Taste Bias and Tech Sheets: Case Studies from the Wine + Data Community
Image Credit: Trackmind
“There should be a support group for data people in the wine industry.”
I said this out loud for the first time shortly after we launched Enolytics. At the time, it was an off-the-cuff response and I was pretty much joking as I commiserated with wine people who felt isolated in their desire to bring the power of data analysis to their work.
Here’s the thing, though. The idea of a supportive community for wine + data has never gone away.
So I’ve been keeping a list of people I encounter who are doing really interesting things with wine + data, and I actively look for ways to highlight that work. The list has expanded to also include people who want to do interesting things with wine + data but don’t know quite yet how that happens.
That’s the part of the list that’s growing quickly.
The more people they can see in our community who are already doing cool, relevant and useful projects with data, the more they’ll be encouraged and inspired to do their own.
Which brings me to two fresh highlights of wine + data projects, and the people already doing them.
A few weeks ago I wrote about our work with NLP, or Natural Language Processing, and that very day I heard from several people in wine who have also done NLP projects. Let me highlight two of them this week, both having to do with the language of tasting notes.
Data Analysis for Marketing and Operations
Melissa Coventry is a data scientist who looked at reviews from the Wine Enthusiast to create a machine learning model that predicts the chance of receiving a 90+ point score from the magazine. In addition, she looked at reviewers’ taste bias within scoring. She originally thought the studies would focus on marketing opportunities, but quickly found that it also showed operational ones as well. Melissa's deep dives into wine data often reveal additional results as her winemaking and operational background lends insights into data analytics that other data scientists might miss.
Winery Messages Embedded in Tech Sheets
John Egan, export manager at Rombauer Vineyards, applied NLP to publicly available tech sheets from historic Napa producers, in order to understand how a winery’s message is embedded in the language. His analysis included winemaker quotes, food and wine pairing suggestions, wine descriptors, and a predominance of words dedicated to vineyards and viticulture, underscoring perhaps the value placed land ownership and investment.
Rombauer was one of the wineries that John studied. He was able to identify – and presumably correct – missed opportunities particularly in relation to that winery’s competitive set.
Neither Melissa nor John hold a standardized job title or belong to a department that’s specific to data. But just because the work of data + wine is still being defined doesn’t mean that it isn’t happening. In fact it’s exactly that – a fresh and rational, methodical approach – that makes their work so valuable for an industry like ours that is so heavily steeped in tradition.
If you know of more people who ought to belong to this emerging “support group,” or if you are one yourself, please let me know. I look forward to highlighting more of this community’s work in the near future.
Final Note: I appreciate those of you who also reached out about last week’s post on racial bias in data. Let me share these three resources that have been brought to my attention:
97 Things about Ethics Everyone in Data Should Know, a forthcoming book by internationally recognized thought leader Bill Franks
The Center for Humane Technology, started by the Tristan Harris, who was formerly Google’s Design Ethicist
Women in Data, a global organization that started in Sacramento, whose mission is to increase diversity in data careers
I hope you’ll have a look at all three resources in order to continue the conversation.
Thank you, as always, for reading –
Cathy