Happy 2022 (it’s not too late to say that, right?)! I’ve just returned from two weeks traveling around Spain, and while I did not have a positive COVID test the day before my flight per US reentry requirements, Omicron wouldn’t let me have so much fun without reminding me that balance must be achieved in the universe, so I am now writing to you from COVID isolation on my COVID couch.1
I live in California’s Central Coast area, halfway between LA and San Francisco, and it is a great place to live if you like wine, which I do. Since moving here nearly two years ago, I’ve gotten an immersive look into the business of wine production. The more I learn, the more impressed I am; there are countless variables that winemakers need to understand and control for, and I can’t begin to fathom the amount of expertise they must have in order to create such delicious product year after year. Instead, I’ve been satisfied by experiencing wine through various (pandemic-era) events in the northern part of San Luis Obispo County, one sip at a time.
It was at one such event last October that I learned more about Perinet, a small-production winery located in the Priorat region of Spain that specializes in Grenache and Cariñena grapes. After dinner, where each course was paired with a different Perinet wine (the perfect time to make the best decisions), my friend and I agreed that we had to go to Spain to visit the winery for ourselves.
After the paella-packed Valencia leg of our trip, my friend and I rented a car and headed up the narrow, windy mountain roads to reach Perinet. Our amazing host gave us a private tour that began with a detailed history and geography lesson of the region.2 The unique slate soil of Priorat means that the grapes have a different terroir than those grown in more common limestone or clay soil throughout the rest of Spain, and tiny microclimates throughout the 76 sq mi region result in completely different flavor profiles between vines that are planted only a mile away from each other.
We then moved on to the fermentation and barrel rooms and of course, were able to sip and compare along the way. This particular winemaker chose not to blend until after fermentation and aging were complete so that he could seperate and sample all of the different variables that would affect that vintage. In the large underground cave, row after row of small barrels represented dozens of control and test samples.
For example, grapes from a north-facing side of the slope tend to need longer on the vine than grapes from sunnier south-facing vines, and the longer the winemaker chooses to leave the grapes on the vine, the more sugar they develop. Once harvested and fermented, the winemaker can then decide if they should be aged in oak or not, what kind of oak,3 and for how long. Even the toasting level of the oak barrels (“high, medium or light toast”) and whether or not he uses a barrel with a toasted “head” (the endcaps of the barrel) were variables to consider. Through a mix of descriptions and sips from different barrels, we were able to start putting together an understanding of the outcomes of options the winemaker had at his disposal.
My friend and I dove deep with questions for the host on the details of his process, and it was halfway through the tour (and after several sips) that we learned he had studied chemistry before becoming involved in wine production. Once he discovered that we were both scientists as well, we further geeked out on his experimentation process. His eyes lit up.
“I have to show you this. Same grapes, same vintage, both American Oak barrels with the same toasting, but each barrel has been made with a different tree grown in a different area with different soil. No one else is testing for the terroir of the oak itself!”
We tasted from each of the two oak barrels. Maybe there was a difference, or maybe by that point we had had too much wine to tell.
I had to ask, looking around the room at all of the relatively small barrels, all containing unique expressions of wine, “How do you decide whether you want to keep these separate when you bottle the wine, and do you have to make different labels for bottles given every variation you have tried?”
“Eh, most of these for example will be blended together as a single vineyard Grenache, otherwise I wouldn’t be able to make money.”
An obvious fact suddenly hit me. All of this information we just learned about will be lost in the blending when the wine is put into bottles. Information like the toasting, the side of the mountain the grapes were grown on, information that an amateur non-sommelier like myself would never know just by looking at the bottle’s label. And even if I was able to recall every detail of our tour, and find out which barrels ended up in which bottles of Perinet in a year or two’s time, my knowledge would mean nothing for the following year’s vintage when the winemaker would make different choices depending on the weather that season.
It made me a little sad, at first. As scientists, we want to know. We see the world around us and want to make sense of it, and understand all of the constituent parts that make up the whole of reality. If I try a different bottle of wine from a different winery, it may be fantastic, but how would I know why I liked it so much? How would I know how to decide what other wines I might like as well?
