Champions of Data + AI

Data leaders powering data-driven innovation

EPISODE 3

Improving Human Experiences
With Data

As data leaders design, develop and iterate on new data and AI products, it’s critical to maintain the human connection — especially in consumer-focused businesses. Jon Francis, Chief Analytics Officer of Starbucks, joins us to share insights from his experience leading data teams across some of the most iconic brands, including Starbucks, Nike, Amazon and Microsoft. Tune in to hear Jon share the challenges and responsibilities data leaders face in maintaining a healthy balance between using data ethically and driving new insights.

Jon Francis
Chief Analytics Officer
Starbucks
Jon Francis is the Chief Analytics Officer at Starbucks. He is responsible for analytic and data strategy, data science, and market research for the enterprise – with specific support for customer, marketing, digital, pricing, product, partner, ops services, and store.

Jon joined Starbucks in November 2015 supporting data science for the Starbucks Rewards program. Over the past three years, his remit had increased to include analytics, data and measurement sciences for the entire marketing organization. During this time, Jon drove the data science and accompanying cloud analytic technology strategy for the marketing organization. Specific accomplishments included standing up machine learning to support personalization and targeted marketing, deployment of a recommendation engine leveraged within the Starbucks App and Google Voice ordering, and establishing next-generation financial measurement capabilities for Starbucks’ paid media investment.

Prior to joining Starbucks, Jon was the Director of Consumer Data Science and Technology for Nike, where he drove the transformation of analytics and big data, cloud-based architecture for the Global Consumer Knowledge and E-Commerce organizations.

Before joining Nike, Jon applied his analytic and data science expertise in several industries from marketing, digital platforms, retail (e-commerce and bricks-and-mortar), healthcare and telecommunications at major brands such as Amazon, Microsoft, Expedia and T-Mobile.

Jon has an M.S. in statistics from Oregon State University and a B.A. in math from St. Olaf College. He lives in Seattle with his wife and two daughters. He enjoys tennis, music and theater, and traveling. Jon’s favorite Starbucks beverage is a Venti Caffé Americano.

Read Interview

Speaker 1 (00:00):
Welcome to champions of data and AI brought to you by Databricks.

Speaker 1 (00:10):
In each episode, we salute Champions of Data and AI, the change agents who are shaking up the status quo, these mavericks are rethinking how data and AI can enhance the human experience. We’ll dive into their challenges and celebrate their successes all while getting to know these leaders a little more personally.

Chris D’Agostino (00:33):
Welcome to the Champions of Data and AI. I’m your host, Chris D’Agostino as data leaders, design develop and iterate on new data and AI products. It’s critical to maintain the human connection to the products, services, and brand that you offer, especially in consumer focused businesses. In this episode, Jon Francis, Chief Analytics Officer of Starbucks joins us to share insights from his experience leading and being a member of data teams across some of the most iconic brands. Jon shares with us, the challenges and responsibilities, data leaders face in maintaining a healthy balance between the ethical use of data and driving new insights. He also talks about the role, the data platform plays as an enabler to improving consumer engagement and revenue growth, Jon welcome. And it’s great to have you with us today. Got a couple of just sort of icebreaker speed questions for you in terms of, so it’s Thursday afternoon. What what was your last caffeinated beverage? And do you have like a go-to beverage or do you just mix it up?

Jon Francis (01:39):
I’m pretty boring, man. I I go with the venti Americano every day. For awhile I was dipping into Irish Cream Cold Brew, if you haven’t tried it except well too much sugar for me. So I really try to be careful about my sugar, you know, here we are with new years. So really it’s the venti Americano.

Chris D’Agostino (02:00):
Awesome. And then shifting gears to your time at Nike, I’ve a friend at Nike and he talks about just how he gets inundated with products from, from as part of his role. And, and so he just has more pairs of shoes than he, he knows what to do with, what about you a favorite attire, favorite types of shoes that you.

Jon Francis (02:19):
I mean, I’m a, I’m an aging sneaker head and so I have the same problem. There’s way too many shoes in my closet. My favorite is probably air max nineties. I just liked the style to get retro lifestyle. It’s pretty cool.

Chris D’Agostino (02:35):
All right. And final one, you know, you were at Amazon really early on, right. And at a statistician, if I remember correctly and you know, was there a time where you realize like, Hey, this online shopping and Amazon in particular with their approach to it, that this was going to be really big?

