps us to understand what’s going on and these were early attempts and of course if you capture things on paper, it’s easy to do. It’s very flexible. You can draw anything on a paper, but the flexibility has its downsides. It doesn’t really scale on paper. You don’t get much in return. So, while the SCA then established all these new standards and helps to understand our coffee production it has its limits in outcome. So, in that regards, it wasn’t a real business intelligence solution. It was a data capturing solution.

But in the end, SCAs goal and our goal was to produce great coffee and they developed standards to help us guide us there and I think that’s where it basically ended.


10:30 The importance of data collection when roasting

Andreas Idl: So, this is basically down the chart to that. So, assuming we have green coffee on this side. There is quite a black box in between and we have brown coffee in the end that hopefully is great. If you think about a roastery, I think the term of a black box is quite correct. Probably not the complete black box as we’ve drawn here. Let’s assume it’s a black box with a few windows in it. So, somebody cut the window in here because we know a few things about roasting, we know there is a thing like a first crack. We know about defects in roasting like baking. We know there is a thing like Maillard reaction which we can’t actually not measure at all at the moment. So, we actually know it’s there, but we can’t see it and many other chemical processes in the roast we can’t see at all. But basically, we can see temperatures. We might even measure a roast color. So, basically we have the tools to glimpse a few things that happen actually in the production. We don’t see everything.

So, the point is we have a few windows. We can see a few things, each add some light into the production, but we can’t see everything and even worse to that, that we don’t see everything. We have connected parameters so for example, you might do your super-duper best development of coffee, development phase after the first crack. You did everything right, the color is there, the time is there, the temperatures are there. You don’t bake it. It doesn’t help you at all if you bake the coffee before a Maillard. So, there are a lot of connected parameters here. So even if you do everything right except one thing you can’t save the coffee so you don’t get a great coffee in the end.

I mean what’s very obvious. Let’s assume you buy infected moldy green coffee. You can do whatever you want. You can be the best roaster in the world in the best barista, doesn’t help. So, if you think about the roasting depends on the good green coffee, the brewing between depends on good roasting and all these processes in between depend on each other and that makes it even a bit more tricky because we need to understand these correlations and we have two options to target that problem and we have a very complex problem here. Either we remove the black box or cut as many windows as needed to see everything and that’s maybe not possible or we target that with a business intelligence solution that basically helps us to connect or collect the relevant pieces of data and we integrate these pieces of data into information that helps us to understand what’s going on and that’s basically what then this talk comes down to. How can we do that and what does that mean to our business?


13:30 What a business intelligence solution can look like for a roasting department

Andreas Idl: So, a business intelligence solution starts at the point where we need to define what data we need to collect in our roastery for example, in our operation. It needs to help us to collect the data, sum it up to some useful reports so we can understand what’s going on. Very simple concept actually, not so easy implementation.  So, this a very, very simplified view on a roastery. I left out stuff like post roast blending or further processing like grinding or packaging. No brewing is in here. It’s the core process basically before that and so what we do in the roastery, we have our green coffee inventory here. So, that’s the green coffee we have in our roast operation. Before that, we have coffee at our Importer on the ship at an external storage site and we move waiting between these two locations.

We have a roast machine or many roast machines in one or many locations. Roasting single origins or singles or roasting blends. Taking weight out of the inventory. We have a production planning going on and that’s in between what orders are coming in, what we have in inventory, what capacity we have on the roast machine, so it affects all of these things. Additionally, we have roast the coffee. That can be assessed again with cupping or color measurements, so we have a lot of quality control going on here. We have production stats and production management here to see how production is going, what volumes we are doing, planning further ahead.

We have quality analysis focusing on the roast process and on the cup quality which again affects what we do in the roast machine because we might change profiles over time depending on what we figure out. The black arrows here are basically processes and you see it’s going forth and back and in circles and if you look where relevant information is coming from, it’s coming from several places. On the green coffee side there are things coming in like variety or moisture, water activity, defects etc. The age of the coffee or the screen size.  When we roast obviously we see times and temperatures. During a roast we collect roast curves, rate of rise curves, gas pressures, drum pressures, rotations etc. whatever I want to log. Also. I might already see here a few rows defects. We then have the roasts which then go to a cupping. We measure the color, we do production cupping, we might even measure the moisture. We measure and weigh it.

