Welcome to "AigoraCast", conversations with industry experts on how new technologies are transforming sensory and consumer science!
AigoraCast is available on Apple Podcasts, Stitcher, Google Podcasts, Spotify, PodCast Republic, Pandora, and Amazon Music. Remember to subscribe, and please leave a positive review if you like what you hear!
Dr. Thierry Worch, a sensometrician at FrieslandCampina, started life in sensory as a project manager at OP&P Product Research in Utrecht, The Netherlands, before completing his PhD in in collaboration with Pieter Punter, Sébastien Lê, and Jérôme Pagès.
At the end of 2012, Thierry joined Anne Hasted’s team at Qi Statistics, where he ran clients’ projects and gave diverse trainings. With Qi Statistics, he also continued a long collaboration with EyeQuestion, where he produced the R-routines included in the EyeOpenR software. Pivoting in 2019, Thierry joined the Global Sensory Department of Friesland-Campina, where he provides his expertise in Sensory and Consumer Methods, Sensometrics, and Data Science.
Besides publishing various papers related to Sensometrics, he is also the co-author with Sébastien Lê of “Analyzing sensory data with R,” and is currently working on “Introduction to Data Science for Sensory and Consumer Scientists” together with Dr. Julien Delarue and Dr. John Ennis.
Transcript (Semi-automated, forgive typos!)
John: So, Thierry, it is a pleasure to have you on the show. Thank you very much for being here.
Thierry: Yeah. Thanks John for the great introduction and obviously, it's a pleasure to be talking with you today on this podcast. My first one ever so quite interesting experience.
John: Well, I'm sure you'll be a great guest, so very happy to have you here. So, Thierry let's talk a little bit, I think you have kind of, you know everybody's background is different when it comes to sensory. And you had a little bit of sensory knowledge before you started working at OP&P. Can you take us a little bit of kind of your journey you came into sensory more from the statistical side, is that correct?
Thierry: It's pretty much the statistics side so I was with you on projects, for instance, and Sébastien Lê amongst other as a teacher and it was part of the agronomical school. So we learned a lot about what is agriculture and agronomy. But then you specialized and I specialized in applied statistics which some of my classmates went to completely different field than myself, like in genetics or in plant research or in biology or things like this. But obviously part of it was also sensometric so I was probably one of the first year when the packages were actually available and we were learning it. So we were kind of the first pioneer to spread it to the rest of the world, I guess. So there was quite an interesting experience.
John: Yeah, it's fascinating. It's interesting, it's a bit of a bioinformatics. I didn't really know that so much about your background.
Thierry: Yeah, I mean, the people from my from my year at that time so because we were like fifteen hundred I guess. No, 150. We're 150 but eventually we were thirty in applied statistics all the other or applied mathematics, however you were to call it. All the others would be going to environment research or marketing obviously because it's everything that goes around agronomical and agricultural things. So obviously you have finances, you have marketing, you have statistics, you also have like environment breeding for cows or whatever, plant research, genetic, everything. So obviously the background that you end up having is quite huge because you learn a little bit of everything until you specialize in when that's the moment to but I always had this joke at school that we were the few that really enjoyed full time what we do because most people didn't go for applied statistics because they hate statistics. But anybody had to do some, which was doing that because we like it. So at the end, we had nothing to surprise us when all the other one eventually did statistics and they didn't like it. So yeah, that was just my own joke.
John: Well, you know, it is interesting, there is a deep connection between statistics and agriculture, of course. Right? I mean, Fisher and many of the early statisticians were interested in agricultural problems, even in the language remains. Right? Split plot design, these plots are like plots of land. Yeah, very interesting. I didn't actually know that about you. So we've been friends for about a decade. So let's go talk a little bit then about how, so even back in those days you were programming with R? I mean, were your early experiences in statistics in R or where it was our subject came later into your career?
