Adel Nehme, the host of DataFramed, the DataCamp podcast, recently interviewed Megan Brown, Director of Data Literacy and Knowledge Management at Starbucks.
Adel Nehme: Earlier in the year, and our 2022 trends report, the first trend we discussed was how organizations will accelerate data, culture, and literacy programs. One thing that has shown this to be true is the rise of specialized data roles within organizations like head of data, culture, or director of data literacy that focused purely on the data literacy and culture conversation.
So who better can help me understand these roles other than Meghan brown? Megan is the director of data literacy for the analytics and insights team at Starbucks. She founded the data literacy team to drive analytics, product adoption, and to translate complex concepts for business stakeholders.
Throughout our conversation, we speak about her experience, launching a data literacy function, how she defines data literacy. What are the main levers she looked to improve as part of her program, the importance of executive sponsorship, evangelizing and marketing, the data literacy program, how to approach humanizing AI and data science, and much more.
If you enjoyed this podcast, make sure to subscribe and rate the show, but only if you enjoyed it now let's dive right. Megan. It's great to have you on the show. Thanks for having me here. I am really excited to be chatting with you today. You're someone who's really at the forefront of the data literacy conversation and you're paving the way for a specialized data literacy function within organizations.
Do you mind briefly walking us through your background and how you would describe your current role at Starbucks?
Megan Brown: Sure. I started out as a fifth grade teacher through teach for me. I didn't think I was going to teach fifth grade. I thought I was going to teach something in high school, but what I learned is that I'm pretty good at designing a curriculum.
For math, but for English language arts, the best practices didn't really work the way that I was told they would. So what I did was I went to grad school to study reading comprehension, because what you do is you follow the hardest problem. Earned my PhD in cognitive experimental psychology. The lab I worked in did what was previously known as neural network models.
Now deep learning Corpus analysis. Now NLP. And then experiments, of course, because we're psychology, you've got to do experiments while I was there. I also spent a lot of time learning more quant methods. So econometrics and looking for the similarities between the neural network models we were building and things like structural equation modeling.
So it was a lot of fun. That was my hobby. At some point, after 2008, my friends started leaving to be data scientists. And I was like, oh, it won't work for me. I'm different, whatever. And at some point it just became clear that they were having a lot more. Fun in their work. They were solving challenging and interesting problems and they were going much faster than academia.
So I crossed over and started being a data scientist around people, data science. So at first my bread and butter was predictive attrition models for employees and large companies trying to help their managers lead a little better by giving them some anticipation of like why that person might be at risk of.
And since then I've done a few different things. So I moved into marketing, data science, then marketing analytics, and I've been in people, leadership in the pandemic. So I've moved around to, and I guess I should say now I'm on the knowledge management and data literacy team. The first time my org has done something like this.
Then if you think about it, knowledge management is making the work we have produced over the last three, four or five years, really easy to find. And data literacy is making sure that the people who find it know how to use.
Adel Nehme: That's really great. And the background in teaching academia and applied data science is really perfect for leading a data literacy function.
I want to set the stage for today's conversation and kind of unpack and break down what it means to drive their literacy within an organization. So let's start off with the basics. How would you define data literacy?
Megan Brown: We have an array of skills with data at Starbucks. Some people are relatively nervous about it and others just are in the data every day.
And can't imagine making decisions without it. We have an array of talents in our audience. So actually defining data literacy gets a little challenging in there because some people you just want to use the dashboard and have. Inform their decisions with it. Other people you want to have coding and actually running some kind of simple, straightforward analysis from the right data, with the right metrics definitions.
So it depends on the audience. It's a terrible answer, but. So in some sense, it's a continuum of skills. And so ever since your department was conceived, you and your team have been working on launching and iterating upon a data literacy program. So I'd love to understand where are the main components of a data literacy strategy and how did you go about architecting your program?
We actually thought we were going to be teaching courses, right? That's what we thought we were going to be doing in the first month. Turns out you've got to make your work accessible to people before you can teach them about it. So we ran into this issue where people had projects tucked away. They'd never properly launched them.
And so we, we knew about them cause we were in the org and we wanted to bring them into a class, but we didn't have what we needed to integrate it really well. So we invested a great deal of time in knowledge minutes. Which for all purposes is really cleaning up our habits about how we share our work.