So, other sets of information happening at that point and then in that itself, we actually get new sets of data like deviations between roasts, defects of roasts, correlation between roasts. So, if you look at that, data happens to be coming to life or need to be collected at very specific points in production. So, that means I have no single point of collecting the data. It also means that most likely different people will need to collect it. It also means it’s coming from different devices and that’s turning out to be more complicated because I need to make sure I have all those things because again, I want to see interconnected things between processing steps. If I don’t capture the former point, if I don’t measure my rate of rise in Maillard it doesn’t matter later on I can’t see if I did it right or wrong and so it has no effect at development time. But this is a bit of a mess, so I tried to clean it up and oversimplify which is again green coffee, roasting brown coffee, roasting coffee.

So, looking at what we can capture in every step. From origin at the green coffee we can get information pieces about altitude and location where the coffee is coming from. The processing. Is it a washed coffee? It is a natural coffee etc. What’s reactivity, density? It has an effect on how the coffee is being turned over in the drum. Screen size again, the same thing, a defect. Cupping flavors already when I buy coffee, I cup it. That should give me quite an orientation and also storage time. How old is that coffee? That might tell me also how fast I need to roast it, or I might expect aging effects really sooner than later. If you think about roasting we talk about weight, blending ratios, temperature curves, rate of rise, gas rotation, pressures, times, development times, overall times. Modulation timers, for example, like proposed by Repose. Also, the time of the day. Did I roast in the morning? Was it the first batch in the morning? Was it the10th batch before the evening? Is it the last roast? How was the weather that day, what seasonality was it foggy or not?

So, basically at what day was it? That’s all relevant things to understand what this profile is about and then the I have roast the coffee. We measure color moistures, the weight loss cupping scores again, brewing parameters for the cupping and these kind of results. So, let’s say if you focus on these three steps, that’s basically what we need to collect and that’s a lot do. If you think about, you capture all that by hand. That’s a lot of effort and I understand if actually nobody actually is doing that because that’s probably not worth it. But still, all those parameters are thought to have an effect on the quality and I think we can agree on that more or less that these are kind of important parameters. That’s why we see these parameters when we buy green coffee. We capture these parameters when we roast coffee and when we cup coffee. But we have a dilemma here.

So, first we want to capture all those things to control our quality, but we don’t have the time to do that in an efficient way. But we also cannot afford not doing it because then we have a high risk of failing. We have a high risk of a problem in the final product and not achieving a great coffee and that’s basically in between, so a dilemma basically means. I think the original definition was a bull. Okay. So, you’re standing here, and the bull is running to you and it has two horns and you can basically choose if you want to if you’re hit by the left or the right horn but that’s the only option you have and that’s a bit of what the dilemma is. So, you spend all your time in collecting the data or you have a bit of risk in what you do and then again, you need to have a look.

So, we need to capture basically data as efficiently as possible and we want to automate as much as possible in terms of what we can collect. So, temperatures during the roast obviously we can collect automatically. A cupping not so much. We taste it and it takes the time to cup but then comes actually the point how we make that more efficient and there is a limit. We cannot speed up a cupping so the only thing we can think about if you capture all that stuff, what’s the best benefit we get out? What our return of investing the time to capture that data and this is basically what the business intelligence system introduces. So, it helps us to capture the data in one location. To capture it once and as automatic as possible.

On the other hand, it will use the data to give us various example of feedback that help us in various aspects and I will show you for example, actually why collecting cupping data helps us later on.


22:30 An explanation of the Wisdom Pyramid and how it underpins business intelligence systems

Andreas Idl: But this is something we need to keep in mind. If you invest the time to capture data, we need many, many benefits and many different kinds of output to create and make up for that investment to give us a return on investment and there is a concept too then. This is the philosophical slide in this in the show. It will end after that. So, this concept is very old and it’s something my English teacher introduced when I studied at University. He said all you IT guys, you need to think about that, and that structure follows me all those years actually. It’s called the Wisdom Pyramid and you will see that IT actually in fact changes that pyramid a bit nowadays.

