On this episode of “Next Round,” host Zach Geballe chats with Jennifer Estevez, co-founder of Palate Club, to discuss how a custom algorithm and sommelier experience intertwine in her app. Using her own sommelier background, Estevez collaborated with data scientists to match consumers with the right wines for them.
She explains that the algorithm was created to help consumers understand integral components in wine such as acid, tannin, and sweetness. Finally, listeners will learn how Estevez sources her wines — partnering with producers from various countries and never using bulk wines.
Tune in and visit https://www.palateclub.com/ to learn more.
Or Check out the Conversation Here
Zach Geballe: From Seattle, Washington, I’m Zach Geballe and this is a “VinePair Podcast” “Next Round” conversation. We’re bringing you these conversations in between our regular podcasts so we can explore a range of issues and stories in the drinks world. Today, I’m speaking with Jennifer Estevez, the co-founder of Palate Club. Jennifer, thanks so much for your time.
Jennifer Estevez: Thanks so much. Excited to be here.
Z: Yeah, we’re glad to have you. Let’s start with just a little bit of background about you. How did you get involved in wine, and what were you doing professionally before Palate Club?
J: Yeah, absolutely. I got involved in wine when I was 21 years old. I ended up working at a Ritz Carlton as my summer job and had a delicious glass of white Burgundy for a wine training that our wine director was hosting. It was just delicious. It was a Pierre-Yves Colin-Morey Saint-Aubin. I still remember it, and it made me fall in love with wine and the potential of wine. I loved it so much that I ended up moving from where I was, which is Arizona at the time, and just picking up with $450 and a suitcase and heading out to San Francisco to try to dig into the wine world a little bit further. Once I got to San Francisco, I worked around the city at many different wine bars and Michelin Star restaurants. One of the most memorable was RN74 and Michael Mina. I just dug in from there, continued my studies, and got my advanced sommelier certificate around the time that I also began consulting on restaurants and hospitality programs. My genesis into Palate Club was really synonymous with my genesis into my company OMvino, which I realized around that time there were just a lot of needs that were specific to the food and beverage industry that weren’t being met. Marketing communications in the food and beverage industry via OMvino were definitely something that was a need, and they needed people who understood food and wine. Similarly, in the online wine club space, there was just a gap between that true consumer connection and these online wine club applications that were over-promising and under-delivering.
Z: That’s actually a great lead-in to where I wanted to go next. We have certainly on the podcast talked about wine clubs in the past. Anyone who likes wine and is on social media gets lots of ads for them. I think one thing that is interesting to talk about is, with some of those other clubs, they claim to be able to match you with the perfect wines and say to leave everything to them. Let’s start, first of all, with how Palate Club works. How does the matching component of it work from the customer standpoint?
J: Great question. With a lot of these other wine clubs, they try to either do one of two things. They try to define the consumer profile, and they try to define the wine profile. There are really no online wine clubs out there besides Palate Club that I’m aware of that are doing both and have the data to back it up. What makes Palate Club really special is it was created by wine professionals along with professional data scientists. Essentially, what we did was create a custom algorithm in IP where we could input up to 200 scent and taste characteristics about each bottle of wine and to the back end of Palate Club. So the wine profiles are very detailed and we put a lot of time and thought into what characteristics of the wine matter for the consumer palate. Then, the consumer portion of the experience is very easy. They rate things on a scale of one to five, so we just put time into analyzing the data that came up and understanding what trends were occurring in these different patterns that we were seeing — so we can more accurately adjust the algorithm to really deliver a satisfying consumer experience, just as if a sommelier was selecting these wines for these people. We also create great wines and we source great wines for Palate Club, which I think between those two things is the really big differentiator between us and our competition.
Z: Sure. So let’s start with one part of this that I’m particularly interested in, which is this idea of all these different components — smell, taste, etc. — that you and your fellow wine experts at the club have assessed these wines on. When you first started out, how did you figure out which ones really mattered?