Thierry: No, I really came from that time because I think Pieter was one of the pioneer at that time of pushing towards R and that's why they actually did develop SensoMineR and FactoMineR because they saw the advantage of such software and the fact that mainly it was free and I guess they were kind of they saw that it would become big. They actually organized one of the first use R conference. So at the time where the number of people was still held, that you could handle. Let's say as a school organizing because schools in France are not like schools in the US. It's not the same size, it's the same capacity and so on. And they still managed to do that. And so I had a background in R through school. And then obviously I developed more and more through collaboration with Eyequestion software that I had at that time because I developed a routine for them. So I basically develop my skills even more there and then well, quickly was just my go to software and yeah and I'm still using it every day now. And I guess that's where if I would take my current work vocabulary, I would say this is what I get my energy from. Like programming, writing some kind of code, developing things and seeing things actually on screen eventually that do what you want to do which is actually quite cool, especially when you start automating stuff and getting nice application out, all these things. And that's really the beauty of R in the sense that at the start you feel a little bit afraid of what you're doing and you're a bit scared because it's in your language. It's complicated to be honest, R itself, it's not a beautiful software. So you really want to get into it. But once you discover this and it's a new world opening and every day you have a new thing to explore. There's so many things coming out, and every time you think about something that is original, be sure that someone already developed the package for that and that will help you and that's quite fascinating to my point of view.
John: Now, I totally agree and I think it's really turbocharged the field. I think if you look at, I actually think sensory analysis is pretty advanced compared to some of the other adjacent fields, especially in the use of multivariate analytic methods. And I think that the fact that R, FactoMineR, SensoMineR exist. Another package I would give credit to is a sensR. I think that sensR has done a huge amount of good in advancing Thurstonian modeling that Thurstonian modeling was around my father was beating that drum in the 80's, you know, but led to I think the mass adoption of Thurstonian modeling was the ability to do it easily and a freeway.
Thierry: Yeah, I'm perfectly in line with you. And I think sensR is really, I associate sensR to a little bit like programming in general, like Thurstonian model I guess by the time that your father was developing all these theories and all these models, it was really use because it seemed unreachable for most people because it was too complicated and you could not calculate things by hands. You needed like to have the computer tools, which means most likely, either you need to buy software or you need to be able to develop them yourself which requires quite some skills.
Thierry: And it's very scary for people to start with that but then eventually, once you see that you can reach it because sensR made it for the Thurstonian model, for instance, then you realize that it's not that complicated and actually it's way better than the proportion of this for instance. So which had at that time, like in the 80's, the advantage that was so simple you could calculate it by hand. And that's really what made the things now available. And now when you talk to people, Thurstonian model is just like its regular. If you do like any discrimination test, pretty much you have this Thurstonian model. They know what a deep prime is and they use it. Well, I guess when I started really joining the sensory and consumer world, I admitted I was kind of afraid of it. I thought like, yeah, but why would I go that way? Because proportional discriminators, I understood why it is like a computer but sensR opened the door so that's great.
John: Actually, I think it was played a big role in Tetrad testing, rising to the forefront. The fact that that theory was programmed to people could computer programs. I mean, there were of course tables or software, virgini had a V power program that was I think, very neat and useful, you know, that you made available. There were you know, but I do think putting it in R really mainstream, a lot of that. So actually be interesting to think, what do you think are the other things that now are potentially going to become more mainstream within sensory analysis? Given the ability of people to program? I have my own kind of list, but I know, for example, cluster stability is something that you've worked on that you find interesting. What are some of the things that you think people aren't doing enough of now, but they will in the near future once they realize that they can do it in software?