So, at the very base level we have data and as we remember data is everything so there is nothing that’s not data. It’s basically an objective fact, I would say but it gives very little meaning. So, just imagine you roast the coffee and at minute six you have X Fahrenheit. That’s not a lot of information actually. There is not much you can understand except in that moment, but not overall. So, data is kind of here, a roast color again for example, not much telling us unless we integrate it.

So, the next level is information. The idea behind this I take all the data, I sum it up and I correlate data. I connect data, I structure data and suddenly it becomes understandable. For example, I have my roast color and the roast curve and suddenly I have a meaning out of that. Alone, not so much. If I add cupping to that I have even better information set so I can better understand what’s going on. So, basically information makes all the data accessible to us. That’s what it is. If you think about, out of coffee example. So, a red light bulb is not much, that’s data. But, if it’s on the street, we actually know to stop the car. So, the context give it a lot of information and above information is knowledge.

That’s the point when we start understanding what that means. An example in coffee was that roasters started using the rate of rise curve and they cup their coffees and they had high frustration. Before having rate of rise, this coffee tastes flat. This didn’t develop the acidity. There is nobody in this coffee and still the bean temperature looked really, really similar and they didn’t see a difference and suddenly we had that information of rate of rise available and actually they’ve seen a lot of difference and they figured out if rate of rise is doing this, this is not acceptable to my product quality. This is not great coffee for me and so, basically they started understanding, seeing a correlation between various information sets over time. So, they understood what the information means and here we can actually become creative.

On top of that, the final piece is wisdom. That sounds a bit yeah, wise like a man with a white, long beard very old man, and I think the reason to that is because it takes a lot of time to achieve that. To get a lot of experience to be a very experienced person in what I do and it’s probably why we associate wisdom a lot with old people but basically, this wisdom says that I gathered enough knowledge to actually start to be creative. I start to implement the strategy, to steer my process, to steer my business, to change ways. I have enough understanding to do that and we all know that when we learn stuff first, everything is kind of untouchable then we start to understand a bit especially if you learn another language.

Suddenly, we understand really what’s going on. We can have conversations and stuff. We understand what’s being meant after being able to read it and then we can actually use the language like ideally as a mother tongue level, then we have reached the wisdom level and the same in coffee. So, with that rate of rise example before what roasteries did over to time. First, they understood this rate of rise is not acceptable to my quality standards, but next they started to redesign their profiles, so they never run into those rate of rise forms of having a too low rate of rise or too high, for example. So, their assessment on how to design a roast profile dramatically changed by that and what IT here basically does.

IT is really good at collecting data. We are not so good in that. It’s very error-prone if you write stuff down. It takes a lot of time. We are not good at repetitive things and we are rather slow compared to a computer and computers are also good in collecting and combining that inflow data into information so we can actually read it. So, computers basically start to occupy these two levels and that’s actually what then again a business intelligence system is in our company and we are here. We gathered our knowledge out of that and the wisdom then to use that information to faster change, to improve processes, to understand problems, to increase our production.

So, please keep that pyramid in mind. I will now start to look at a few examples in coffee that happened and that followed that pyramid pretty well.


28:10 Applying the Wisdom Pyramid to a coffee roasting department

Andreas Idl: So, let’s look at a few examples. So, again, we build up the pyramid here. So, assuming we are a roastery and we roast coffee and we cup those coffees. So, the data sets we actually collect here are temperatures as curves and cupping scores, flavors whatever.  On the information level we started to correlate those things that it can actually be seen in the same field. So, I see my roast curves, I see my cupping results in one thing, and I can see that over time again. So, I look at that stuff regularly and suddenly I understand that again the rate of rise example, if the rate of rise is too low companies suddenly understood that this is a problem. We always get baked flavors if rate of rise is doing that and we didn’t clearly see it in the bean temperature before that. So, hey actually say, okay this is probably an issue. This is a visualization of the problem. That’s a new set of information.

