How better to capture New Year cheer than in the form of alcohol? When Havas helia joined forces IBM Watson (a data analysis tech platform) to bottle the spirit of the season, they ended up creating the world’s first ever beer ‘brewed by data’.
The customer engagement agency helia analysed thousands of New Year related messages on social media and matched them with a range of emotional states, before analysing different beer recipes and categorising them according to human characteristics, using Watson Personality Insights. The top ten beers that matched each of the most shared New Year emotions on social media were then identified, and each recipe analysed to decipher the most commonly used ingredients: honey, the Nelson Sauvin hop and the Hallertauer hop. And thus beer was brewed! A crisp cream ale that genuinely tastes mighty fine (we were one of the lucky ones to try it).
LBB’s Addison Capper chatted with Havas helia executive creative director Steven Bennett-Day to find out how the agency pulled it off and what the process could mean for the ad industry of the future.
LBB> Where did the idea to brew a beer using data initially come from? Had it been around for a while and you were waiting for the right time to do it? Or did it actively come out of the task to celebrate the New Year with IBM Watson?
SBD> The idea came from wanting to prove that data and creativity together can create something interesting.
When our industry is discussing data and creativity, the main focus is around programmatic. Sure programmatic is going to be hugely influential in the future of our industry, and how we have to think as creatives, but the possibilities of bringing together data and creativity can create so much more.
To prove it we wanted to use data, creativity and technology to interpret something abstract – the feelings and emotions around the New Year – and turn it into something people could physically experience.
What could be better than a beer that lets you drink in the optimism of the New Year…?
We didn’t set out to use Watson, it was the right tool for the job.
LBB> How did the rise in popularity of craft beer inform the project?
SBD> Absolutely not at all… I much prefer wine or gin!
Seriously though, craft beer is culturally important and it was important to us to make a product that had a cultural grounding – something tangible and easy to understand. Creating something physical from something so seemingly abstract meant that what we created had to be universally understood.
LBB> The first step in brewing your beer by data was to analyse thousands of New Year related messages on social media. Which platforms did you follow and over what period? Why?
SBD> Twitter and Facebook. Simply because those are the channels on which the most emotion would be displayed. We wanted the broadest demographic possible.
LBB> Were you surprised by any of your findings?
SBD> We weren’t too surprised to find that a majority of the sentiment around new year was optimistic… although there was also a hint of ‘dread’…
LBB> You used Watson Personality Insights to categorise each beer according to different human characteristics. Can you explain how that works? What does the Watson Personality Insights do?
SBD> Because this really only makes sense in the whole process – here’s the whole process …
Once we had the top ‘emotional’ keywords associated with the start of the New Year from Facebook and Twitter, we then scored thousands of New Year-related messages against a set of 24 pre-defined emotional states. This produced 38 separate emotions with a clear tie to New Year, ranging from optimism and excitement and resolutions to joy and dread. The top ten: Amazement, Anticipation, Generosity, Indulgence, Excitement, Happiness, Joy, Love, Good cheer, Reflection. Then we fired up a couple of IBM Watson programs – using Alchemy to analyse 2800 beer recipes, pulling out the frequency of ingredients, their recipes and tasting notes – and how people reviewed the beers. Next Watson Personality Insights to analyse each recipe and characterise it according the values, needs and emotional states – essentially profiling each beer as assertive, friendly, intelligent, etc.
Then we correlated the frequency of ingredients against corresponding emotional scores for the beers, identifying the flavours most commonly associated with our top New Year emotions. The beer that tastes like New Year optimism should contain honey, Nelson Sauvin and Hallertauer: honey for love and cheerfulness - nelson sauvin for optimism, imagination and resolution - and Hallertauer for excitement and emotion. We also added notes of citra as this relates to ‘intellect’ and we wanted the beer to taste a bit clever.
LBB> How else do you think these kinds of findings could be used to create other products?
SBD> Well since this project I’ve dreamt of opening a bar where the more you drink, the happier you become…
Connecting the slightly intangible world of feelings and emotions to products and utilities is dawning on us fast. Imagine an alarm clock that understands you are still tired as you were out late, it predicts your emotional state based on past patterns before you wake up… It could re-set your alarm, re-arrange all your meetings and book you a massage followed by a coffee with that friend it knows always cheers you up by how your body responds… The algorithms for that are already here. It’s connecting them to human need and warmth that keeps them one step away.
