10 Parallels Between Whiskey Tasting and Data Analytics

In today’s world, the power of data analytics is everywhere. From agriculture to healthcare, from shopping to dating, from the vehicles we drive to the way we do business, our experiences are increasingly shaped by data analytics. This is true even when it comes to whisky tasting, although in this case the analytics process is driven by our senses and our reasoning rather than sophisticated algorithms.

This is a topic that is close to my heart, given that I’m a director of data analytics who moonlights as a whiskey sommelier. I often have occasion to reflect on the amazing parallels between the principles of data analytics and the process of tasting whisky.

With that thought in mind, let’s look at 10 of the ways in which data analytics and whiskey tasting share common ground.

Let’s take a step back and see how we got here.

Back in the 1980’s, when data warehouse vendors like Teradata provided the ability to pool data, business owners asked even more demanding questions. Then SAS and SPSS, whose origin owes to government and academic interests, developed tools that allowed for “what will happen” questions and not just “what happened.” Fast forward to now. Fueled by math smarts and entrepreneurial spirit, we now expect Amazon to recommend books when we shop or Uber to send strangers with an empty seat to our address. Doubt what I’m talking about? Ask Siri or Alexa.

Back to whiskey. By 1954, the U.S. saw the number of distilleries collapse into four companies. Courage and curiosity brought back independent distillation following the rise of microbreweries in the 1970’s. Pioneer Tito Beveridge planted the craft distillery flag in Texas in 1997 after he observed the first seedlings in Kentucky and Tennessee. What followed was a wave of craft distilleries with shoots emerging in California, Texas, New York, Colorado and Washington. It’s now a multi-billion business with 27.4 percent growth.

Amplifying this trend, cocktails popularized by shows like Mad Men or House of Cards put whiskey in our collective consciousness. We should no longer expect only wine to come in flights either. We can skip our way through whiskeys too. We arrived at the new normal: whiskey tours, bourbon runs and the rise of The Whiskey Sommelier. Whiskey tastings now pop up like daisies in a sun-drenched field. Whiskey tasting is an active sport involving all five senses and your brain.

Now let’s loop back to data analytics and look at the threads that tie these two worlds together. In particular, let’s look at 10 reasons why data analytics and whiskey tasting share common ground.

Deductive Reasoning

When online retailers look for patterns in clickstream data, they engage a practice known as deductive reasoning. I see this and I see that. Therefore this other thing is highly correlated. Ask people at Walmart why they stock strawberry Pop-Tarts in the front of the store before a hurricane. They will tell you it was because they saw a pattern.

Same is true for experiencing whiskey. An active whiskey drinker analyzes what she experiences. Does she note a familiar herb like heather in all of the Speyside single malts? She is looking to establish a premise AFTER collecting data. She is not making a grand statement like “all Speyside single malts have heather” after tasting just one. Inferring generalities from specifics then running experiments to prove the theory is inductive reasoning.

Feature Detection

Data analytics pursues feature detection to find what will predict an outcome. Features are like column headers in a spreadsheet. Lenders inspect aspects of home mortgage applications to see what attribute or combination of attributes will shine a light on those who are worthy. This process is not unlike whiskey tasting.

Consider tasting wheels. They are just circular spreadsheets. Each spirit can be ranked in intensity to those specific flavors. Those features originate from grain selection, fermentation, distillation, barrel type and aging. These are puzzle pieces that whiskey lovers adore assembling in their minds. They are getting to know the whiskey like characters in a novel.

Classification

One of the short cuts to dealing with large populations is to bucketize them into groups. The Boomer, Gen-Xer and Millennial labels are nothing more than a classification exercise based on birth years. We make generalizations about each group’s interests. Consider that Red Bull has the Millennials in their cross hairs while Metamucil aims at the gray hairs. This classification technique works well with whiskey comparisons too. Bourbon by law needs to have 51 percent corn. So in a blind test comparing spirits from different grains, *look* for the candy corn aroma. It’s a signature to this class of whiskey.

Propensity

Propensity is a fancy term to describe what might likely happen. It’s how data analytics deal with the unknown. We see a drop in the price of oil and the propensity for Houstonians to leave the family cell phone plan rises. The same principle underpins whiskey when it comes to food pairing. Chicken piccata, an Italian dish served with a lemon and caper sauce, is likely to go well with a Rye. Why? Because the rye grain has lemon on the nose. So the propensity for the match is high.