Thierry: Well, first, before answering your question, I think that's one of the things that I love about the statistics in sensory and consumer world, is that I think to me is one of the few fields where you can really do everything because you have univariate, you've got multivariate, you can really pretty much apply whatever you want. You can apply the way you want. We have the data sets to do that. Now, I guess to answer your question, I think the future for new development will definitely be going towards bigger data. I don't want to call that big data because I don't think we have the data suitable for that, at least not for now. But the bigger data and that would be connecting like more information from different projects, making like studying, working with databases. Obviously you have way more information than just answering consumer data. But product, in most cases, you can have like if it's your own company product, that you probably have ingredient products or you can have like instrumental measurements, or you can actually take instrumental measurements from competitor's product to whichever product you do. And then, of course, you can have access to all the sales information, all these things so at the end, you can have a very large amount of information that is very valuable and trying to find a way to connect all that, I guess that's definitely going to be the long term future. Yeah, it's just a question of having the right databases, connecting them so having the right, say, IT group behind you if you were bigger company but that's going to be the next big thing for sure. It will be a more and more and yeah, definitely the way to go, I guess.
John: Yeah, now I totally agree with that. I think that one area where we have been kind of lagging behind as a field is that we tend to collect, run a study, collect data, analyze the data from that study, draw some conclusions and then move on. And if people go back and look at work that's been done before, typically that involves looking at the past reports. Very rarely are people, there's exceptions, of course, and some companies are further along. But I think that, generally speaking, if you were to take, you know, what is kind of the median in terms of what's happening in the world of sensory. It's looking within data set rather than connecting across multiple data sets. I mean, obviously, sometimes I mean, in a drivers like a project, you might have sensory data and consumer data and you bring them together. But it isn't very sophisticated. It's not like there's a database that has all the data from all the research that's happened in the category or the company is connected that can be easily accessed. So I definitely agree with you on that front.
Thierry: I guess, yeah and I mean, I guess that goes back to the distorter as what I was saying about R, and getting to learn it which is maybe for some people quite difficult, but it's a matter of going out of your comfort zone and working on a project by project separately and then forget about it is what we used to. So that's the comfort zone because you rewrote the same project, probably standard, you know, which analysis to do which buttons to push, depending on which software you use. And that's very simple. So you're happy with it. But once you start saying, okay, now I'm going to start connecting different data sets and start looking at historical data and bring them together to try to get a bigger insight, that's where you go out of your comfort zone because suddenly you haven't done it. And that's maybe why some people are reluctant.
Thierry: And then to continue on your question, if you go outside the statistics, I think obviously the difficult situation we are right now in with the pandemic, also the positive side of it in the sense that we had to embrace some of the technology because obviously we noticed clearly for people who have been to EuroSense and thanks to your great talk as well, that the technology is there and provides us with a lot of great tools that for a long time we were just maybe scared to use or we thought like the old methods are working fine. So why changing? And then suddenly you're in a situation where you can attest as you used to in CLT so you need to be at home. So what could you do? And to control better than having smart speaker, having things more automatic, having more e-meetings, working from home, doing a lot of things like that shows that actually the technology there and was there probably already 10 years ago which is too afraid to use it. So because, again going out of the comfort zone.
John: Yeah, that is interesting. I mean, I think there's always been this tendency that, okay, we need results right away. And yes, we could do this other thing, but there isn't time to do it. There's never time to invest in the processes. And I think that Covid forced us to invest in our processes. And a lot of things are going to change. I think focus groups to a large extent will continue to happen remotely. There'll be some amount of bringing people in for, you know, especially you have to cook something. There's some challenges, you know, but there's a lot of advantages doing focus groups in people's homes. We just had Lisa Beck on the show and she was talking about this, about transcribed focus group data doing analysis of the text to go along with what she would normally do as a qualitative researcher. So I think you're right.
Thierry: The part where I disagree with you is the example you take about the cook. Because in sensory and consumer, we like to control for everything, and now that's why we have all these CLT-test and all these things. But on the other hand, we go more and more to it needs to be more natural and needs to be more like in line with what people do. And they can do that as a cook, just like I'm used to my computer. So if you give me another computer, the layout will be different, things will be different. I will be as comfortable in my own computer that I said that the way I want. And I can imagine for cooks, they have their own kitchen and their own pan. They have their own tools. So when you tell them, oh, you're going to come to that center, you're going to cook that steak on that thing that you're not familiar. I can imagine that for them. It's also not as natural because my mom, who loves to cook well, she's used to a kitchen. And when she goes elsewhere, it's not the same because, well, you don't know how it works and you need to sit at the table the temptation for these things. And that's why I think it's when you're doing home-use test, it doesn't matter. The pan is not the same, because at the end, if you compare five products, the difference will still be there regardless of the pan that you use.