Ideas like food made for the emotions felt at a dance festival, clothes that respond to how a city is feeling… this beer is proof it’s a reality right now. It’s how we connect the creative thinkers with creative data understanding and human interests.
Addison Capper enjoys a nice, cold, data-brewed beer...
LBB> You identified the top ten beers that matched the most popular New Year emotions. What were they? How do they relate to the emotions?
SBD> We thought carefully about answering this, as we’ve been asked a few times, and we decided not to answer it in this way… Sorry. Purely because our intention in identifying those beers was to assess the ingredients and how they correlated to emotions – not to chart or rate someone else's products.
However, we did find some significant patterns in most beer on the market:
By far the strongest recurrence according to our analysis of 2800 recipes, was with the personality attribute ‘imaginative’ which applied to 67 per cent of the beers we analysed - implying that beer fires the imagination. Second to that was ‘intellect', which might be what leads to us loudly sharing our opinions in the bar. The third strongest correlation is with ‘cheerfulness’.
LBB> Can you tell us a bit about the beer that you eventually brewed? Which style was it and what ingredients were in it?
SBD> The result is a fine tasting cream ale, initially in a limited edition of 500 bottles.
The ingredients most people may recognise are:
Nelson Sauvin – correlates to ‘Imagination’
Citra – correlates to ‘Intellect’ (a citrus flavour)
Honey – correlates to cheerful, agreeable and harmonious (and it’s a common flavour)
A hop called Galaxy – correlates to ‘Curiosity’ (an Australian circus/fruit flavour)
Corn sugar – correlates to ‘Love’
Styrian goldings – correlates to ‘Hedonism’ (a mild, but bitter hops)
Hallertaur – correlates to ‘excitement seeking’ (a very commonly used hops with a bitter taste)
Yeast – correlates to ‘altruism’
LBB> Who did you collaborate with on the brew and why?
SBD> Our beer was brewed in partnership with an expert independent micro-brewery in the Peak District, UK, High Peak Brew Co.
We chose to work with them because Ben Millner, the head brewer was very open to experimenting alongside us.
Ben's brews are also unfined and unfiltered which meant we got the closest taste to the data we could get. That was important to us.
LBB> How was it to collaborate with them? They are creatives in their own right… how did you work together?
SBD> We kept a close link with Ben when the recipes were coming out of our data experts and Watson, and he agreed to brew it, whatever the data showed. Which was one my key criteria in making this project. Even if you think it’s not going to taste good, lets make it…
All micro-breweries are craftsmen/creatives. Not many craft beer enthusiasts were that keen to let data form a part of the artisanal craft of brewing, but Ben saw it as exciting in the same way we did.
LBB> Who was the beer distributed to?
SBD> We’ve sent beer out to our favourite people, people who are interested in creativity and data, and also to our clients as a symbol of what we can do for them. At the moment it hasn’t been on general sale but we’ve had a lot of interest.
LBB> What were the trickiest components and how did you overcome them?
SBD> Mapping emotions to an inanimate bunch of ingredients…
LBB> Any parting thoughts?
SBD> How data affects our industry is one of the big questions to unfold over the next couple of years.
I believe right now we’re looking at data and creativity as two distinct things. Creative AND data. Ideas AND numbers.
And in the process of creation, creative comes AFTER data, as a response to it.
The obvious view is that data empowers more accurate creative work because it’s based on a level of empiricism. The optimistic view is that data allows us the opportunity to find creative reference points we could never have seen before.
Accuracy and targeting of messaging is very much the first toe in the water.
Although data can tell stories if you know how to read it, it’s not usually in a way that can make you cry.
I think that creativity will migrate further towards data to influence the questions we ask of it and the way we use it. And data will migrate towards creativity to find the cultural points that start to touch emotions.
In short the ‘handover’ between data insight and creative story telling will become longer.
Meaning both creatives and data experts are going to need to think differently.
And thinking differently is what we champion in our industry, and what we aspire to as the spark of creating something new and genuinely interesting.
So this beer to us is a symbol that creativity and data are not at war.