Michel Bonnet - Strong in Science

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!

Michel’s career has focused on Consumer & Product Research & Sensory Design. He has held a number of senior research positions in the industry, including Director of Insights & Sensory Design at Danone. He is also a former Head of Product research & Sensory Design capability at Unilever.

Since becoming an Independent Consultant, Michel shared his expertise with research teams around the globe. His focus is to help teams overcome their challenges by bringing an external and creative viewpoint to deliver people and product insights that will drive business decisions and product design.

Michel has been one of the first in the industry to make the Sensory & Consumer Science function an Insights powerhouse. Since the start of his career, he has been a strong advocate for People Centric Innovation projects to launch products that overcome consumer’s tension and truly enhance people’s lives.

LinkedIn

Transcript (Semi-automated, forgive typos!)

John: Michel, thanks a lot for being on the show.

Michel: Thanks for having me, John.

John: It's a pleasure. Michel, we have many shared interests and I think also some might actually be interesting because I think we have some shared viewpoints. I think there's a few kind of subtle differences in the way we see things. So one thing that I idea that we share in common is that when people start thinking about new technologies, there's this tendency to just simply want to do whatever you're doing faster, more efficiently. This idea of more with less, that kind of thing. But I know that you kind of share this viewpoint, I know that's not enough. So can we talk about what else you think people should be doing with new technologies?

Michel: Yeah, sure. I mean, we discuss this a couple of times. We see that in the industry. People see digital at first and AI now. Technology go faster, often to go cheaper. But beyond that, there is this idea to say, oh, but we will do the same.

John: Right.

Michel: On the short term, we can expect to do something, we need to be faster, maybe cheaper eventually. But the big opportunity of AI is to do new things. And for sensory professionals, that we are, we want to understand to explore the product experience. The extent that the consumer goes through when they use our products. And we know that the way we research has bias. For many, many years, we've been asking questionnaires long often. Now they are shorter with digital but they are simply false rationalization. The question we ask people about memory of the experience they have. We've been able to get closer to the in-moments and we talk about that very often in the moment. And that is helping us to do new things, being closed to the emotions. We know that in liking, as we used to it, acceptance, deep emotions plays a bigger role, actually than the rationals. So being in the moment, we are closer to emotion. But I think it's beyond faster is better is how can we do things and understand to things that we were not able before. Well, we have this distinction between quantitative which are question direct questioning and qualitative where we are setting at focus group or whatever observations those kind of things. And AI I think as this opportunity to help us merging these not just challenges of qual and quant but being able to kind of really observe and measure at the same time that we are not able to measure before. As an example of years, four years in a universe, we were doing observational studies to understand the behavior on how much water, for example, people are using when they are doing their laundry. Wanted to know how much water they use. It was to understand the nature of the powder and therefore able to calculate how much enzyme wanted to put in the powder then about sustainability and how can we reduce the water they need to use, etc. But with that it was extremely heavy. It was extremely complex. We were putting devices on the washing machine. I mean, there are many, many things. How today the technology can help us with that. But another example, a few years back, I did a similar study on drinks, for example. And when we were very wondering was the temperature of the drink. And we are thinking, we assumed people were drinking the soft drink at four degrees. So we gave the people a portable thermometer to measure the temperature of their drink when they were drinking. It was far from full, it was 10 degrees. So that's the way you sense the bottle and that's changed the readings of the flavor. But what I expect from AI is really to have first, not to give a portable thermometer to the people, but to use a device they are familiar with because with AI we can use this opportunity to do a research by using the new behaviors that the consumer have. So today taking a picture with your smartphone is past behavior. Five years ago or six years ago, we were asking them to make a picture of a bottle or something, it was not really yet completed behavior. When I'm doing cycling, I mean, people take pictures on the bike all the time. Video is even. So asking people to take a picture of the experience of the products is part of their general experience. So I think we have the opportunity with AI to make research part of your normal life. So we really changed the dimensions of what we do.