J: When we were going through the process of creating the systems in the back end of the app, we really had to think about what do consumers really look for, and what parts of the sensory experience are important to them? When we’re looking at things, it’s not as important in the wine, for example, if you get a note of blackberry. We do put that in as a descriptor but where does that factor into the overall relevance of this bottle profile? It’s definitely lower on the list of priorities, but things such as complexity, acid, alcohol, tannin, structural components of the wine that every wine has our higher priority in terms of determining what people really want and what people really are paying attention to. Often, why we feel that Palate Club is so relevant is that people don’t understand what these really important integral components mean. Many times when I worked in a restaurant as a sommelier, people would come up to me and they’d say, “I want a sweet, dry wine.” I would think “OK, we understand that those two things do not work together, but how can we figure out what you want?” We try to figure out and fill in around that as sommeliers. Now from a data perspective, what these people really want based on these ratings. Again, the things that all wines have — tactile sensations and objective qualities such as alcohol, acid, tannin, and structure — all of those things are relevant. One thing that we also found was complexity. Complexity is a really important weight in the conversation as well. Defining that complexity is really important to have sommeliers who understand the bandwidth and the spectrum of complexity to put it in properly from a data perspective.
Z: Very cool. OK, so we’re going to diverge a little bit from my planned line of questioning because now I’m just curious about this. I’ve long believed as a sommelier and wine professional that the way people are taught to learn about wine overly emphasizes flavors and aromas. We’ve all read the tasting notes that are preposterously specific about the exact kind of wild mountain strawberry that you might find in a wine. What really drives people’s preferences is, as you described, much more the structural elements of the wine that most people don’t really learn to identify. Does that sound right? Do you find that those structural elements are a much better predictor of whether a given customer will enjoy wine or not than mulberry, or whatever?
J: One hundred percent. I think that those arbitrary tasting notes are like arbitrary wine critic scores because they’re very subjective, just based on people’s personal preferences and descriptions of the wine. That’s why I think that when you look at apps such as Vivino, it is very interesting from a technological standpoint. They crowdsource the ratings from the people who are creating these ratings on the app using keywords. How accurate is that in benchmarking what the actual wine qualities are? It’s not. When you’re looking at things that are really important to the wine, again, the objective qualities such as acid, alcohol, tannin, and sweetness are very important, along with the main characteristics. We basically create different overarching categories that we also found were very important that were not just objective characteristics. If the wine was more dominant towards earthy characteristics, we could basically tell that by a percentage rating when we rate the wines. That would be something we’d gravitate people towards who have earthy palates. For example, when I’m talking about earth, let’s say this wine has mushroom, umami, savory, meaty characteristics. Those are some of the primary dominating descriptors. Instead of just picking out one of those singular things and saying this blackberry or this mushroom note is the most important thing, we look at these percentage values and see where that arc goes and see that this is the dominant characteristic of the wine. What’s the dominant characteristic of these people’s palates such as the blackberries, the chocolates, the blueberries? Does that stand out more? Do we gravitate them more towards fruit and sweet flavors, and not compound sensations? That’s how we clump all of these different profiles. It’s just creating these generalities between scents and taste that create different subcategories. Also, using those objective characteristics that are inarguably there for every wine. Does that make sense?
Z: It does, yeah. Not being a data scientist, it only kind of makes sense, but the wine part of it makes sense to me. I want to ask a couple of more questions along these lines and then maybe we’ll talk about some other elements of Palate Club. I know we spoke prior to this about a little bit of the program as well. One thing that you mentioned then that stuck with me is this idea that, through data that you collect from your customers, you can start to see that there are these clusters of wine drinkers that have similar tastes. I was hoping you could talk a little bit about how many of those categories there are? Are they roughly equal in size? Is there one big category and a bunch of subcategories? How do you see the wine-drinking public breaking out?