At the end of the day, the answer is pretty basic. As the wine consumer, I don’t really need to know. I can appreciate a wine aggregated from various samples for what it has become and how I want to enjoy it. I can trust the Perinet winemaker to be the expert, learning and sampling year after year, and when I want to know if a particular vintage would go well with a particular cheese, say, I can just shoot him an email and ask.
Shifting into the metaphor here, it made me remember one of the most difficult aspects of moving from science in academia to the tech industry: learning to let go of that deeply ingraned need to discuss all of the different investigations that go into an analysis when presenting the final results to stakeholders.
In academia, our customers (scientific journals) wanted us to prove that we had considered and controlled for every variable; the difficulty was in determining whether or not we had thought of them all. In the tech industry, our customers (product owners and business leaders) only want to know the key takeaway and what to do next. The most difficult part instead becomes knowing when to aggregate information, especially after creating all of that information entropy to do the analysis in the first place. A well-communicated result, like a good bottle of wine, leaves the customer satisfied. A savvy consumer may pick up on the underlying complexity and ask to understand more, but that’s really up to them.
As Data Science managers, we can help our team find the right “expression” of the analysis to bottle up for our stakeholders. By setting up a little intentional structure around how results are presented, we can take the cognitive load off of those who are responsible for doing the analysis and help them accept that most of the work they have done will never be directly seen by their stakeholders.
I’ve used the concept of a “one-pager” template for analysis results that highlights the key takeaway right at the top in bold and only describes the high-level context (ie: problem statement) and actionable recommendations, keeping detailed analysis hidden away in other artifacts, much in the same way the label on a bottle of wine denotes only the most important information. Similarly, slide decks that present results can be ordered with the results and recommendations first, and almost all of the analysis details either in backup slides or not at all.
If that is still too hard to swallow, because someone should know about all of that hard work, managers (and peers, for that matter) can play the intermediate role of the academic journal, vetting the methods and giving an outlet to data scientists who want to make sure they feel that their analysis has been thoroughly peer-reviewed.
I could continue to venture deeper into this metaphor, especially in regards to aggregation choices in the data warehouse, but if you haven’t guessed already, dear Reader, this post was mostly meant as an outlet to geek out on wine nerd stuff and get back into the swing of things. Until next time, here are some summarizing tips:
If you are a Data Scientist, acknowledge that you might have a subconscious tendency to want to show off your work over focusing on the results presentation as an artifact to be enjoyed in and of itself.
If you are a manager, help be an editor as someone with a fresh perspective when reading analysis results from your team.
If you are a consumer of data analyses, then feel free to appreciate the complex subtleties that went into creating the results presented to you.
And if you just like wine, well, cheers to you.
<PSA> I’m fine btw, thank you vaccine and booster. The symptoms were so mild I wouldn’t have even known I was contagious if I hadn’t been rigorous about testing, so you know, get tested often, especially before you see susceptible loved ones. If you happen to live in the US, here’s a website where you can get free tests starting Jan 19. </psa>
TIL: In the 19th century, an aphid-like bug called phylloxera ravaged European vines and nearly wiped out all wine production across the continent. To this day, every vine in Europe is grafted on North American rootstock, which was resistant to the bug that had originated in America in the first place. I mean, what?! 🤯
My favorite anecdote of the tour was how our Spanish host described the difference between the effect of French oak vs American oak barrels on the flavor profile of the wine:
“American oak, well you know, Americans are so easy to get to know. They are fun. It’s a party. You don’t have to wait long to become friends. The American Oak offers a sweetness to the wine and does not have to be aged as long to get the flavor.
“But the French, they do not immediately become friends. They take a long time to get to know you. And you need to have a strong personality to stand up to the strong personality of the French. And to use French Oak barrels, the wines must also be strong, like Cabernet Sauvignon, and they must be aged for a much longer time to get to know the subtle flavors from the oak.”