Sometimes data scientists can be a little shy about big presentations, so they tend to hide away from demos and launches. And they think that like a deck or a launch on Tableau is going to cross the divide to the business. And it doesn't always work out. We started a tool, we call the Wayfinder. So it's a very, very simple search.
It's a very data science solution to the knowledge management problem. You have a table with a dashboard on top of it that allows people to search for things like. We did some launch training. We inventoried all of the launch habits. Our org had and made a menu for people so that things other people were doing really well became really obvious.
We also have been dabbling in standardizing data engineering processes, standardizing how we share code and things like that. So we've been internal facing for about the first time. Our current approach. So what we thought we were going to teach these big classes that are going to be very exciting. I think one of the things that's become really clear on the pandemic is people are tired of screens, right?
So we're like, okay, we're going to do this a little differently. Now that we're two years in, we're going at the team. And we're piloting a program that's a six to eight weeks of basically mini lessons from the education, 15 minute lessons on data. I think one of our main concerns is like, what is a point estimate, exactly? What does a point estimate mean? What about all those other data around that point? So really digging into here's why we use averages. Here's the problem with using averages? And that sort of thing. So pretty fundamental at the team level because we know we can get their leader's buy-in and then we're good, good data scientists will measure all the things.
Adel Nehme: That's really great. So in summary and correct me if I'm wrong at the beginning, the efforts is really around structuring how the data team communicates with the wider organization at large. Then you started focusing on the actual data skills of everyone else within Starbucks.
Is that correct?
Megan Brown: Yeah. I found that we have to share what we've got with each other, even within the org, before we can all advocate for each other's work and adoption.
Adel Nehme: And what was the process of prioritization like between focusing on packaging, the data team's output and focusing on the rest of the organization skills?
Megan Brown: Sure. I actually got a little impatient, so it was a little bit of me where. I am a problem solver. If you come to me with a problem, I'll try and figure it out and it can be a data problem, a data science problem. It can be a culture problem too. And I realized I was getting mired down in progressively smaller internal issues where we weren't going to get the return that we'd gotten on our first stuff.
Going out. So we made the decision to pivot. And so it took us a little time to get out of the internal sort of cleanup and organization, but work. But we're on our way. You mentioned here that part of the strategy is organizing the knowledge that the data team is creating, and this is made accessible.
Adel Nehme: And the tool that you call Wayfinder, do you mind telling me how Wayfinder works and what you've learned working on this internal data portal? Like I said, it's a very proper data science solution where. We created a table and put a dashboard on top of it. So that's as fancy as it gets at the moment.
Megan Brown: What we've been doing progressively more over time is going to more API. Ingesting data from places where people are just putting information and folding them into that table in the background. So if you think about your stakeholders and what they care about at Starbucks, they'll care about decks, they're care about demos and launch videos.
They'll care about quarterly summary videos. They might care about white papers. And they definitely care about dashboards generally. So those are the first things we built into the Wayfinder. We also have compliment to the Wayfinder, which is the analytics and insights library, and that's for some of the more technical stuff.
So that was the second thing we did was pull from our code sharing platform and pull from our technical documentation platform. And gosh, what else do we get in there? Our ticketing platform as well. So you can start to see all of the. Parts of our projects come together, but our stakeholders don't necessarily have access to that information.
Adel Nehme: How has the reaction of the organization over time, as you were widening out the rollout?
Megan Brown: It makes our gaps obvious, which is a part of problem solving. If you can't find information on this project, you have to go hassle the. And hopefully when you hassle the people, they put their information into a system that you can pull from for our stakeholders.
We have these things called immersions at Starbucks. It's where you meet people across the org. You don't need an official reason to immerse, but there's like half an hour to talk about what you do, what they do, what their dog does, all that stuff. So at emergence, I'll often bring it up to see if they've heard of it.
And half the time people are like, oh, I love it. I can find anything and I can apply it right away to what I'm doing. And it's so, so easy, so much easier than it had been. So that makes me very, very happy. Other times I introduced them to it. And I hope that the next time I talked to them, they're very excited about it.
It did inspire a bit of work in Starbucks technology called search and discover. So thinking about like, this is just for analytics. So we don't share things coming from other orgs. And it's in this pandemic world where we'll either be remote or hybrid in the future where we're onboarding new people all the time.
