El Bulli: It Could Be More Than Foams and Olives
Explored machine learning applications in fine dinning restaurants.
El Bulli, “an avant-garde, three-Michelin star restaurant that has pioneered the “molecular” or “experimental” cuisine movement in the gastronomic field,”[1]closed down in 2011. 8000 foodies from all around the world every year travel 2 hours in the Spanish mountains for a 5-hour, over thirty courses, 230 euros tasting menu.[2]Yet, El Bulli’s whole-year reservations are gone within one day, leaving millions of people on their waitlist.[3]Ferran Adria, El Bulli’s co-owner and head chef, is considered the father of gastronomy, taught a culinary physics course at Harvard in 2010. Highly regarded with huge customer demand, what caused El Bulli to operate in financial losses, around 500,000 euros a year since 2000?
El Bulli only opened for dinners six months a year with the other six months researching and developing menu. In addition, his large staff team, bold and pricey experimental ingredients, seasonal and irregular supply chain are all contributing to their success in innovation and failure in business.[4]Since 2005, Ferran Adria and his team had to focus more on the process to reproduce their dishes and organization instead of developing new dishes, in the shadow of continuous monetary loss. “As a commercial restaurateur, I have certain responsibilities. When people bring up the issue of reservations, I have to give some kind of explanation.”[5]Eventually, Adria’s passion for finding the meaning of cooking surpassed his patience for running a restaurant. In a parallel universe, machine learning may have solutions to his problems with costly ingredients, customer satisfactions, and resources to develop new ideas.
In a short term, there are machine learning algorithms analyzing FTIR data to quickly identify meat spoilage.[6]Moreover, El Bulli could use machine learning predict the quantities and qualities of fresh local seasonal produces to better plan accordingly.[7]Both methods will help cut down their inventory costs.
Machine learning also offers benefits long term – analyzing data and recipes from all over the world to find out clusters and trends in flavor combinations and profiles. “His team have used scientific research to invent new techniques, such as substituting agar-agar jelly for gelatin, because they found it can be used at higher temperatures.”[8]Algorithms can speed up their innovation and investigations for new dishes, but does not sacrifice the quality and creativity of El Bulli’s food. Machine learning can help create visual maps exploring the differences between 3000 different varieties of tomatoes and cooking techniques respectively. In addition, produce varieties that are less common used in cooking can be identified as outliers in clusters and provide new ideas for Adrias to research on. Furthermore, algorithms can help El Bulli predict customer satisfaction. With customer data collected after the meals, El Bulli could create preference profiles from different regions of the world. “There’s not enough time for cooks to reflect, read, and learn from history,” Adrias explained during his talk at Harvard. With machine learning, he could better recognize and learn from what his customers think about his dishes and incorporate feedbacks for further improvements.
“So, seeing chicken curry as a concept and determining to do something that hadn’t been done before, he developed a dish, now famous, in which the sauce is solid and the chicken liquid.’’[9]
Algorithms can help to resolve his issues with not enough time to create dishes like the one above. However, is a profitable and customer satisfied business what El Bulli wants? What makes them unique and stand out from other Michelin three-stars is finding the delight surprises that Adria’s team created. This is a closer form to art than cooking recipes. Are the 0s and 1s in a machine learning algorithm capable of creating art and exploring the unknowns, with data only from what people have tried in the past?
(705 words)
[1]https://academiccommons.columbia.edu/doi/10.7916/D8MP51CR
[2]Norton, Michael I., Julian Villanueva, and Luc Wathieu. “elBulli: The Taste of Innovation.” Harvard Business School 509-055, March 2009.
[3]Henry Chesbrough, Open Services Innovation: Rethinking Your Business to Grow and Compete in a New Era, 2010
[4]https://hbswk.hbs.edu/item/customer-feedback-not-on-elbullis-menu
[5]https://www.decanter.com/wine-news/ferran-adri-i-m-tired-of-being-a-commercial-chef-46904/
[6]David I. Ellis, David Broadhurst, Douglas B. Kell, Jem J. Rowland, Royston Goodacre, Rapid and Quantitative Detection of the Microbial Spoilage of Meat by Fourier Transform Infrared Spectroscopy and Machine Learning, https://aem.asm.org/content/68/6/2822
[7]Forecasting yield by integrating agrarian factors and machine learning models: A survey, 2018, https://www.sciencedirect.com/science/article/pii/S0168169918311529
[8]https://www.dailymail.co.uk/news/article-1250914/El-Bulli-worlds-restaurant-close.html
[9]https://dinersjournal.blogs.nytimes.com/2010/02/12/el-bulli-to-close-permanently/
This was an interesting application of machine learning that I hadn’t considered before, I was curious on the choice to apply machine learning to food and culinary innovation, and why you chose to consider it for the most cutting-edge and high-end dining sector compared to something more mass market? There are definitely many innovations in machine learning and science/chemistry that can be applied to the field of gastronomy, as you pointed out. And then in the field of profitability it would be interesting to hear how applications in fast food or mass market food consumption machine learning data can be applied to a place like El Bulli.