John: That's interesting. You know, it's possible, actually, that maybe we should be taking more, asking people to make more objective measurements. And this is happening. People are having, you know, consumers take pictures of the food after they prepare. Maybe there probably are people who are taking temperatures, have people to capture the food before and when it comes out of the microwave, this kind of thing. So, yeah, I mean, I agree with you that definitely. There's always been this tension between ecological validity and scientifically controlled experiments. That is, I think that gap is being bridged by technology. But maybe with some of the measurements that are made in home, we could continue to bridge it because you would like to know I mean, if it's the case that half the microwaves out there function one way and half function the other way, and for half of the microwaves, your product turns out well, the other half it doesn't. You'd like to know which microwaves for which. There might be some additional information you'd like to include in your analysis, like the I don't know about microwaves. Maybe the water to microwave might be something as a variable that you could ask the consumers to record.
Thierry: I guess one of the limitations that we reach by doing more things at home is that people will need to read like the actual participants will need to read properly the instructions, and you have to trust them to read them properly and do what you ask them to do and not what they think they need to do because obviously, I mean, it sometimes happened to me, I guess, as well, when I have some questionnaires and I'm not fully into them, and then suddenly I'm asking questions like, oh, what was actually the question again? And you feel like you start getting into automatic mode and you kind of don't really pay attention to everything anymore. And that could be the danger of being home testing,because obviously, if you go to a place and you do the test there, you have the feeling that people are there and watching you. And so you may be more attentive to what you have to do. And or if you are not, maybe you can just go back and say, hey, excuse me, what was the question? Because the moment of I kind of forgot what I was doing and but can you remind me what I have to do there and then they can explain you. What if you home alone? Well, you have this sheet of paper, you've got this email, whatever it is, and you're limited to that. So you need to be very quick and you need to make sure that the people are actually doing what you're doing.
John: Right. And, you know, video is helping, I think with that. Right? Where some people are having interviewers, be on video talking to the person. I suppose a less expensive version of that might be a smart speaker if you at least are going to have some instructions if you are going to follow. So, yeah, these are all very interesting points. So let's get back to our kind of common interest in analysis. I actually am interested in a question that's always on my mind, is to what extent do people need to learn to code? I mean, obviously in our book, we are teaching people who would like to learn to code so that they can use sensory analysis. But what do you think are the important, what are the ideas or the skills that you think all sensory scientists need to have versus the situations in which you would say you really do need to learn to code? It's not going to be good enough to just have an idea of what coding is about. You yourself want to code like where do you kind of draw that line?
Thierry: Well, I think to start with learning to code is an investment. It's an investment and a long time that short period of time, if you are distressed, it's not going to be good for you because you will take way more time than you need to, to do the same thing. The long term, you will save so much time so it's definitely something that you need to see it in the long term. If you see it in the short term, it's never good to code because short term you will never win. The long term you will be winning so much more then it's getting useful. I would answer your question with a very nice analogy that Pieter Punter was always telling me when I was working at OP&P which was basically, he was commenting about most of the people that were working for him at that time. They were not really good statistics. They knew the basics, but they were not like for instance, sensometrician like I was and his argument was, well, they don't need to because they run a project which were quite standardized. They analyzed the data and they do the reporting. Eventually he was checking the reports, but then he said, and if one day something there is goes wrong, imagine them as driving a car. And you, Peter and myself, at that time, we were the people in the garage that if they have an issue with any data set, they would come to us and we would fix the car and then they could drive again. So I think this is a little bit what happens now. Everybody doesn't need to learn to code. It's good to have good knowledge about coding and stuff like this so that you can help thinking or having the way of thinking that is programming oriented. But you don't learn all the coding as long as you have someone that can help you with that, obviously. Well, I'm kind of in that situation, although I'm lucky enough, a lot of my colleagues are interested in learning to code at least a little bit. They are not afraid anymore to open R and see the interface and to run some code and stuff like this but I mean, if they want to learn more, I would be happy to help them learn more. But they don't need to be a second make on that topic because right now I'm there and I can just help them with that. And but it's definitely good for them to have bases they would want to learn to. It's good. But I think you need to have a few people that are able to do that in the company. And then the other one needs to have just the basic level to my point of view, to read the code and get things done when needed. And yeah, at least that's my vision about that.