Why should they have to search across 70 platforms to find a thing? That's the fundamental question? How can we make this easier? And one of the interesting things is metrics and metric definitions end up being at the heart of it because that's the thing that shows. In everything else, you can talk about your sales and transactions metrics and any deck or video it shows up and you should be able to link them together.
Learning Program at Starbucks
Adel Nehme: Now, of course, we wouldn't talk about data literacy without data skills and education. So can you walk me through how you set up a learning program at Starbucks and what were the learning objectives that you first.
Megan Brown: So we have a very, very applied focus. Our main goal is to drive adoption of our tools and resources. So whereas a learning and development org might focus on like, this is a distribution and all of these general things, we're very much focused on like you're in marketing. Here are the dashboards you have. Let's talk about what's in these dashboards and let's talk about the decisions you might have.
With this information. And so that's really the heart of what we do is we'll do some basic data literacy constructs. We'll do some concepts like around machine learning, for example, but really ultimately in service. Of getting people to use our tools, to make decisions in their routine work lives.
Adel Nehme: What was that conversation like with the marketing team and how did you adjust the learning objective of your program to fit any given team's business outcomes or business objectives?
Megan Brown: So we're, we're just piloting now. And what I'd say is we know every team has a different set of needs. So right now, in our pilot, we're working with the leader to figure out where they think their team's opportunities. In the future, in order to scale, we actually need to turn that into us.
Adel Nehme: I think companies sometimes tend to fall into analysis paralysis when designing data skills programs, or data literacy programs, because there's so much to teach so many levels of competencies that you need to think about different personas that need different skills and so on and so forth. Can you walk me through how you manage this complexity and who are the main personas for your learning?
Megan Brown: We developed some interview based personas based on people's roles and their interest and the skills they have and how they learn. So our fundamental persona that we're focusing on is something we call the patron.
These are the folks that typically come to us with requests, but may or may not use the output of those requests in their work as it goes sometimes. So we have these consolidated 8%. They go from everything to like builders and advocates and leaders from like the folks who actually keep our business going, who might just not use data because they can avoid it.
They might use some data because they know where it is, but they don't know where the rest of it is. And then maybe there's. Like almost an analyst with their data. They know where it is. They've asked enough questions to get everything they need. They can maneuver a dashboard really easily. And so they just have leveled up their skill enough that we can catch them on to other things like machine learning and get them really excited.
And that's, that's really how we've done it. So we've picked one or two personas to go after. And that's narrowed the space quite a bit. I will say while we're organizing information for our technical personas, the builders and advocates, and we're really digging into sort of the bulk of the starbucks corporate population
Adel Nehme: When approaching these different populations, are you adopting a tool agnostic approach or are you focusing on upskilling for specific tools?
Megan Brown: We go after specific tools being used at Starbucks we're we have to be very, very applied. It would be very frustrating for a lot of our partners, our employees too. Be told that things are possible and then not be shown what exists. And it would be even worse if we told them that things are possible.
And then what they have is actually this.
Adel Nehme: Exactly. I think the real beauty of an applied approach is you're able to get that aha moment on tools that people already familiar with. And that's. So I've seen you discuss this before, and you mentioned how executive training plants, the seed to accelerate digital literacy within their teams.
Can you describe the process by which you create data literacy champions by having these learning sessions with executives?
Megan Brown: The way I think about it is an organization will not change unless their leader really wants it. And. It's almost, I would prefer to work with the organizations where the leaders like, yes, my people need to use more dashboards to make decisions.
Their metrics are different in every presentation. I never know where they're getting their data from like great. Let's solve some of those problems. Then, then a leader who's like, yeah, Data. That's cool. So honestly, we do go for a bit of the low hanging fruit when it comes to like the teams we work with, because my team is relatively small and we need to invest in the right places where we think we're going to have an impact.
Adel Nehme: And that always comes from leadership that is willing to work with you on these data transformation project.
Megan Brown: Yeah. If the leader is not willing to say to their teams, that using data to make decisions is important. And then to incentivize somehow whether it's like social or positive regard, incentivize people who are actually shifting to use data to make decisions, then you're not going to get the change you want.