They start to understand. So, they understand. So, they started to understand baking in their company, with their machine in their context. We don’t have an objective standard when baking starts I think. Most couples also disagree, and it also depends what your roast machine is, when it really starts. How we see that actually, it depends on the used sensors but basically in their company, companies understood, okay baking this is how it happens. This is our minimum rate of rise we need to have. These pumps we don’t have, or this pump is still okay. So, they introduced new standards and new ways to think about the coffees and what they basically did they started to redesign the profiles. At the moment when Scott Rao wrote his roasting companion book.

He already talked about a design of rate of rise that was very specific actually. So, at that time in the industry and here’s a person who already knew that if we do that, we don’t hit the baking phase. So, that was already implemented and most people in the industry started to think about that, but we reach that with the rate of rise curve and the process of looking at rate of rise curve and cupping over time. This is an example of how that looks so, this is the bean temperature and the rate of rise and there are some great curves in the back. So, we compare that current roast with a lot of others, and they all end here. I know it’s pretty light, I hope you can see it.

But these green coffee roasts I mean that really touches the zero line a lot. Sorry, this is Celsius. But it’s zero rate of rise for some time. We also see it in the bean temperature here. It’s so obvious and here we get the cupping results. We get flavors and cuppings curves and basically here with the green one we see baked and if we get that presented over time, we will start to see a correlation and say, “Hey, this is too low. We have a problem. This shouldn’t happen.” And then we are actually at the wise level and change that. But that’s a very obvious example, but we have slight changes in how things affect quality again because some processes are interrelated. They are not so obvious or there are processes we cannot see.  Again, a pyramid.

So, we collect the same data actually. So, we collect the cuppings curves, the roasts and the time when that was and now we aggregate it differently or the system does that for us. We aggregate it by time at the profile and suddenly we will see if you do that. If you have a cupping curve of each roast, a profile over time. We actually see how the quality develops over time and there’s some interesting facts to that. So, first I can see how consistent my quality actually is. Is it going up or down. Does it go up or down between different roast machines or locations or does it depend on our whatever fog is coming in on the weekend?

But what many people figured out, they’ve certainly seen that a  specific green coffee ages too fast and it’s just a bland component and the quality slightly decreases over time and you don’t see that on a single cupping score, but you see it if you wrote it up in time and it might on the long-term improve their storage because they have a storage problem or they simply need to swap green coffee for that specific profile. But people can take action here and they can do it because they understand the quality development over time. More concrete that is as simple as this so each bar here is a cupping score of a specific roast and this line here is the cupping score of our target quality of our target roast quality and you see it’s going a bit up and down and while you see here that there is actually no bad cupping result here it’s stellar. 87 is the worst score. We can actually see that the quality on average is going down a bit.

So, obviously it’s easy to see that on auto track manually on 10 coffees or so or 12, what it is, but it’s rough to see if you roast for many months and a lot of the same coffee. You get so many batches, but you will still see a tendency and you see if you’re on track, improving or declining quality.


34:20 Applying the Wisdom Pyramid when buying green coffee

Andreas Idl: But interestingly, we use the same data and we aggregate it a bit differently and we get another output and we can do the same for coffee buying purposes for example. Again, we have cupping, roast and time. We aggregate the usage on the volumes per profile per time. So, a bit different than before so we get new meaning again. What you didn’t get out, we can suddenly see a run rate. So, if you have that usage data, I can project it to the future and say hey this green coffee is running out in three weeks. I need to definitely restock from my trader or whatever else or I need to buy new green coffee.  I can also see the usage over time and see what did I really use between mid-November until end of December last year.

So, when I plan my coffee purchases for this year. What did I really use and it’s sometimes not so obvious because you buy for example a Columbia for blends or being used in two blends and you buy from the same country, you buy a single origin or two single origins. So, you want to group that out and see at what quality level, how much coffee did I really use in that specific time frame and that helps to buy the coffee in the end which again also affects the quality because if you buy too much coffee it will age or you need to use it for something else and if you don’t have too little coffee, that’s an operational problem additionally. But basically, again same data set, a different outcome, and different strategies I can actually take on and I hope the point gets a little clearer. So, as long as we have the data stored in the same system in the same business intelligence solution I can use it for different purposes.

