Resources
Feb 21, 2023 |
1h 1min
|
592 views

Data Science Hangout | Emi Zambada, Nationwide Financial | Building a story for the C-suite

We were joined by Emi Zambada, Assistant Vice President, Annuity Distribution Intelligence at Nationwide Financial.

In talking with Emi about data science in insurance technology and managing teams through digital transformations, we also learned more about building a story for the C-suite in our own organizations. (40:26)

Storytelling is critical.

I go back to thinking about how we speak to little kids. You need to speak their language. They are not going to speak your language. If you spoke in Shakespearean language to a kid, they’re not going to understand you and they don’t care. You’re going to lose them.

Listen to your leaders.

Listen to the earning calls or whatever ways that they present to the organization. Pay attention to the words they use. Pay attention to how they describe things and the metrics they share.

Use the words they use.

Words are important. You can say the same thing with different words. We talk about data foundation, but if you say data integrity - you might say, well data integrity and data foundation are the same. If they use the word data foundation, use that. Use their words so they can understand you.

Keep it consistent.

Think about the old times - and the old fashioned newspaper. The New York Times always looks the same. They have the same structure, the same format. The articles and the layout are similar every day.

Imagine if they tweak things and move things around - you will spend a lot of energy and time trying to figure it out. It takes mental energy to figure out the format and the structure that you don’t care about. You want to read the news. All your energy is to understand the content of the news.

The same thing with storytelling and dashboards. Find a way to use the words, the structure, and the metrics that they use and are familiar with so that they don’t spend time and energy trying to understand your dashboard - your story.

For example: the metric that you’re tracking is going this way, we might need to take action on this.

Be very specific.

Don’t say a lot of things. Less is better. Be clear, concise, and very specific on the things that you really care about.

Repeat the things you care about.

Silence doesn’t mean they agree. Watch them, look at their expressions and repeat.

A mistake that we can make is that we don’t recap what we talked about last month. You can’t walk into a meeting thinking that they remember everything that happened last month. They don’t. They have done so many things since then. Don’t feel offended, it’s normal. Go back to that and say, last time we spoke we did this. Now we’re going to do this and let’s bring it to the present.

Speaker bio: Emi leads strategic business intelligence for Nationwide Annuity distribution by leveraging data, analytics and technology and partnering with senior leadership to identify trends and measure efficiency and effectiveness of distribution. Emi is recognized as a thought leader and was recognized by MIT Chief Data Officer Magazine´s “40 Under Forty Data Leaders in 2022” and is a recipient of the Hispanic Association on Corporate Responsibility’s “2022 Young Hispanic Corporate Achievers Award”. A proud Mexican immigrant, Emi serves as an advocate for minority groups in AI and tech.

► Subscribe to Our Channel Here: https://bit.ly/2TzgcOu

Follow Us Here:
Website: https://www.posit.co
LinkedIn: https://www.linkedin.com/company/posit-software
Twitter: https://twitter.com/posit_pbc

To join future data science hangouts, add to your calendar here: pos.it/dsh (All are welcome! We'd love to see you!)

image: thumbnail.jpg

Transcript#

This transcript was generated automatically and may contain errors.

Emi Zambada, Nationwide Financial, Building a story for the C-suite

Happy Thursday, everybody. Welcome to the Data Science Hangout. I hope that everybody's having a great week. I'm having an awesome week. We are in Palm Springs, California for our work week. So the whole, well, most of the Posit company is all out here just kind of working together and just getting to do fun things with each other because we don't see each other all the time. But if this is your first time joining us at the Data Science Hangout, so nice to meet you. I'm Rachel. This is an open space to chat about data science leadership, questions you're facing and getting to hear about what's going on in the world of data across different industries. And so every week we feature a different data science leader as my co-host to help lead our discussion and answer questions from you all. So together, we're all dedicated to making this a welcoming environment for everybody. And we love when we can hear from everyone, no matter your background, level of experience or area of work. It is totally okay to just listen in. I was actually just telling this to some of my co-workers as well. But there's also three ways you can ask questions and provide your own perspective on certain topics too. So you can jump in by raising your hand on Zoom. You could also put questions in the Zoom chat and feel free to just put like a little star next to it if you wanted me to read it out loud instead. And then we also have a Slido link where you can ask questions anonymously. And then if this isn't shared in the chat right away, I'll put it there in a second. We do share the recordings of each session up to our Posit YouTube and our Hangout website. So you can always go back and re-watch or share with friends.