Jon Francis (02:54):
Yeah, I think I ironically it was, I was sent to the center in Atlanta in the year 2000. This was back when we couldn’t get enough seasonal workers to go pack boxes in the store. And I was in the distribution center in McDonough, Georgia, and I was standing there in absolute awe of the sheer volume of product and just the breadth of inventory, everything that was sitting on our shelves. And this was 20 years ago. It’s probably exponentially different now. And just the idea that, well, how do you create capabilities and seamless experiences for customers to be able to find and interact with these products? And so to me, it was actually standing in that distribution center and just understanding the breadth of product that was out there and, you know, the way that digital can create seamless experiences to unlock that. I think that’s when I first saw it.

Chris D’Agostino (03:52):
So you’ve got a really impressive career. I mean, you’ve been in the Seattle area for a while. You’ve been working with data and AI for, for quite some time and at organizations like Starbucks, Nike, Microsoft, Amazon, Expedia. So if there seems to be a bit of a theme of a business to consumer play here, as well as obviously data. And so just wanted to better understand, you know, are these intentional in terms of the B2C play as well as going broad across a bunch of verticals and industries or, or have you really just kind of followed the data as part of your

Jon Francis (04:28):
Yeah, no, I think it’s a great question. I, I think what’s really driven me first and foremost is a passionate around customer and marketing applications and digital capabilities. And so you see that threaded through a lot of where, where I’ve gone from a career perspective. I think as of late, probably in the last 10 years, I think the thing that inspires me most though it’s less about the data domain. It’s actually going into organizations that have built amazing brands almost in the absence of data. So if you take Nike and Starbucks has two great examples, incredible leaders who were visionaries and had a strong gut-based instinct around whether it was Nike or Starbucks and to go into a company like that, and actually be part of a transformation and go from what historically as it may maybe more gut-based to more of a data-driven organization and really help lead that transformation. I think that’s what really has me passionate now. And it hasn’t been for 10 years or so.

Chris D’Agostino (05:35):
So John, you said at the, at sort of the beginning that you like going into organizations that have a really well-established brand and you know, maybe haven’t used data as well as they could and, and come in and really assist in that data transformation initiative. I would imagine with companies that have that brand recognition, there’s a couple of attributes that go along with it, which is, you know, obviously have scale to have that kind of brand recognition. They likely have the resources to invest in these massive transformation initiatives and they collect a lot of data. So along those lines, you know, are those the types of things that you look for for an opportunity and when you’ve done it, you know, tell us a little bit about how you get executive buy-in and, and form your data science teams and your ML efforts, and really get this thing going and, you know, establish some traction

Jon Francis (06:27):
The biggest need. And if I just reflect back on, you know, the last 20 years or so, especially in these organizations where they may not be as forward thinking on their utilization data, it’s actually more than anything, more than the resources, more than the data it’s about cultural transformation and the change management of how you take organizations on a journey and really to think differently about how they do their work and where there are opportunities based on the use of, you know, whether it’s research, whether it’s machine learning capabilities, whether it’s statistical inference. I think that’s probably the biggest part of it. And I guess to, you know, how so, how I’ve led through that cultural transformation, there’s been a couple of things that have been helpful. One, I really tried to take a first team mentality and what I mean by that is, you know, first and foremost, I’m a leader with them, Starbucks, secondarily, I just happened to lead the data function.

Jon Francis (07:33):
And why that’s important is because I it’s really important for me to think like a business owner within the organization you’re respective of the capability I’m responsible for and being clear about where their challenges are, where the opportunities are what consensus and alignment needs to be gained and driven getting that credibility organizationally to say, well, it’s not just the data guy that sits off in the corner and has crazy ideas, but rather he’s part of this team and cares about transforming the organization. So I think there’s a bunch there just, that actually has nothing to do with, you know, hiring the best data scientists and having the best tools and capabilities, but it’s really around how do you earn the credibility and have a seat at the table to drive that transformation and influence where the organization goes. So I think that was a big part of it.

Jon Francis (08:27):
And, you know, I think the second part for me is really just immersing in what happens in the business. And, you know, even if I think about Starbucks, I’m in the field a lot, I certainly was more before COVID and traveling to different markets and spending time in stores with their store managers and the leadership in the field and really understanding what makes the business tick and how the stores work and why are those jobs hard? How could they be better? And, and then thinking through, well, what are the applications you can build the data products, the capabilities that you can bring to bear to help those areas versus I think what you see a lot of times and Alec functions is, “hey, here’s sort of a Wiziwig, a cool application”. That’s trying to find a problem. I try to start with, let’s start with the problem and make sure we’re building the most rational solutions that are going to be easy to comprehend and, and adoptable within the organization.