John: Yeah, that's really interesting and there's a bunch of kind of follow up questions I have about that. I mean, one thing that I like is the use of these technologies in a way that is still consensual that someone if they've agreed to be in an experiment. Right? But they can go about their lives as opposed to surveillance, which also happens.

Michel: Exactly. It is science and I think this is one thing we need to be clear about science, we talk about scientific research. So there is an experimental design, there is a setup we see half the time. So it does not change the propose and somehow with approach. That change is the relationship we have with the respondents. Part of your life we want to do research as part of their life as close as possible and behavior and how they not just react with pure liking like it or not. But really what is the food experience from the moment they buy it to the moment they dish the pack somehow without interfering. We have done those observational studies where you go to the home of people and you look at them and but you are there. You're not supposed to be there. But they want to please you somehow and still the same. It's like when we do nowadays, you know, communities where you talk with people via zoom, for example, what they do with their friends, they zoom all the time so they can do zoom with you without feeling they are doing a survey. And that's something which I believe AI has the opportunity for us to change.

John: Right. Obviously, the covid situation has accelerated. Like you're saying, it's changed people's behavior as people have had to accept these digital tools in order to live their lives. They've integrated them into their lives. And now not unusual that like during the part of the focus group, you know, with eight other people on Zoom, that no longer seems like a strange thing to do. In fact, it's very comfortable.

Michel: Exactly. And I think that's the big thing for me is people might finds themselves more comfortable in the new set up than the old one.

John: Right.

Michel: We all have done those focus groups where you spend 10-15 minutes making people at ease because they're not in the usual place because they don't know the other guys. Now, you can imagine things where people are more and they are used to. You know, for example, in sensory, at the moment there's covid, a lot of discussion of doing sensory at home rather than in your lab. With a good strong line to months, we did some 10 years ago, maybe 10 years ago. They did for products that was in the UK. We decide to do sensory at home because what we found at the time is that the people were not doing the gravy at ease in the lab although they had the kitchen and everything they needed. We discovered that something was wrong and we decided to do the generation home. At the time, we didn't have Zoom and all those things, but some words came out at home which were not out in the lab. Of course, at the time, we have a mixture of stations in the lab and at home so in the lab but we found out so much at home because they were home at ease. We were not there. There was no booze or whatever. And those attributes helped us to explain the consumer behavior better actually. That was some time ago without all the facilities we have today. So I think we can work much further nowadays.

John: Yeah, fascinating. You give me an idea here for Uber, for focus group moderators where you could have get a whole bunch of people who are skilled interviewers. They're part of the network and when someone's ready to make the gravy at their house, they say, I'm ready for an interview. And any one of these interviewers can take the job and they can hop on and do an interview.

Michel: Actually, that's funny you say that because I'm working with a client right now where we want to do some observation at home and think about they need to buy our AI tools so we're not there physically, but we want to do a final interview but in different countries. And with covid, the guys kind of really troubled in a way that they would have language barrier. So we are thinking of taking some local people from the subsidiaries to attend to those interviews, although they haven't been part of the food study. So it's a little bit the same idea where you can take local expertise to help you and still using the more central team or whatever you want to call it to be engaged so that is the same idea.

John: Yeah, it's fascinating and of course smart speakers help with that to a point, although a smart speaker, I think there's a tendency for people to think of them as qualitative measurement instruments when they're still really quantitative. They just have a kind of qualitative feel. So they're on demand whenever anybody is ready for the survey. But it's still not going to be the same as asking someone questions while they're making the gravy. I mean, you can walk them through it. But it is interesting.