One added note I just wanted to throw out here, if you are hiring, feel free to share that in the chat. I never see that as like spammy or anything like you want to share any open positions there, feel free to put them there. But with all that, I'm so excited to be joined by my co-host, Emi Zimbada, Assistant Vice President of Annuity Distribution Intelligence at Nationwide Financial. Emi, thanks so much for joining us. I'd love to have you just start by introducing yourself, maybe a little about your role and something you like to do outside of work.

Thank you. Thanks, Rachel. And definitely Yellows because you're in Palm Springs. I'm here in freezing Columbus, Ohio with summer snow. So definitely, you know, I wish I were over there, but very happy to be here with all of you. And I appreciate the time. And as you say, Emi Zimbada, I work for Nationwide, not the property and casualty, they are insurance. We work in the financial services, specifically in annuity. I've been in the company for over four years. I've been all over the place in working for Whirlpool, Siemens. I started my career or, you know, education in accounting, but life take you to places. And now I lead the what we call the distribution intelligence team that is nothing else but using technology and data analytics to support our sales teams to be more, more effective and efficient. So again, thanks for having me here.

Thanks for joining us. And what's something you like to do outside of work?

I like, I like reading. I like have a beer. I go with cheap beer, Bud Light. I like sports and have a little boy and spend a lot of time with my wife and the kids. So yeah, and not a lot of free time. But whenever I can, definitely, you know, like to travel, go places, visit Mexico. I'm originally from Mexico.

People development and personal branding

That's great. So, I mean, I know when we were speaking briefly before the hangout, you mentioned a lot about people's development on your team and how you get joy out of helping people develop. And I thought that would be a fun place to kind of kick off the conversation and just learn a little bit more about how you do that with your team.

Yeah, absolutely. Yeah, definitely passionate about personal development, helping people with their careers. Nationwide, we have over 100, 150 data scientists, over 200 data, data engineers or data related roles. So it's a big group. And it's known that in the data science space, people are mostly introverts. They like to be heads down, coding, and they do amazing things. But one thing they are not very naturally good at is thinking about their career, building a personal brand, networking, things that are really important as you build your career, right? And as you have aspirations to become a future leader. So I'm naturally an introvert, but probably push myself to be out there and build a relationship.

So I always tell people that it's very important to get the work done, get the results. Results are what matter, but also it's important to be out there, build these meaningful relationship networks. And don't be shy about connecting with people. You never know, you will make a new friend, you will get into a new partnership, whatever that might be.

One thing I want to acknowledge here, and before we continue, Rachel, is I don't want to ignore the situation that's happening right now in America with Ty Nichols and the not very good news happening with the Black community. So I want to make sure that we acknowledge that, we support them. One thing that we do, I feel very well at Nationwide, is the diversity and inclusion. So helping people from all different backgrounds and ethnicity and races. So I want to make sure that we all acknowledge the situation that is happening and support organizations that help those groups. So I'm wearing here my hoodie from Latinx in AI. I'm a big fan of that organization. I try to help them as a Latino, but then I want to make sure that if you have the means, anything can help. I've been a mentor, financially supporting. There's a lot of groups like a woman in AI, Black in AI, AI for all, you name it. So go out there, support those groups, and do whatever you can to support that community.

Absolutely. Thank you for that, Emi. And if you have some links, I know you just mentioned a few groups too, but if anybody wants to share some of those links in the chat so that we all have them here too, that would be great.

Yeah, absolutely.

ChatGPT and AI in the workflow

I see we're starting to get a few questions in from the chat. And Jared, I see you just asked one. Do you want to jump in? I see you unmuted, but I can't hear you yet. So I can ask it.

I can see the question, so I can read it.