You'll basically get the early adopters and no one else because everyone else is busy with their jobs as they are.
Getting Leadership Buy-in
Adel Nehme: Can you walk us through what a difficult conversation with a leader, who's hesitant and investing in the data skills of their team looks like. And how did you go about that conversation?
Megan Brown: I would say actually the harder conversations are where people say they're supportive. Not that this happens at Starbucks happens any number of places. They say they're supportive, but in practice, they don't really want to invest the time to help their organizations. And so it won't show up in the first conversation.
Right. Cause they'll be like, oh yeah, data literacy. That's great. I read about it in this article last week, whatever. But then when you actually ask for their team's time, that's when the psychologist in me comes out and I talk about those learning actually happens and how behavior change actually happens.
And that's, I guess that's my go-to for winning people over is if they don't have hands on practice with things they're going to use routinely and that takes time, then they'll never change their habits
. For many business leaders there's often trade off between short-term business goals and investing in skills that accrue benefits in the long run.
Adel Nehme: How do you convince stakeholders to invest in a long-term play like upscaling?
Megan Brown: So we have erred on the side of being as close as possible to the short term business objectives. If we can improve people's work this week, we're really happy. Especially, let's say you're in marketing and they have, we have these three marketing dashboards and we know you only use one of them.
Cause we track usership. We can introduce you to the other two. We can talk about the concepts very briefly and get you using them. I think what we want to work towards is actually having a crew of people that want more. So what that second level of education looks. We're not really sure yet, but we want more advocates asking harder questions of our analytics and insights team.
So like the more meaty questions that aren't just like a pivot table, things that might have a model and to get there, we need to move past the applied, but the applied is where the need is right now. So you mentioned here marketing, I think an under-discussed lever within successful or data, culture, or data literacy programs is the importance of evangelism in marketing, the value of acquiring data skills.
Challenges and Best Practices
Adel Nehme: Can you walk me through some of the challenges and best practices you've gained along the way when marketing and evangelizing the data literacy program at start?
Megan Brown: Data scientists aren't necessarily marketers. One of the first things we learned in the first quarter we were doing this was like, Hm, maybe our current channels aren't working. So we started expanding the channels we have. We've done a lot of, we use a lot of user research before we stand up a new solution, both into internally and externally. And what we, one of the questions we ask is how do you get information on the data. And sometimes the answers are unsatisfying. Sometimes there's no way that they get information on the data, which is challenging because I know there's a number of people within Starbucks trying to get information out into the world, but it means we don't have the right channels.
And I think we've built a number of them. We have this really great org newsletter that gets to a lot of people and I try and have my team get something into that newsletter. Every time it goes out. One of the current challenges is people's habits have changed in the pandemic. So well, newsletters used to be really effective within Starbucks.
I think there's a bit of exhaustion. People are in however many hours of meetings a day and don't necessarily have the time to read very deeply. So our, our messages have gotten shorter. And drive people to resources, but I think there's also one of the things open questions for myself is I think people in different orgs are getting information in different places that we don't have access to.
And so part of my question is, okay, where are these places? What are they checking routinely? And how do I get access to them as an outsider? How do I get my message in there? And so it feels a little bit more distributed. Then it was at the beginning and it's challenging. And we have a bunch of new folks who have only ever worked at Starbucks remote.
And in the before times, Starbucks was a very relationship oriented company and you got a lot of information through relationships and we're really looking to break that right, get it out of the relationships, get it into something systematic that can be easily found by folks. Who've never been to the Support Center.
Adel Nehme: That's really great. And I think something that comes out from your chat here is the complexity of communications, that massive matrix organizations. So how do you deal with the complexity of communication and large organization?
Megan Brown: I mean, even within our own org, there are two main channels of communication. One is in use by folks who are more business facing and folks who are in leadership roles. And the other is in use by folks in more technical positions. And so even within our own org, if we have something we want to say, we have to get it into both places. And then we probably also have to get it into our Friday coffee tasting as you do at Starbucks, we probably also have to get it into our all hands because.
People are busy. They're not necessarily taking in a great deal of information right now. So really as many places as we can get the message in and get an entertaining message out there in the world that kind of. Feels a bit different from the rest of them. So how important do you find executive sponsorship of your work impact reception across the board?