J: Yeah, we’ve divided it into eight main categories and then people can also take a look at their detailed profiles and see what characteristics come up from a percentage standpoint and what those characteristics mean in the app. For the general categories, we give them fun names like the Sicilian and the Sexy Beast. We give people these categories to think of themselves in terms of their wine palates, but we also give them more context to what it means. If you go in the app, you can click on all of these different descriptors and tactile sensations and ask, “What does this mean? What does it mean that I like medium-plus acid? How is that relevant, and what wines does this correlate to?” We’ll have some Q&A there. I went in the app myself in the early days of the Palate Club and wrote in different descriptors, and we’ve revised them many times over. To answer the other part of your question, there’s not really a trend. That’s the thing that people really get stuck on. I think often, they don’t know what they like and why. People’s tastes are so diverse. We’re often so much more open to different types of wine than we think. It’s really individually based on what people’s preferences are that they may not even understand. The world of wine is so vast.
Z: Yeah, that’s a fascinating point that I’m curious to hear more about. One thing that I think you and I probably have both experienced in our time working the floors of restaurants is, you go to a table and you get a sense of what people liked to drink in the past. You have to make a bit of a decision as a sommelier or a person recommending the wine. Do I want to say “Oh, do you like Napa Cab? That’s what you’ve said to me. Great, here are the seven Napa Cabs I have. Which one is in your price range? I could also talk a little bit about them and we’ll differentiate.” Or, you might also have to make the decision of, “I’m going to try and see if I can’t open something for you that you might not be familiar with.” Maybe it’s a different Cabernet from a different part of the world or a different variety entirely. Part of the hard-to-define skill of a good sommelier is knowing when and where to make those suggestions versus when to get what the person wants. Then, also kind of the actual accuracy of those suggestions that go outside of someone’s existing flavor profile. How does Palate Club balance giving people more of what they’ve already shown they liked, versus potentially allowing them to explore something that might be a little less familiar?
J: Yeah, absolutely. That’s something that I definitely think we tried to create and we’ve had really great feedback from our current customers on it. What we have is a randomization option, and the option isn’t totally random. We’re not just throwing bottles out there in your club shipments hoping they stick. It basically creates deviations from your profile. Each time when we have a deviation, we’ll know that these people like X, Y, Z. Now, let’s say they like wines that have earthy high acid, like structured wines. Let’s say they are just Barolo lovers. Now, what can we think about that ties into similar characteristics as Barolo and at least a few of these categories? OK, maybe we’re going to try some Dolcetto or maybe we’re going to try something structurally similar, but from a different region. Whatever it is, we’ll latch on to just a few of those characteristics, and then the system auto-generates recommendations based on that. For example, if you like Napa Cab, maybe you’re going to love some Aglianico. Whatever it is, just something fun and unique from a different area but that has similar structural and taste profiles. It can get more or less similar to what your profile is like, based on the level of randomization you select. That’s an option in the order section when you’re going to your checkout.
Z: Let’s switch things up and talk about the sourcing of wine. Obviously, technology is a big part of this and having the ability to offer quality suggestions is good. At the same time, you also have to have quality wines. How have you connected with suppliers, producers, etc., and what does the catalog of wine such as this look like?
J: Yeah, absolutely. For Palate Club, we have traveled literally all over the world and just met with producers from different countries to try to find really unique selections. We work with small distributors here in the States, and we just try to find wines that are unique and offer value and add a diversity of price points. We favor sustainable and organic, artisan wines whenever possible. Biodynamic as well. That’s really something that we just don’t want to compromise. Many of these clubs just source bulk juice and put a label on it, and it’s just not offering quality. Their user retention is lower because of that. That’s something that we’ve done from the beginning. It was really fun and a pleasure to do. I traveled to Australia, France, Italy, Germany, South America, and South Africa. Basically, with one of the other co-founders of Palate Club, we traveled everywhere just to find these different producers, to build these relationships, and to just champion what wine is really about, which is connection and quality product. That connection and a quality product are not just in the wines, but also for the consumers. It’s really important for us to deliver that whole full-circle experience.
Z: I think another thing that can be potentially challenging for wine clubs online is, how do you establish categories as far as pricing goes? People are going to have different budgets. Is it a narrow price range, or is there a wide range of price points for wines through Palate Club? How does that work?