It's okay. I'll read it. But the question was, with buzzwords like ChatGPT, has AI impacted your workflow as of now?

Yeah. Specifically with ChatGPT, we started using in a limited way, to be frank. There's a lot of great stuff about that technology and that tool specifically. There's a lot of things that we need to make sure, like privacy, security, making sure that it's used appropriately and we don't erase the data or secret solve the code. So we have used it, like we use Google and Stack Overflow and the many other tools that are out there. So I want to make sure that we do it in a meaningful but educated way. So we've been trying to release some statements about what is it, how to use it, where not to use it, but definitely evaluating what is coming in the future.

I think AI in general has impacted your workflow. Absolutely. We started doing AI and machine learning models over 10 years ago, like probably a lot of the big companies. We have a lot of models in production, producing results. We have a patented model factory in place when we have all the AI models implemented. So there are many, many different ways that we use them from underwriting, pricing, claims, acquisition, you name it. So we can always get into any specific detail, but definitely it's one of the many ways we use it.

Bridging data science and the business

Thank you. I just wanted to remind everybody, we shared the Slido link in the chat too, if you wanted to ask questions anonymously, but if you just join, feel free to just raise your hand on Zoom, or you can use the Zoom chat as well.

But something I also wanted to ask you about, Emmy, is in the Hangouts, we talk a lot about communication across the business. And I know this is something that you've probably heard a lot that you focus a lot on, like bridging the gap between data science and the business. And I'm curious if you have any tips for all of us.

Yeah. A big part of my role now, and I'm fairly new in the role, like three months, but even my previous role, I was leading the advanced analytics team, that was kind of the translators in between the business and the data scientist. And even now, as a data intelligence, I'm in between ourselves, people and data scientists or business intelligence team. I think one thing that is really, really important is understanding the business, understanding the people using the different technology that we have. And I will split it in two ways. So one is the leaders, and the other one is the frontline associates or people in the field. So you need to understand both. You need to make sure that both are good with whatever you plan to build. And let me start with the leader. So definitely, you need them. You cannot ignore what they want. They know about the strategy. They have the funding, right? Money is very important. So without money, it's very hard to do anything else. So make sure that you understand where they want to go, what they're planning to do, what are the challenges they face, what are the things that they want to achieve, and ask questions and get to know them personally. I think that helps, you know, get closer to their heart. And explain what you do and how it works, but knowing that they don't live in our world, and they don't want to learn all the stuff that we know because they don't have the time and their focus is to run the business, not to understand AI and data and analytics, but knowing what is enough so they can get a good sense of what we do. And they can, you know, they can relate and think about how that can be used for the work that they want to be done.

So again, you know, not to be a brown nose, like I said, but really important to know them and get to understand them. And then you have the frontline people, like salespeople, call center, whomever is going to use the technology that you have. Make the things easier for them. Don't build something and say, well, they need to figure out how to use it, because they're not going to use it. They are going to, they are busy, they have a lot to do, they have a very tight metrics, whatever else, whatever that is, like a sales goals, call they need to handle, you know, whatever that might be the case. So you need to make it very easy so they can, that doesn't drastically change the work they do. Think about like, you know, Netflix, when you watch Netflix and you have recommendation, you just watch TV and then just get, you know, the different recommendation that you just select that. So you don't need to know what is going on there, or you don't need to understand the algorithm or change the way you watch TV. It's just a helpful tool.

So for all salespeople, you know, we tend to make mistakes. We give them a lot of information, a lot of data points, a lot of dashboard, and we overwhelm them. And then they get to the point that they don't know where to find it. They don't have the time to look at it. They get, you know, there's too many data points that they get this phrase with all these dashboards and situations there. So we need to be, instead of giving them a lot of dashboards and cool visualizations and things that they might not even look at it because again, they get overwhelmed, tell them what is the action, right? Like who they need to contact and why, and tell them, you know, what to do next. So they start the day and say, all right, what do I need to do? Okay, you gotta call these people and you're gonna talk about this and you're gonna then do, you know, do these activities. And a little bit of why helps in simple English. Definitely don't build, you know, precision and inaccuracy and anything like that, or don't share those things. But then, but you need to build the trust. Like, for example, things that are ways to