We have had a great deal of change in the pandemic. So what I'd say is the great resignation actually hit leadership first. And so our pathway was really, really clear and then became less clear. Right. And so we find that the person we thought was over that thing that could become a champion is no longer over that thing or whatever.
We really find ourselves starting from scratch more often than not, unfortunately, but that's what happens when change takes over really.
Adel Nehme: And from a strategic perspective, how has executive sponsorship helped you break down these silos? Even, before the great resignation?
Megan Brown: We were brought into a number of conversations and able to give demos to audiences that we wouldn't have had access to otherwise.
Yeah, it was really. Our SVP at the time, using his relationships with others to get us into the room. And that, I mean, it was delightful and we're hiring a new SVP. So at some point I looked to take advantage of that again.
Adel Nehme: So when evangelizing data science within the organization, do you find that there is anxiety within the broader population? That needs to be assuaged. For example, data science, data skills, machine learning are often associated with automation and job loss. Whereas more often than not the evidence points towards augmentation and not automation, how do you convey the message effectively that data skills augment jobs and do not necessarily automate them?
Megan Brown: I think actually this is one of those places where sometimes we use prescriptive analytics and that really rubs people the wrong way. Because if you're prescribing something and they can't actually make a decision as a human in that process, they're going to reject it. I think to some extent, What we need to talk about is a portfolio approach.
This is also from education. So you need data in your portfolio. As you make a decision. Every decision has its own set of things. Certainly people have a lot of experience. They have a lot of contextual business information that we don't have. And that our models may not surface. And so everything needs to be in context and that's, that's the art that's the hardest part really is getting people in the business skilled up to be able to include data or consult with a data or decision scientists to include data in their portfolio. And then we also have to work with our data and decision scientists to improve storytelling and business acumen so that they can be a part of the portfolio without too much wrangling of language that needs to happen to crossover.
Adel Nehme: And how have you approached these conversations around automation?
Megan Brown: I mean, I joke about it a lot that the robots are coming for our jobs, but honestly, maybe they'll come out first. Um,
I do. I use a lot of humor. I think we also culturally have a bit of a fear of math. Like I've dated people who put staples through their thumbs in elementary school to get out of a math test. So numbers aren't necessarily where we feel the strongest in general. So part of it is they're worried that their career won't be able to keep.
With the demands coming at them about data analytics, machine learning. And part of that is how we communicate. If we're not paying attention to the business context, and we're going into the deep technical layers of the thing with a business audience, they're not really listening, we're talking to ourselves.
And so that just adds to the fear of math. And then I think there's so much hype around data science. That it verges on the robots are coming. We won't need humans to make this set of decisions anymore, but actually the narrative should be, we're augmenting your job. We can take these really basic decisions that are really annoying for you, and we can help make them, you can check us all you want, but then you get to make this tier of decisions that are like, we're not really capable of recommending yet.
And maybe it shouldn't be.
Adel Nehme: I couldn't agree more about humanizing data science and AI and using humor and down to earth language to describe machine learning and data science. I think data scientists tend to make mistakes sometimes in communication because they adopted technical hat and this reinforces the fear of math and a lot of the audiences that they speak with.
Now, of course, with communication and active participation within the programs, you must also receive a lot of feedback on how to improve the program itself. So can you walk us through what. Like for a data literacy program. And what are the main inputs you prioritize above Isles when looking for improvements?
Megan Brown: So, as I said earlier, we do a lot of user experience research. What we've learned to do over the last year and a half, two years is to take an idea out with some wire frames, to talk about it more broadly, to use that, to help us make decisions, and then to get. Because we're the first ones doing this at Starbucks in this specific way.
We really don't know. We don't know if people want many lessons on teams. We don't know if they want like a three hour thing where they learn concepts of data science and concepts of machine learning. So we really do pit things against each other. We have the satisfaction measures and like whether people enjoy it, but we also do.
Test people. We quizzed them because we want to see them do the thing we taught them to do. Otherwise we like, no matter how much they enjoyed it. So, so we try and be real friendly and lightweight about it, but we absolutely do ask them to show us what they just learned.