This post reminded me of our Gap case where the company experimented with using big data to predict trends in consumer taste and shorten the cycle for developing new clothing items. One of the major concerns we raised is that the result could be a regurgitation of ideas that already exist in the market and a convergence toward the lowest common denominator. For industries like fashion and food, the trailblazers are the ones that set the trends and introduce products that customers have never seen before. I question how innovative El Bulli can be if they are riding the coattails of fads that have already peaked.
This is a restaurant that I really wanted to try before it closed down but unfortunately was not able to do so. I am skeptical about a couple of points made.
Firstly, while I think it is theoretically possible for Ferran Adria to use machine learning to predict stocks of local produce in order to better manage his inventory costs, I believe he will face a lot of difficulty in obtaining the data to do so. In addition, most high-end fine dining restaurants do not actually use “mass-market” produce and typically have their own local farmer who’d deliver a specific produce. Given this context, I doubt that it is practicable for him to apply machine learning to predict quantities of local produce. In addition, I also think it would be difficult for him to obtain data on the quality of the local produce in order to apply machine learning.
I am also skeptical about the use of machine learning to process customer feedback. Given that his restaurants only cater to a relatively small number of customers every year (as compared to more mass market restaurants), I think the effects of using machine learning to process customer feedback and obtain customer insight is rather limited. In addition, I also wonder if this impedes the innovation process since he will “constrain” himself by limiting his ideas to customer preferences, which may or may not be as informed as his own.
Agree with LW above, a very interesting application of machine learning that I had not considered before! Thanks for a unique perspective.
As we discussed in our Aspiring Minds case, I wonder if the application of machine learning in this avenue could introduce flavor / combination biases into Adria’s dishes. In this vein, I completely agree with your statement that this seems to be more of an art form, than a science. As machine learning infiltrates the ‘creatives,’ I wonder how we might further innovate to protect authenticity and originality. Perhaps algorithms can be trained for that as well?
What an interesting application of Machine Learning!
I agree with mrrobot’s point of view and drawing of similarity between the Gap case and a Michelin star restaurant using AI. Given the reputation and expectation of innovation from such restaurants, AI could kill the novelty of the restaurant. Though insights from different parts of the world about trends would be helpful to a restaurateur – the role of the chef or the human element of innovation cannot be eliminated in an art such as cooking.
Really interesting post! I love El Bulli – there are rumors that they will re-open next summer! I find it very exciting that artificial intelligence and machine learning have barely penetrated the food / gastronomy industry. What is clear is these new technologies currently only make sense for businesses that fortunately have the sufficient money to invest in them (i.e. high-luxury restaurants). Yet, I wonder how much of an advantage this would actually give them – wouldn’t they be better off investing the capital in better customer service or more exotic ingredients? Also, how much of the “art” would it take away from the whole experience?
Interesting topic! I am doubtful on whether master chefs like the Adria brothers would like to gain artistic or gastronomic inputs from machines. I believe that the artisan movement will continue to grow as aggressively as machine learning and AI as a form of counterculture. (There is evidence that machines can produce art though: a computer has composed a classical music piece which critics deem the same quality as master composers.) However, I liked the application on whether machine learning and AI can develop to a point where they can become a full restaurant management team (versus hiring staff). Part of the failure of El Bulli is despite its gastronomical achievements, it was not managed well. Instead of forcing artists to gain business sense, outsourcing key business decisions could allow chefs to focus completely on their craft and potentially draw up forecasts on how profitable each dish is.
Interesting topic and insight! I agree that this could be super helpful to fine dining restaurants to help them optimize otherwise inefficient process and better estimate customer preferences. However, I wonder what solutions already exist to help restaurants improve their business and, if so, at what cost. Is it or will it be feasible for one restaurant that already has extremely low margins, if profitable, to spend more money for something that has long-term benefits. I don’t think fine dining restaurants optimize for their bottom line so I definitely think the use of machine learning in this industry is a big possibility.
I have read about the use of machine learning and artificial intelligence in food supply chains such as ordering food items, managing inventory etc. which lead to less food waste and more efficiency which I think is an important global issue right now that we need to tackle. However, I am skeptical about the use of machine learning as a potential tool for high end restaurants in creating and predicting their menu / flavors. In my opinion it seems to be more of a fad than anything that will yield tangible benefits. Designing an interesting dish is fun and something that chefs are probably passionate about so why do we need to use computers in order to do that?
Really enjoyed this application of machine learning! I had never thought about it in the context of fine-dining restaurants. In answer to your second open question, I would echo some of the points made below. While machine learning would give El Bulli an efficient way of collating food trends from all over the world, I do believe that the restaurant’s goal was to give customers an unparalleled culinary experience that goes beyond anything that they might have tried in the past. Therefore I don’t think machine learning is the answer here because it doesn’t provide anything ‘new’ and therefore the chef, as an artist’s role remains central to the process. However, I do think you raise a valid point at questioning whether there is a role for machine learning in responding to the large volume of customer inquiries in order to maintain a relationship and encourage them to continue trying to get a much coveted reservation.