Chris D’Agostino (09:31):
And so, Jon, you know, we talked a bit about the change management and going in and, and, you know, being that business leader, thinking about data in terms of how it can assist the business in its goals. We talk a lot at Databricks about this overall sort of 10 step strategy for how do you go from in on-prem, you know, small data sets and not using data science and ML to moving towards cloud-based initiatives and being able to really leverage all your data across different business units and, and unifying these personas. And so we talk about not only the technology platform that’s needed to enable that, but also the cultural changes. So it’s consistent, I think, with, with your approach and going in and making sure, even as a technologist, you’re thinking about things from, how does it impact the business? How do I partner with the business to make sure that they’re successful? And then we’re not just building technology for the sake of technology?

Jon Francis (10:30):
Yeah, I think the other part of it too, is just thinking through how you can test and learn and start small. And, you know, instead of trying to boil the ocean, I like how you framed it. It’s, you know, start with some use cases where, you know, you can drive incremental value and maybe it’s not the perfect architecture. Maybe it’s not all the data you need or want but can you identify some use cases start small prove value to the organization. And then it almost becomes a flywheel where it builds on itself and people get excited about what they’re saying and the value it can drive. And then how, how can we get more of that? I don’t know, is that consistent with what you see with some of your other customers as well, how they approach the space?

Chris D’Agostino (11:14):
Yeah, exactly. You know what we talk about taking use cases and trying to bend them into different categories in terms of performance and feasibility. And so from a performance standpoint, I like to talk to customers about sub-second response times for use cases multi-sector and multi minute, because I think as you think about the use case and how quickly it needs to execute, what kind of data is needed in order to satisfy the use case. And then of course the algorithms that you might apply, you can start to figure out, well, these are the use cases we should prioritize first. And, and similar to you, this idea that you want to look for these quick wins, try to find something that’s in six weeks or less, that you’re not waiting. You know, the business isn’t waiting eight months for you to show results with ML, right? They want to see quick wins, but you also, as you said, you may not have the infrastructure. You may not have the data sets at the ready to be able to go big with a certain use case, but if you can start and be practical about it and pragmatic you’re gonna, you’re gonna get some traction, I think.

Jon Francis (12:15):
Yeah, I think it’s also in like, you know, I grew up as a practitioner as a statistician data sciences building models. And I think the challenge also is, you know, building something, I’m glad you mentioned scale because a lot of times a statistician or a data scientist might approach a problem as well I’m going to pick the hardest nastiest algorithm in terms of how I approach, how we’re going to solve this, this business prompt, but then if the can’t scale, then the utilization isn’t there. So I think that always needs to be top of mind, as you think about solutions that like, ultimately, if you want this to be running in production and driving value, you need to think about that way. And seeing that performance, especially for customer facing applications, if it’s on a website or an app that speed is going to be able to, in essence.

Chris D’Agostino (13:06):
So in terms of data products, you know, we talk a lot about at Databricks that the goal, as you move into this new data architecture and you collect more and more data, and you’re looking for insights, and oftentimes companies are looking for insights across business units. So there’s this push towards this sort of data product mentality of, we need to produce data products that are reusable. We have a particular data pipeline that does the refinement, does the data validation, the data curation. It makes sure the data is of good quality and it’s of a large enough scale that you can train machine learning models with it, for example, to, to bring in new use cases, new insights, how do you approach data products and, and forming your teams and identifying the potential? So the data products of course feed into the use cases, but data products is sort of a first-class citizen within the data ecosystem.

Jon Francis (14:01):
Yeah. And I, you know, I’m glad you mentioned insights because, because it always starts there just in terms of, in whether it’s through analytics that we’re getting off the behavioral data, or we’re doing ethnography and we’re talking to customers or we’re talking to our employees you know, I, I think it always starts with that. So when, you know, one example I’ll give in terms of just to kind of paint a picture around how we think through that, you know, obviously is a lot of retailers are facing during COVID. You see a lot of employers like ourselves working from home now and, and not in their usual routine of going in and getting a coffee in the morning on their way to work. And we’re seeing more shifting out to the suburbs in terms of, you know, customers who are now going to stores closer to home and not doing into cafes as much they’re going into drive throughs more.