Building Data Culture and Data Literacy
Adel Nehme: Now, looking at the future. I think your role is very fascinating because you sit in a function that blends data science, learning change managers. Solely focused on building internal data, culture and data literacy. A part of me feels like the data literacy role is going through or the data scientist role underwent 10 to 15 years ago. I see much more often now organizations really investing in a data culture manager or their literacy manager. So I wonder how do you see this role evolving or function evolving within the industry over time?
Megan Brown: So in my experience, a lot of data science, orgs erred on the side of being. Clearly only technical to some extent like you had your leaders, but the vast majority of roles on the team were heads down data science, decision scientists, data analyst roles. What I think will happen and I think is happening is some data scientists are really good and interested in pitching to the business.
And just by experience, they'll get better and better at it over time. Some folks are not. And what, so I think the big question is what does an organization do if they discover that they have a technical folks who really just want to build technical things, how are they going to translate? Right. And who is trusted to translate?
Because I think there's a lot of tension. I experienced a lot of tension, even as a data scientist around who gets to speak for whom. If, when they give the presentation, they're going into the deep technical weeds all the time. How does that crossover to the business? So there's a bit of like how much storytelling do we expect of a data scientist?
There's a big question there. I have my answer, but other people have other answers. And then I think we also have a lot of experience and I've worked at three decent sized companies. Where if people aren't data literate, they outsource their data questions to the analytics team and they tend to be really low level and not where the analytics team really wants to be spending their time.
So really leveling them up. So we get more interesting questions and then trying to figure out what do we do? What do we do around storytelling? Who tells stories? How do we zoom people out of the details? I think that. For a lot of data scientists. And I remember experiencing this a while back by the time you have your pitch to the business, you've left out all of the flaws and features of the data you've left out.
The modeling decisions. You've left out the feature decisions. You really just telling this very, very, very high level story. and that can feel really unsatisfying when you've put in two months before. Into that thing. But if you go into the details, maybe that thing doesn't get used
Adel Nehme: In some sense, do you see data science splitting off in organizations with teams that are focused on applied data science and others focused on the adoption of data science?
I, I do. I think it depends on how much data science we expect people to do and how much communication weights people to do. I think we're super unclear on that right now. And you will always have folks coding production. Data science and someone there needs to pitch, but they don't need to pitch all the time.
But the closer you get to the business and immediate business need, the more you need to communicate and the better you need to communicate,
Advice for Anyone Trying To Break Into the Data
Adel Nehme: I guess we're going to have all the time in the world to find out. Now, as we close out, what would be your advice for anyone trying to break into the data, culture and data literacy conversation within their own organization?
Megan Brown: I'm the type of person that always ends up to. No matter what my job was, I always ended up teaching because none of this is rocket science. I wasn't born knowing analytics. I wasn't born knowing data science. So I just think it's something that can be shared. So I'd say the first step is to try teaching, right?
If you have people arriving into your org and the great resignation who don't know the thing, you know, your value isn't in just your knowledge. Your knowledge is very important. The depth of your knowledge is very important. You can easily be as SMI if you spend enough time on it, but I would never hoard knowledge.
I would make sure you were sharing it and make sure that people around you knew you were sharing it, make sure your leaders know you're sharing your information. It's a sign of a team player. It's a sign of good citizenship. And it also means you're not insecure, right? You're not worried about that person being able to run the same model as you, because you keep learning yourself.
Adel Nehme: That's awesome. Megan, I had a fantastic time chatting. Do you have any call to action before we wrap up today's episode?
Megan Brown: One of the things that I have found most valuable is spending a couple of days with our stakeholders, watching how they use our dashboards, what tools they get information through and what they're ignoring, because they're there, they're missing signals.
We're sending out because we're not putting them in the right places. And so if you notice that you've built a dashboard and two people use it, go to those two people and ask who else should be using it and ask if you can shadow them for it. Because then you're going to find out exactly what's missing from your dashboard or what's missing in your communication about your dashboard.
That's preventing other people from using it.
Adel Nehme: Thanks for coming on data friend.
Megan Brown: Yeah. Happy to thank you.
You've been listening to data, framed a podcast by data camp. Keep connected with us by subscribing to the show in your favorite podcast player. Please give us a rating, leave a comment and share episodes you love that helps us keep delivering insights into all things. Data. Thanks for listening until next time.
← Back to blog