Michel: I think I agree with the idea. I mean, of course, we still have to learn how to use things like this. We have to find a new devices, of course, but we have to work on the one we have. And I agree with you, of course, we imagine them as qualitative and we see the algorithm today functioning well into data analysis. But although we can ask some questions and maybe it can be more a question that today we are used to, it's a life. It's not a question like before where we were asking a question that some consumers want something in one way and all this? And that was when understanding in a different way. Here you can have something which people are more used to. We see on LinkedIn, some people are posting something. You can be creative, I think, around those things and how you can use speakers. A little bit of quantitative or more some qualitative research. And even though, for example, on the drink example I gave when we were taking temperature, taking pictures of the bottle, because we wanted to understand how many people were drinking where they were having to drink the whole bottle, because again, the temperature and the time plays a role. So we asking them to take a picture and before taking the picture, we are asking them to mark on the bottle with the marker the level of the drink. So we then have someone saying, oh, this amount of millimeter or whatever. Today, with AI we do that for us. So you don't ask people to have a marker. You just take a picture and that's it. And then the computer will tell you the amount of water you have left in the bottle. So this I don't think where it's going to make our life easier. And you can then imagine more complex studies to get into the behavior of the people.

John: Right. Yes, that's fascinating, Michel. You're really giving me the awareness. Okay, so one of the central issues in science is that you can't observe something without interacting with it to some degree. Right? Historically, we've been sometimes very far from the actual moment of product experience for various practical reasons. Technology is taking us closer, but it's also allowing us to interact less, that it's a more lightweight observation because there's more we can do on the backend.

Michel: So we have discussed about the opportunity which is we cannot yet imagine everything. And what you just said is for 4 years that for all those years in sensory, we've been really studying the product or product variation. And we’ve been in the lab because we want to kind of study these variations between the product formulations without any impacts of environmental non-studied variations and here we have the opportunity to put the people in a context so that we have emotions and full mindset. But the risk is to have all those noises. So this is as well something that AI would have to balance. There is the opportunity but still it is a science as we said earlier on and keep reading the sensory of what it is and control what we are doing. It's not just an opportunity, it comes with some challenges as well.

John: Yeah that's quite interesting and I know you have some thoughts on meaningfulness. You know that we do these analysis especially the more black box a model is the more you have to, I mean, I think that there are technical approaches that can help us to have an idea what's going on inside a model. But it might be interesting for our audience to hear some of your thoughts around meaningfulness when it comes to these more advanced analytic techniques.

Michel: Yes. Well, when I called meaning from that side I don't know if it's standard word, but it is the fact that as I said, AI is somehow black box. We have many things we don't know. Until now what we have was statistician cleaning the data. If you wanted to put two sets of data together, you had to clean the data and making sure that there were meaningful. What we see today was big database, for example, is they merged everything. You can say, oh, it's quantitative and it's different, whatever. But what is the meaningful behind that. I have an example which is a personal one which happened to me last week, actually which is nothing to do with sensory but I think its means what I have in mind. I'm doing cycling and cycling is an endurance sports. You have a watch which takes your kilometers you've run and all those things and the climbing you've done and whatever. There is an app where you can connect your tool and on the app you share with your friend’s classical behavior. Everyone does not. So all the travel to France and everyone download information there. I went to my device as it connects to this app and now everything I do is on the app so my friend can see it. My device gives me numbers and the app give me slightly different numbers. What the people see actually is me doing more climbing that I actually did. So I spoke with the guys and they said, oh, yeah, there is maybe something you can do, whatever. And I see this with my group of people who have the same problem and I cannot do anything. So I think this is the purpose of the meaningfulness for me is until now, when we had one set of data you have statistician working on everything and not like me doing more climbing than what I do the same number of kilometers and the others would be see as an outliner and think, what's wrong here? When you have an algorithm doing that, you might see if it's overlooked. So maybe it's not important, maybe it is. And that's something for me we need to be to be careful of. I have another example which is more in sensory field. A few years back, I did work with a statistician to put together five preference mapping on ice cream and the end of that company was about all we want to have the same value in flavor all the countries. And the statistician went into the data and those five preference mapping were done independently. So we had a problem of it was not the same flavors or somewhere, but anyway, so we looked at the data and we decided that we could not solve them. And what we found out was, no, you cannot have one. However, we found the same clusters in all the countries. So that was a big insight for the company. We said, well, guys, you have the same clusters in all the countries, it just happen that the size of clusters are not the same in the country. Now do your strategy or whatever. Before I talked them, what's important is after that I asked, well, what would have happened if you have put the Australian data in it? And we can have the same clusters. We didn't have any conclusion at all. So I think the question is, what if you ask an algorithm to say, I want to merge five sets of data, what would be the answer? The algorithm being able to say, although there is one which is really different, we're not sure. And that's what is behind a meaningfulness. And with this complexity of analysis, how are we going to ensure that what we have is still meaningful.

