Heineken uses AI to optimize the color of Heineken beer. Why is color so important?

“Many factors play a role in brewing high quality beer, and one of the key quality parameters is the color. We want our beer to be the typical Heineken “gold”. However, currently 7% of our beer brews do not satisfy this requirement. To get the right color, these batches are remixed with other brews. This takes up production capacity that cannot be used for other purposes.”

It is surprising that the color of beer is not constant. What is the complexity behind it?

“Unfortunately, beer brewing is not an exact process: it is more an art than a science. For this reason, we have master brewers who play a very important role in overseeing the process. Beer color can vary from brew to brew due to changes in the process, changes in the ingredients, and a variety of other factors that have impact on the beer color.”

You use a machine learning algorithm to predict the beer color. How does it work?

“We developed a machine learning model that looks back at what happened at the past 10 brews and what ingredients were used. It considers the ingredients that go into the next brew and it predicts the color of that next brew. It then advices on how much color malt to add to get to that perfect Heineken yellow.”

How is this algorithm used in the beer brewing process and what are your results so far?

“When we tested the application of our algorithm in real life in the brewery, we found that the beer color of the brews was 30% closer to the perfect Heineken yellow. In our experimental setup, the prediction of the color is shown to the brewing operator and he changes the recipe to adjust if needed. But this experiment also has other, broader effects. For example, master brewers now have a better understanding on how beer color is impacted.”

What do you see as a next step for this AI solution?

“Now, the prediction of the color is shown to the operator and he manually changes the recipe to adjust if needed. In a more ideal situation, you would use the machine learning model to change the recipe automatically. Then you would need to install some safeguards of course, for example automatic triggers if certain safety barriers are crossed, or a warning if the model cannot make a good prediction. Other than that, we need to take this solution from experimentation into production, which comes with IT challenges.”

What does the implementation of algorithms mean for people and the organization?

“People need to get used to it. For brewing operators, it implies a different way of working: instead of deciding about the proportion of the ingredients based on measurements and discussion, they now do so in collaboration with an algorithm. When you apply machine learning to business domains, it usually means that people must change their way of working a bit or a great bit. For this reason, change management is a crucial aspect, and translators play an important role.

Beer brewing is an art. These are the steps taken in the Heineken brewing process.

What are analytical translators, what skills do they need and how do you build them?

“We recognized early that to be successful with machine learning, it is not enough to have a central team of excellent data scientists and engineers who can make all this analytical magic happen. To bring real value with these machine learning technologies, you need people that connect to the business and understand business processes. People that can identify opportunities and translate them in terms of what you need from data, algorithms, and AI technology.  We call these people translators and we have developed the Analytics Translator Program to train them.”

What does the future look like for Heineken with respect to AI?

“We will see AI in more business domains and more business processes than we see it now and in a more mature way. Right now, we are transforming successful experiments into operational products. On a longer run, we envision a Heineken analytical app store that contains all AI solutions, which are plug-and-play for operating companies to improve their processes. This also means that people need to become more data savvy. Just like we have English as our business language, data should also be one of the languages we speak.”

What are the main challenges to reach that goal?

“We need to focus on more than just analytics, data management also plays a crucial role. We run into the challenge that much data we have is not harmonized. Clean data is important, as important as good algorithms. But cleaning and harmonizing data and storing it in a safe and accessible way is time-consuming and challenging.”

What advice do you have for organizations on the same journey, with respect to training their staff?

“Build the skills of internal people. This way you ensure you build the knowledge and skill to create value with AI in-house. And make it a board priority: enthusiasm at board level is crucial.”