John: Yeah, it's very interesting. I do think it's important for people to be aware of what coding brings. Right? I was on a call actually earlier today with a client where we were talking about doing some analysis and there were lots of ways to cut the data. Right? And every time you cut the data, there was a particular chart you could make that was going to be interesting. And so the client was saying, well, okay, why don't we maybe just do a few charts and we'll see if it's worth doing. And then I said better to me, we can do one chart or all the charts. It's the same amount of work for me. I'm just going to write the code to make the chart. And if I need to like, I say how about I run off a thousand charts and send them to you, we can look through it and it's a different way of working. You know what I mean? That it's one thing to run a few and look, it's another to run a thousand and flip through them and say, okay, like you're saying, meta the meta analysis, because then you can look and you can see, okay, is this actually as we flip through all the way to cutting the data. Is this something we want to pursue versus if you're going to do it by hand or do them, you know, one at a time in Excel stat or something like that? It could be very time consuming. Right? So I think that way of thinking, thinking about automation, being aware of what is automatable and what is not, you know, I think is really important.
Thierry: Yeah, I fully agree with you. But I guess that's where, for instance, my role comes in, in the sense that I'm helping automating stuff. But then my colleagues are using some of the basics on how to run my code. If it's still the code, if it's still in an application, it's easier for them because they don't need to open R, but if it's code, then they just need the basic to learn the code. Now, where I'm trying to be maybe a little bit more pushy is when we have interns and that I'm kind of the adviser of the interns, then if I see that they have a certain curiosity, I usually try to push them to learn things and be curious and to learn about it because they're interns. Usually they are like graduates or they just come out of school. So they have their entire professional life in front of them. So for them spending the time at that time, particularly because, well, you're full of energy, you're young, so you can still learn quite easily you have time then if you can create that passion and for them to learn it will be a huge benefit for them. Now, at some point, I'm going to hold their hands until the end. So at some point they have to invest themselves. And once the internship is over and that I am not working with them anymore, then they need to continue on their own. If they lose it, they lose it. That's on them. But at least to my point of view, that's the best moment for them to really get a grasp on it. And I'm really trying to push when they come to me, I try to push them to use R as much as possible.
John: Yeah. So what advice do you have to people who are actually interested in learning to code? I mean, what would be the process that you would recommend for someone who wants to get up to speed quickly, get like, you know, like what would be the plan you would, supposed I'm an intern and I come to you and I say, Thierry teach me. Like, what should I do? I want to learn to code, what should I do? What would be the path you lay out for me?