Iteration

Data analytics professionals, just like master distillers, like to experiment before they lock down on a data model. They have ideas and tweak as they go. Internet properties like Facebook play with the shade of blue to see which gets the most clicks. So why not expect that with whiskey? Whiskey blenders like Compass Box continue to push the envelope for their blended malts. They were famously cornered by the Scottish Whiskey Association for a rather unconventional aging process in the original recipe of Spice Tree.

Establishing a Baseline (Supervised Learning)

We know what is normal for blood pressure because doctors have measured this vital sign for years. More than that, they have correlated both positive and negative outcomes to the data. They know a patient is high risk because of the histories of hundreds of thousands of patients. The role of data analytics is to determine what is “normal” based on a given data set. However, normal becomes useful when we know the outcome of a certain event too. In the land of data analytics, when we establish a baseline with known outcomes and ask algorithms to pick out things that predict the future, we are engaged in supervised learning.

The act of building a baseline for whiskey tasting comes from personal experience. Blind tasting after blind tasting helps the taster single out the single malts from the blends. A corn mashbill from rye or barley. Secondary casking versus single. The more a whiskey taster experiences, the bigger the sample set, the broader the foundation, the more the taster knows. This foundational knowledge helps whiskey tourists know when they have left the paved road and are launched on an adventure.

Anomaly Detection

Anomalies get a bad rap. That is until you understand that all parents want their kids to be normal, but never average. Being above average earns gold medals on the downhill and early acceptance to that hard-to-get-into college. This is not normal. Seeking anomalies is the job of talent scouts. It is also core to data analytics because it’s something from which we learn. It might be the use of a product the designer never expected. Ask the Pfizer about its original intent for Viagra.

The world of whiskey tasting presents a similar opportunity. Greenspot Irish Whiskey has green apple all over it. Westland Single Malt tastes like chocolate. And Hudson Four Grain has a barnyard quality to it. You can almost hear the sheep. When you hang notes in the air like Pavarotti, you get noticed. Anomaly detection is a different kind of appreciation. Whiskey aficionados aim for this.

Normalization

Eighty percent of a data scientist’s time is spent wrangling data — filling in the missing elements in a table so the columns and rows are ready to be analyzed. It is hard to draw conclusions when the artifacts are missing. And this same rigor is pursed by the whiskey trade. Evaluating whiskey must be done if and only if the spirits are served in the same way and the same time.

We know that wine oxidizes in the glass. An angry glass of cabernet becomes approachable after an hour once it breaths. Time also plays into whiskey, except oxygen is not the factor. When alcohol evaporates from the glass precious olfactory volatiles escape with them. A whiskey freshly poured might be feisty at five minutes and friendly at 50. So it’s important that we treat data and whiskey with the same level of consistency: same glass shape, same pour size, same time out of the bottle.

Enrichment

Business data only gets better when we add diverse data types like geo-spatial, event-related or weather data to it. When a Texan shops online at happy hour on Cinco de Mayo and the cart was abandoned, there should be no surprise. Oh, it was sunny that day. This context-driven awareness adds a deeper understanding. Modern analytics is all about enriching structured data with unstructured to gain a better experience.

Likewise distillers are aging whiskey in second or third casks. They take a completed product and finish it off in wine, sherry, port, rum or Sauternes casks. Or it might take the form of the *blenders’ art* like Johnny Walker, Dimple or Compass Box. These distillers ladder up the experience by marrying whiskeys from different places. Hudson Four Grain ages the same spirit in three different sizes of barrels each with a different char to get that exact expression.

Shaped by Taxation

Fans of history have no trouble remembering when Alexander Hamilton rode to Western Pennsylvania in 1794 to help his boss lay down the Whiskey Rebellion. This was before Mr. Hamilton was a Broadway sensation. He was our first secretary of the treasury and was hungry to pay down the national debt with a whiskey tax . Our friends in Scotland struggled with same issue in the mid-18th century where distillers were taxed based on still size and not production. In 1787 taxation was the tipping point in the split between the Lowlands and the Highlands. Religion, language and affiliation with England may have proper historians talking, but a true Gaelic Highlander knew the real argument was over whether blended whiskey was a *real* whiskey or swill to appease the English.

Likewise analytics in the English-speaking world found its voice because of taxation. First appearing in Britain in Roman times, the practice became a consistent effort in 1801.  Its mission was for allocating precious resources. And since counting people one by one takes longer than a single decade, statistics found its place in economics.

So the next time you raise a Glencairn glass of the *water of life*, just remember that as you ponder notes of heather, sea air and the smell of warm biscuits, you might actually be thinking like a data scientist.

Cheers!

For additional information on Dell EMC data analytics solutions, please contact us at [email protected] and follow us on Twitter at @DellEMCbigdata.