This again how they can look. So, we see different green coffees here being used from Guatemala. The amount of green I’ve used in a timeframe, comparing that to the same period before so growth or decline. Roasts per week, batches per week and what I currently have left if any so this is obviously empty but assuming you go to the show flow afterwards you cup [34:40 inaudible] coffees. That’s nice information to have in case you really like a coffee and you get a price and you have a quality assessment you can you can decide on that. And here comes a bit a new example. So, all these three examples are kind of in the past and show that with the same data we get a lot of information and strategies out. This is a very new one that’s an experiment by Chris Korman. He’s working at the Crown and he published that, I thinking Daily Coffee News two months ago.

So, Chris is basically logging all his roasting, all his cuppings. He’s also logging moisture and water activity. So, he’s a bit in a good decision position actually because he has a water activity device which most don’t, and water activity is a really interesting parameter actually because he tells us a few things for example water activity is a strong indicator for storage and it’s not being used only in coffee. Use it, for any grain of everything you dry actually in our food industry measures water activity because it tells you first if there’s a lot of free water. What it basically is, what the parameter is, it’s very it’s more likely that you get an infection on that product so microbes or germs or whatever else is likely to go in and interact and use the water.

So, a too high water activity that thing might get moldy although the overall moisture is still in an acceptable level. It also tells us a bit about shelf life. A too high water activity or too low decreases the shelf life. for the same reason. If there’s too much free water, it more likely goes in and out basically and moves around that’s not good and it has also the effect on green coffee for example that it tastes aged or whatever. Too early basically, but it seems also to have an effect on roasting and then here it becomes really interesting. So, what Chris did he figured out depending on water activity, roasts work out differently or have different dynamics in them around Maillard phase and they darken differently so the color change is more dramatic, and he wanted to do a small experiment and how he did that. He measured the water activity of coffees. He took a sample. So, he took a sample, he split it up in two bags and one bag he stored well and one he didn’t store well. He basically put it out of the window and that changed the water activity dramatically.

He did that for different kinds of coffees. and then he roasted those coffees. First the one with the high water activity and the other one with the low water activity and he roasted them on I think a Probatino and it was with the same settings and he tried to preheat the machine well, so they are really under the same conditions. So, not really production conditions. I would say. So, the orange line is the high water activity and the blue is the low water activity and then he roasted the same thing. So, we see a similar start temperature and similar ending. But we actually see here that the high water activity goes lower into the turning point, but then it decreases much faster although we don’t apply more gas to the system. So, the rate of rise is much higher at that point and it was even more strong. This is not what we should do at home actually.

So, we see the higher activity in orange and the lower in blue again and here Chris opened the airflow to 100% opened. So, that means a lot of water was going out of the system or a lot of heat capacity and you see that lower water activity actually dropped even a negative temperature growth. So, it really dropped while the high water activity at least leveled out. So, it’s not an ideal roast, but the effect is much, much stronger than we see here and he did that with several coffees and what he basically concluded as an idea is that water activity actually tells us how we should treat the coffee in Maillard when we have a lot of water in the system. Is that now an outcome like we know that rate of rise of X means baking. No, we are not there but what Chris did he has now an idea and with his business intelligence solution basically collecting that data he can report now on every roast he is doing to see if he has that correlation over time. It’s like seeing the rate of rise being too low over months and he’s collecting the cupping as well.

So, he will build up a database that tells him if that idea is actually true or not. In case it’s true, he generated new knowledge and secondly, we will figure out new strategies how to attack that problem basically. How to make that not happen and it might again change our procedure and our thinking how to design those profiles. We don’t know yet, but he has the tool to do that.


42:20 How a business intelligence system helps make sense of coffee business data

Andreas Idl: So, we come back to the simplified process. The business intelligence system sits on top and what it does it helps us to collect all that information as efficiently as possible, as good as possible. It stores it in one location that also was the advantage of Chris or at least advantage of Chris right now. He has all his roasts from the last year’s and all the water activities so he can immediately report on the old data sets as well and get information out if his idea is actually true or wrong in his set up.