Jon Francis (14:55):
And so, you know, we, we obviously can observe this through our behavioral data. And that insight yielded an opportunity for a data product, which is really around our deeper machine learning application that we built on top of Azure and partnership with, you know, capabilities from Databricks. But what we’ve it’s supported us, the ability to do is look at our drive-throughs where we have digital menu boards. And we know that we’re driving more customers through the drive-through. They don’t get as much of a chance as, as they have, you know, when they go into a cafe and really understand what’s on the menu and explore and talk to baristas. So we took this opportunity because we were getting more customers through the drive through to say, well, let’s elevate recommendations personalized recommendations through the drive, through to those customers that that’s based on the context of their basket, what the weather is like, what inventory is what top sellers are. And so we use all those to, to run reinforcement learning algorithms at scale that are then powering recommendations in the drive-through. And it’s been really well received by customers and it’s driving incremental business and attached for us and it’s performant as well. So I think that’s an example where we took an insight and we ran up through to a data product that is meaningful from a business perspective.

Chris D’Agostino (16:29):
Yeah, that’s really cool. And I mean, that’s a great example of an organization being able to adapt quickly, given the circumstances of COVID. Talk a little bit about the in-store experience we will, one day, hopefully soon, get back to the in cafe experience and the thing that’s always struck me as, you know, the, the ability to go in, you talk to a barista, you can sit down, you can do some work, or you can meet with a friend and have a conversation. And so it seems to always be centered around, you know, just improving that experience. How does AI play a role in, in, in the store to assist that?

Jon Francis (17:06):
Yeah, that’s a good question. You know, I think the way that we try to think about it and, you know, the, almost the overarching mission is, you know, this is not about robots selling coffee. I’m like, that’s not our value proposition. Our value proposition is around human connection and that special relationship our customers have with the baristas in the store. And, you know, Chris even think about your own experience as you go into your local store. And the barista usually knows you and remembers your drink, might even remember, you know, you had a big meeting yesterday and asked how it went or knows about your kid’s soccer game. And so that that’s special, right? The coffee’s amazing. Don’t get me wrong. That’s obviously will always be there at first, you know, front and center for the organization and the business, but we’re also in the business of connecting people.

Jon Francis (17:58):
So the way we look at AI then is, you know, what, what role can they play to make that connection stronger? And so if you think about, on whether it’s on the customer side or the barista side, one of the things that we can make their experiences easier, and whether it’s on the barista side or how we can automate inventory replenishment, or how we think about IOT applications or the customer side, where we can build more seamless experiences around ordering and delivery. I think that’s really how we approach that problem. It’s, it’s more about how we can take a lot of the components of the day-to-day for both customers and our partners and let them focus on that human interaction between each other.

Chris D’Agostino (18:47):
So, John, you talked about the example with the drive-throughs and trying to personalize the order process there. So the, these kinds of innovations, of course, you’ve got a lot of different stakeholders. We’ve talked about how you prioritize use cases which ones will drive value for the business and the fact that you and your team have got to think like the business owners as well. How do you how do you solicit feedback as you’re innovating and you’re building products, how do you get that feedback loop and, and give us some, some sense of the range of personas that you, you work with inside a Starbucks, for example, or in your other other other ventures, you know, the other organizations you’ve supported, you know, how do you set up those stakeholders to provide that feedback?

Jon Francis (19:33):
Yeah, that’s a great question. And, you know, I think whether it’s your, your customers or your employees, and that can be employees in the field that are working in stores or employees in our, our corporate headquarters, you know, we’ve got a few mechanisms through which you know, we collect feedback and this is why I really love my job because it’s not just all data science. We also do ethnography and survey research and design. So what we, we tend to do is build capabilities and to you know, the features that we’re launching to create milestones and points of interaction, where we can collect that feedback and in real time from, you know, from whether it be from our customers, whether it can be from our, our partners who are the, you know, that’s what we call our baristas, we’ve got stakeholders in the building.

Jon Francis (20:28):
And I think what I’ve learned through that too, is that, you know, that we try to be as customer centric as possible, but there also are times when you have to be considerate of constraints and I’ll give, I’ll use the drive-thru as an example. You know, if you’re recommending, let’s just say a product that, you know, might make sense from an algorithm, but in actually my potentially slow down the drive-through now you’re inadvertently slowing down the operations within the store. And so the point being that it’s super important that you get these signals from all over, whether it’s your customers or from other peers in the organization, because it’s, it can’t all just be machine learning. Like you’ll never get into some of those qualitative signals without getting that feedback directly. So we really try to be holistic in terms of whether it’s from a survey, whether it’s from talking to restock in the store whether it’s talking to my appearance supply chain or an operations around the capabilities. So we, we build that into all the products that are, that we’re really thinking through.