John: Right. Well, yeah I mean, I definitely am on board with the idea that there should be a human in the loop. And I also think that we don't really have so much data that I mean, it is true sometimes you hear about these digital initiatives where some tech people get involved and they basically just try to take all the data in the database and put it together in one way model. But in this situation, there's not really, you know, it looks like a lot of data to us, even five different category for instance really isn't that you still need an expert to help you. Yeah, I mean, it's interesting because this is the idea of bias, right? Bias is a filter that helps us interpret the world to an extent.

Michel: And that brings us back to the very first question you had about people just wanting to do faster. That's a risk. You know, you say go faster, I'm not an expert. Oh let's merge the data and ask the algorithm to do it. And you lose the expertise in fact, which is going to have meaningful data insights that you have data and then you merge them and you get some conclusion which will lead them to insight. And that's I think it's an opportunity, but it requires from us even more science. It's not just the permission to ease it somehow definitely not.

John: Now, I totally agree, and I would say definitely my experience that if there isn't some way for subject matter expertise to find this way into machine learning, an app based analysis of sensory data it will almost certainly fail. That's my experience. There has to be an entry point because just kind of you know, people talk a lot generally like bias as a social phenomenon is often a bad thing. Right? However, when it comes to analyze like building models sometimes bias helps you to make sense of, you know, if there's pattern or something based on your experience in the past in the product category, you might actually be biased to understand what that pattern is whereas a machine may not. Okay, Michel, we are almost out of time. I do want to ask you what technologies you're most excited about? What are the things that you think are most exciting right now?

Michel: Well, today and because I haven't used it too much yet. I think the speaker is something which is becoming a behavior with people and the speaker can be the speaker like Alexa or whatever or the speaker on your smartphone or whatever is really only because we see if you use WhatsApp, all the things, for example. People don't write anymore. They just give you a message. So I'm really into exploring that. How high we go to be there with the people we are sitting.

John: Right. So voice activated technology in general, not necessarily just smart speakers, but any sort of voice technology.

Michel: I guess it is the combination with voice and pictures.

John: Yes. Okay, that's fascinating. That's interesting. I'd love to talk to you about your thoughts on sound but we are actually out of time. I think sound fits into all of that too and it's something else. Okay, Michel, supposed you are going to give some advice to somebody who just finished their say Masters in Sensory or Food Science something like that or maybe consumer research. They're going to go into product insights, in some of the areas that you've worked in, what would your advice be to a young scientist just starting?

Michel: Well, I guess is three things I would say. The first is really study science. I mean, we are researchers. So kind of compromise on science. Be strong on science. Whatever you choose, be strong on science. The second thing is because you are strong in science, you can be pragmatic on a plan. Don't be a one hundred percent person, be pragmatic. The world is moving fast, we need to move fast. Industry needs to move fast. Your project will have whatever the university or an industry that we have to move fast so be pragmatic. Don't be a hundred percent person. But you can do that only if you're strong on your science. That's the two first one. The third one is if you are in the sensory world, you need to love product and people. You need to care for products and people and sending the tensions and solving those tensions and the way to find insights and share your insights and convince people to go with your insights. You to love products and people if you don't love products and people do something else.

John: Okay, great and supposed someone wants to connect with you after the show, what's the best way to get in touch with you?

Michel: Well, LinkedIn of course. When you're in LinkedIn you would get to my website and email.

John: Okay, great. Well, this has been wonderful, Michel. Thank you very much. I’ve personally learned a lot so I appreciate you.

Michel: Thank you very much, John.