So, business diligence grabs all that, stores it in one location and then we start to get out our benefit. We can learn about baking patterns. We can learn about defects in roasting. We can use the same data to monitor our quality over time and see lighter or our long-term problems. We can use it even for coffee buying. We can use it to test new theories. So, there’s many, many outputs on that and it’s so similar when I did that lecture actually I was thinking, for me, it’s very obvious that these outcomes come out. I see data and information as multiplication out of that. But, it’s so similar to the point on return on investment. I invest this amount of resources in time and devices. That’s what I do to capture the data and then it really comes up. What do I get back? And that justifies to do cuppings, that justifies to record other things and invest all those resources, but the crucial point is if you write that down on ae paper, if you write that down in an Excel sheet, which is already a bit better.

Those formats predefine what you get out, so you have no multiplication. So, you have no good return on investment and business intelligence system actually needs to do that for you. So that’s the important piece. That needs to help us to get all those outputs and also to store the data because maybe there is another Chris or Chris has a second idea. So, we come up with new theories. They might prove right. One of those will prove right at the point, but Chris can use all his history data set to validate that and to figure out so the point when he builds up knowledge and the process of building up knowledge is much, much faster to him because he doesn’t start with data and information gathering again. He sits on top of that. He can start to process more knowledge more quickly and that’s what these kinds of tools do. We do the same thing in accounting systems or ERP systems in bigger companies. It’s the same process. These systems cost a lot of time to maintain and put in the data, but we have specific benefits from that that justify the investment and for coffee if you think about, it’s great coffee clearly.

That’s our goal. It also helps us to steer our quality and our processing and everything regarding that BI system for coffee must have quality as a core part instead of only efficiency and I think that at the moment when we still don’t understand any process happening or all of the process happening sorry this is the approach we can actually do nowadays with IT systems we have. Capture a few points, correlate as much as possible to circumvent the fact that we cannot see a perceived Maillard reaction really, we don’t know what’s happening except theoretically but not in the moment and that helps us around.

So, maybe what reactivity or modulation charts help us to steer better through that phase and give us a tool to do that and that means I come back to my initial point. That means I also come more or less to the end of the lecture and that’s then obviously the last information and as a reminder that’s on Wisdom Pyramid Level 2 because we can all see that, and we can all understand that. That’s basically just an example for the end but thank you for coming. Thank you for your time to be here. I would love to get a few questions if there are any.


46:50 What is the “rate of rise” in the observable data and who owns this data – Cropster or the roaster?

Heather Ward: A member of the audience is asking ‘What is the Rate of Rise in the observable data and who owns this data – Cropster or the roaster?’

Andreas Idl: So, the rate of rise is an algorithm applied on a temperature curve. So, here we apply it on the bean temperature.

What’s really important here to understand Cropster doesn’t own data. So, our privacy agreement with any customer or user of the system is that this is data of the company using that. So, every company has its private account and aggregates the data into that. What we basically share is knowledge or understanding. So, today we know that rate of rise or too low rate of rise indicates baking, but it doesn’t necessarily tell us on your specific machine, on your specific sensor setup which your machine ships with at what exact number that is. It really depends also on that. So, the best thing is that we have now the information at the knowledge and industry to talk about that problem and you need to know that problem and you will need now to decide on the strategy how to improve my roast from this not happening.

So, we want to enable the user to do the analytics and we have a query backbone basically helping you to do that and the graphical output to that, also some automatic checks, but you can set that in whatever area you want to do.

Thanks again guys. Enjoy the show.

48:30 Outro

Heather Ward: That was Andreas Idl of Cropster at Expo in 2018. Remember to check our show notes for a full transcript of this lecture and visit worldofcoffee.org for tickets to our next run of lectures!

This has been an episode of the SCA Podcast. Thank you for joining us!

Subscribe to the #SCApodcast on iTunes, Stitcher, Soundcloud, Pocket Casts, or RadioPublic.