Chris D’Agostino (21:35):
Awesome. And are you all using say like NPS scores to track how well the products are received and what that experience is for, for you, for the baristas, the partners?

Jon Francis (21:47):
Yeah, we, we tend to, we have our sort of version of that. And we also you know, we look at a lot of qualitative data and then we do mining and against that in addition to just having one-on-one conversations, I think that all plays a role in terms of how we evaluate success.

Chris D’Agostino (22:05):
So, John, you talked about you know, switching roles going from a practitioner to leading teams and, and in, you know, trying to elevate their, their voice in, in, in the process and make sure that their needs were heard and, and maybe a higher level of appreciation within the organization for what’s involved in the data science space. That, you know, I would imagine as somebody that is in statistics and hands-on, you’re probably a bit like me as a software engineer, a bit more left-brained logic based. There’s an empathy side to this that comes in and the ability to, to build relationships and represent stakeholders and things like that. So for people that are listening in today and you know, learning about your background and the interesting work that you’ve done and, you know, sort of the early hands-on experience that you’ve got and how you’ve led that into an executive role leading teams do you have advice for people that want to be in your position? They want to be that chief data officer within an organization. And what advice would you give them as, as they think about that?

Jon Francis (23:13):
Yeah, that’s a great question. And I think you hit the nail on the head, Chris, I love the word you use, empathy, and, you know, I, that was not a muscle I organically had and I built in, and I think it’s really just empathy around and it comes back to the point I made earlier. It’s about really understanding what problems the business is trying to solve and spend a lot of time listening and immersing in the business. And in less on, well, here are capabilities that I think are cool and could potentially solve problems, but it feels like you’re pushing a Boulder up Hill more versus starting the other way. And really thinking first, like a business owner. I think that’s one. And just again, having empathy for folks in the rest of these organizations that don’t think about data on all day long, or don’t think about data science, I think that’s an important one.

Jon Francis (24:09):
I think the other one I would say is don’t underestimate your role and elevating data literacy. And I think maybe it’s a bit self-serving, but what we’ve also found is it’s hard to get people excited and inspired if they actually don’t even have a foundational understanding of data and the role and impact it could play. So what role would you have as a Chief Analytics or Data Officer, and really actually upping the collective literacy around data and how people think about it, everything from defining KPIs to building strategies around analysis things that may not seem too sexy. But I think it’s almost a cost of entry to, to, to be credible and help the organization better understand the value of the capabilities that we’re building. And again, that’s a little self-serving, but ultimately I think it’s a path to success is really helping others along on that journey.

Chris D’Agostino (25:14):
Yeah. And I would imagine given, given the consumer focus that you’ve had in your career, that the ethics component of data and that personalization is also, you know, as part of educating stakeholders and making sure that people understand, like, what is the art of the possible with data what’s pragmatic and realistic, and then what’s ethical, right? Like you can do a whole range of things with the data that you collect on your consumers, but not all of them are great for business. They’re not all great for the consumer. And they may not even be legal in some cases.

Jon Francis (25:47):
Yeah. We, we debate a lot creepy versus cool and even things that are not, that are legal may just be creepy and you have to think about the brand impact that it would have. And do we want to show up in a certain way and do some of these things that ultimately might disenfranchise some of our customers. So I think that’s an important question. And you know, something to think, think through how you think about privacy and how you educate your team on that too. You know, Chris I think the other one that I think about as well as is, you know, and, and maybe along the lines of empathy the word you use is thinking through the impact like machine learning is not going to live in isolation of human decisioning and the real magic and the power is actually when you combine what machines are good at and where they can automate you know, advanced decision-making combining that with what humans are really good at.

Jon Francis (26:50):
And I think that’s been a big aha for me, just in the last 10 years, especially between Nike and Starbucks, is that the machine learning by itself will never solely play a role in the absence of the role that whether it’s a marketer or someone in operations that that thinking and experience that they have either to help the machine learning perform better or to augment what you’re doing in terms of the decision from, from the algorithm. So it has to be a balance. And I think that’s a little bit earned in science that people just have to be comfortable with.

Speaker 1 (27:26):
Thank you for joining this episode of champions of data and AI brought to you by Databricks. Thousands of data leaders rely on Databricks to simplify data and AI. So data teams can innovate faster and solve the world’s toughest problems, visit databricks.com to learn how data leaders are unlocking the true potential of all their data.