erode the trust is you tell them, hey, contact this person and the phone number is wrong, or contact this person and the person is no longer working for that company, or hey, do this and they try it and they try it and that doesn't work. You're gonna lose that trust and that's definitely not good. So make sure your data is good, you validate some of the things we're sharing and ask for feedback, right? People forgive if you give them something that didn't work or the data wasn't good, but if you are arrogant and you say like, but you gotta use it and I'm sure, you know, what I gave you is good, they're just gonna ignore it, you know, in the future. So again, start with your leaders, understand them, get closer to them personally, professionally, and then, but don't lose, you know, don't lose the focus on the end user.

I love that, thank you. And when you say don't, like, overwhelm them with dashboards, how are you serving that information to them?

So, you know, the dashboard, I see one of the comments there by dashboard AI is, I think there's nothing wrong about having dashboard, the problem is that you need to share with them what is really meaningful for the work that they do, keep it simple, and if they want to get more detail, probably have a way that they can drill into details and be specific on what they need to do, right? Like if you say, hey, here is a dashboard and you can see how much they sell and the opportunities and what can be done, but be very clear on what is the, you're asking for, the eyes or the expectation, right? Hey, you're going to start talking from here to here, and this is why, and this is, you know, data points that might help you with your conversation. So keep it simple like that, and if they want to explore more, you know, give them the opportunity to explore more. Some people still like to get it in Excel and play with that, so it's okay, but be specific and don't make it to like a black box that they don't understand what's going on and why.

Marlene, I see you asked a great question in the chat a bit earlier. Do you want to jump in? I actually think, I mean, you actually kind of answered it already, but I was just curious, really, I struggle with the C-suite in my organization. I can give them flashy because flashy feels like it's new age and bright and shiny and look at what the 21st century can bring you, but you try to drill in like here's the takeaway from what you're looking at. The only problem is I have severe anxiety about live demos, so I was just curious, do you prefer live demos maybe, or that question was way more open-ended than what I'm getting into, but that was the idea. Something, you know, vivid and moving, or do you prefer static and something that they can just kind of stare at and take in and really kind of get everything that's on the screen?

Presenting to leaders: keeping it simple

Yeah, and a great question, and I don't think there's a magic formula, but I will tell you that the higher you go in the organization, the more you behave like a little kid. The attention span is very limited, right? They have so many things going on, so much stuff that they need to handle that if you put a lot of things or you try to demo something very complicated, you lose them, and it's very hard to get them back, their attention back to what you want them to do. So like a little kid, right? You play with a little kid and they get bored and they move on, they do something else, and you say, no, no, no, but here is this other toy, and they're like, you know, they're doing something else. So I think the leaders a little bit behave like that, right?

So you need to be, hey, what would they really care? What is in their mind? What is the things that they're worrying about? They want to increase sales in a specific product, they want to reach a new market, they want to target a new channel, whatever that might be, right? Or they want to reduce number of calls, whatever that might be, and be a start high level, right? Like, hey, you want to do this, you know, I see this in the data, and you follow and maybe ask a question or look at them, right? Like it's in person now with the video conference, hopefully they're on camera, you can tell, right? Like, are they nodding? Are they distracted? They go from camera, you lost them. So be mindful about reading the people, not only what they said, but what they do with their expression. So keep an eye on that. And as you talk about, and again, start high level, don't worry about the technology, don't worry about the, you know, don't talk about the tool, don't talk about how you made this all such, you know, like, but keep it something that they care. And I think hopefully, if you engage them, they're going to start asking more specific questions. Like demos, you know, to your specific question, but like demos are dangerous.

the higher you go in the organization, the more you behave like a little kid. The attention span is very limited, right? They have so many things going on, so much stuff that they need to handle that if you put a lot of things or you try to demo something very complicated, you lose them, and it's very hard to get them back, their attention back to what you want them to do.