Thierry: Well, the first thing I would say, there's a book that is free online that you can buy it physically. It's not that expensive, which offered data science. That is a great start. You actually introduce me to that book John and it is definitely the scientific book that I have that I open and close the most out of it, because I think that's a great start to learn R And and I would almost say to fall in love with coding because it is neat and clean. It's very powerful and you can do a lot of difficult stuff very easily. So that's like entering data sets and all these things. So I think that's a great way to start. And then it's being curious, obviously, Google is your friend. There's a lot of, most of the problems that you will face someone had it before you. So that's no problem then I'm usually there to at least at the start, I'm there to give the kick start, showing a little bit, explaining how to code, giving them few tasks that they can start and get pushed to it. But yeah, I mean My main thing is be patient, and if you end up sometimes showing like applications or automated reporting and stuff like, look, now I'm changing, I have a new data set, look in two clicks, I've got the results. You still need two days of work because you need to set up the data in Excel that you make a lot of mistakes. You need to run the analysis and if 10 days you learn that the data set is wrong, you need to start all over again. So I guess if you can start, if you use this as a spark for them to just like, okay, I see why I should do that. It's going to be an investment. But if I can reach like I don't consider myself as a degree to all programmer ever, because I'm definitely not, but I think I know enough to make my life easy. And if people think if I can reach his level, I can make this person can make his or her life easy then it's a win-win situation. So why not just trying to get there? That means weekends and evenings reading books, learning stuff outside while you know what it is, I'm sure it's being passionate about something. I don't see it as work, actually, that's the point, I don't see programming as work. I see that as a something which is good because, yeah, it makes life easier and happier when you don't work, really, but you just do your passion.
John: Yeah, I totally, totally agree with that. I mean, if I can get a nice cup of coffee and sit down on a program for an hour, it's Nirvana because it's just so pleasant. You know, you're solving your little mental puzzles and you feel like a wizard. You start doing your stuff and magic happens. And it can be really a magical feeling. You know, you learn these spells basically and you cast them and magic happens.
Thierry: And start early like your daughter. Start very early programming.
John: That's right. She will for sure. And I would add one more thing to this, because I think it's very important if you already have a way of doing the analysis to try to, supposed you have to do the analysis for a report. Well, go ahead and do it the old way, but see if you can replicate it or at least replicate part of it using code and do that as soon as possible. I think that people do make the mistake. I feel like, okay, when I study enough, there'll be some magic point in the future when I'm ready to start doing it to my real data. Now, I would say get started right away, like try to get your hands dirty with your real actual problems and, you know, maybe do it your old way because you have to not get fired. But like, you know, that would be something. So, Thierry, I could talk to you forever. I love talking to you. It's always a pleasure. What advice would you have for them, would this be the advice to a young sensory scientist that to start studying, coding or is there are other bits of advice? Maybe life advice that you would give to a young sensory scientist?
Thierry: Well, I guess the biggest one, going back to the start of our conversation is don't be afraid of change, don't be afraid of things. Just be very curious and don't be afraid to reach out to people for, I don't want to say for help, but like to get stimulated because nowadays, especially, again, with the technology you can reach out to around the world. I mean, look at us right now, you are in the US, I'm in Netherlands, and we just talk like we were sitting together on the nice beer or whatever like we used to do in conferences where we could still go to conferences. But I mean, that's the beauty of the thing. So contenting people don't be afraid of it and of course, be very curious. Always try to see next. I guess that would be my advice. At least it worked out for me.
John: Yeah, we are well suited for this moment, like the way that we want to approach problems happens to be the way that is valued at the moment. So I feel fortunate for that. So, Thierry, how can someone reach out and connect with you following this podcast?
Thierry: Well, I guess the easiest would be to contact me via LinkedIn because that's probably the easiest way to reach out to me. So yeah, I would advise that. I mean, obviously anybody who has my email can contact me there, but it's easier just to go through LinkedIn, send a message and I should have responded very quickly.
John: Okay, awesome. Alright. Any last piece of advice for our audience?
Thierry: Well, just enjoy the podcast and yeah, and in the meantime, thanks for John for having me. It was a great pleasure, obviously. Always nice to talk to you and it's a great initiative that you started. I really love, I must admit, I didn't listen to all of them yet, but it's definitely on my bucket list to do and I'm glad I was actually a part of it now.
John: Okay, wonderful. Okay, thanks a lot, Thierry and we'll be in touch. Okay, that's it. Hope you enjoyed this conversation. If you did, please help us grow our audience by telling your friend about AigoraCast and leaving us a positive review on iTunes. Thanks.
That's it for now. If you'd like to receive email updates from Aigora, including weekly video recaps of our blog activity, click on the button below to join our email list. Thanks for stopping by!