J: Yeah, absolutely. With Palate Club, beforehand, we were doing a tiered system, so we had just different price point tiers. However, we’re moving away from that. Essentially, people can just set what their budget is and how many wines they’d like to get. Then, we’ll generate wines that are appropriate within that category for their shipment and with their taste in mind. It’s really customizable, and it is really flexible. Again, as much as humanly possible, we want this to be a human experience delivered via an app. We want it to be customized. We want it to be hospitable. We don’t want to have to call five different numbers to cancel your subscription, and we want you to get one that you want. That’s really what we’ve tried to do with those different features.
Z: Gotcha. So you mentioned that the main way in which Palate Club generates a profile for someone is that the subscriber rates wines that they received on a one-to-five scale. Then, you look at that feedback and look at patterns in there. A), how did that start for someone? Obviously, with zero wines that they’re picking blind. And B), how long does it take before you get an accurate picture of an individual person’s preferences?
J: With people who are coming on board via the tasting kit — we have a blind tasting kit that we start people off, with four half-bottles, either red or white. They’re all very contrastingly different, and we use them as bookends for defining your palate. Defining the spectrum, the high/low end of all these different qualities that need to be considered when thinking about what a person’s palate gravitates towards. That is the best way to do it and the fastest way to get a really accurate palate screenshot and just a palate profile built out. The palate profile continues to change and refine itself over time the more wine you drink. It’s really fun, and you get to unlock different secrets about your palate in the app as you go. Then, if people come on through the quiz option, we have a regular quiz and then a connoisseur quiz, which we’re developing. Essentially with these quizzes, again, there was just a missing ingredient for many of these other companies when they’re tying in these quizzes to the wines that they have in their selection. I took a quiz the other day where I was just testing it out. I’m always looking at the competitive set, right? I took this quiz, and I was just checking boxes. Yes, I like chocolate. Yes, I like coffee or whatever, and wondering if it ties into anything factual. I did not select anything that was overly sweet, that was overly fruity. Somehow at the end of this quiz, I got Moscato d’Asti and sparkling Riesling. I was very confused. We do think that the general palate, the general blind tasting is more accurate. Yet, with the quiz, we tie in the questions that we ask about the foods, ingredients, and the things that we have in there to actual data points. If I’m asking you if you like bitter dark chocolate, that means for me that you probably like tannin and more astringent, bigger wines. It’s just something that, again, has to be correlated to more of a data point rather than just a consumer preference when consumers don’t really know. I hope that answers your question.
Z: Well, how many of these data points do you need before you get a relatively accurate portrait of the consumer?
J: It really depends. It depends on how the data points cluster. It’s good when you’re taking in data points to get 10 or 12 at least that are outstanding, important data points. However, it’s really subjective. It is based on what the user’s profile gravitates towards, and it’s harder to benchmark that with the quiz itself. With the profile, when you’re building it from the wine perspective, you just need to taste four wines for it to start generating a really accurate profile.
Z: Excellent. Well, Jennifer, this has been super interesting. For people who are interested, where do they go to find out more?
J: You can go to palateclub.com.
Z: Nice and simple. Well, thank you so much for your time. I really appreciated learning a little more about this. I look forward to seeing as the data evolves further. It’s always interesting. We at VinePair are super fascinated by data and how we learn about ourselves and each other through it. It’s cool to see that people are doing that in the wine club space as well.
J: Yeah, thank you.
Thanks so much for listening to the “VinePair Podcast.” If you love this show as much as we love making it, then please give us a rating or review on iTunes, Spotify, Stitcher, or wherever it is you get your podcasts. It really helps everyone else discover the show.
Now for the credits, VinePair is produced and recorded in New York City and in Seattle, Wash., by myself and Zach Geballe, who does all the editing and loves to get the credit. Also, I would love to give a special shout-out to my VinePair co-founder, Josh Malin, for helping make all this possible and also to Keith Beavers, VinePair’s tastings director who is additionally a producer on the show. I also want to, of course, thank every other member of the VinePair team who are instrumental in all of the ideas that go into making the show every week. Thanks so much for listening, and we’ll see you again.
Ed. note: This episode has been edited for length and clarity.