So if you have, PowerPoints are still ruling most of the companies. So if you get a screenshot, and you put it there, and that's something that you prove it works, and I know what you're presenting very well. So get a very good example, get a very good dashboard, get to know that dashboard very well. Because if they ask you something about that, how you calculated this, or how you come out with this, or tell me more about that example, if you don't know that very well, you're going to lose that, you know, trust, and the perception will change about the work that you do. And probably lastly, I will say, be your, you know, like the people advocate, you know, what are the things that they're going to challenge me? What are the things that can go wrong about this presentation? What are the things that they might not like about what I presented?

I'm sure they're not going to like this. How am I going to convince them to like this? So don't prepare for the best possible scenario, kind of prepare for the worst scenario, and that will get you ready for, if some of that happened, right, or all of that happened. So I don't know if that helped. That was very helpful. Thank you so much. Yeah.

Data sources and tools

I want to just jump over to a few questions that were on Slido. And so one was, what type of data do you work with? Where does the data come from? And I guess this is a few questions in one. And also, what tools do your 300 plus people work with? Excel, Python, R, SAS, and what mix?

Yeah, you know, like probably any other big older organization, we're almost 100 years in business. So we have in different business units. So we have so many data sources, so many legacy systems. So data is all over the place, different, excuse me, different quality, different shape, different format, you know, different databases. So definitely, it's a challenge. We try to consolidate and mirror a lot of those data points. We have data lakes, we have data stores, and data market, you name it, right? So I would like to, I say, oh, yeah, we nail it. We know, you know, all the data thing, and it's perfect. So definitely, it's a big challenge. Data engineers are way underestimated. I think that we don't have many. We probably need more and better, you know, we have great people, don't get me wrong, but definitely, it's a very, very much needed profession. And if you are into data engineering, you got it. That's a great place to be.

We use a lot of structured data from like admin system, like kind of reports and transaction data and things like that. But we also use a lot of unstructured data like calls, tags, you know, emails, a lot of third-party data. We have quite a lot of information about the market, about the customers. So definitely, a lot of third-party data and try to combine with internal data. So in terms of tools, a lot of Python, still a lot of Excel, probably, you know, something we are evolving, but still very popular out there. A lot of Tableau. We started more on Power BI, databases, Snowflakes, a lot of, we have big AWS shops, so, you know, S2Bucket and SageMaker. So kind of try to narrow the number of tools that are used to make it easier for the team, but still, you know, different data breaks, different tools out there. We try to, you know, we're big into open source tools, like Python and R and things like that, Spark, but still, we use a lot of commercial tools. We used to do a lot of SAS, no longer a SAS job, but yeah, I hope that helped. Thank you.

Diversity in data science

I see some feedback in the chat that Javier said, if Emmy decides to be a career coach one day for data scientists, I think he'd be very successful. Thank you, Javier. But kind of on that topic, I know you have served as an advocate for minority groups in AI and tech. And I was curious if that is something that you do for people to help bring them into the space as well. And if you have maybe some ideas for all of us about how we could get more involved too.

Yeah, I think it's, you know, diversity is very important in any space, but data scientists is definitely super important because we build the future for companies where we build the new capabilities. We are, you know, truly the future of how we're going to do business. And by diversity, it doesn't only mean like a color, race, ethnicity, gender, can also be diversity of background, right? Experiences, expertise that you bring to the company.

So we have the chief advocate for the Latino group here nationwide. And our executive sponsor is the president of the biggest business unit. And he said diversity is not only the right thing to do, but a good business decision. And what that means is like, if you think about the demographic in US, I might get this a little slightly wrong, but that 16% of the population are Black and they have a $1.8 trillion purchasing power. And Latinos are close to 20% with another, you know, close to $2 trillion purchasing power. And you think about like a woman, right? Like half the population, a big, you know, group growing and getting more into business and increasing the purchasing power. So it's a lot of money there that if you don't tap into that, if you don't build solutions, products for those individuals, and they think different, right? Like there's no right or wrong, but every group, they have their specific needs, their specific characteristic.

he said diversity is not only the right thing to do, but a good business decision.

And you need to understand those so you can tailor your products, your solutions for them. And they feel good about buying your product, right? Because they feel like you care about them. You are socially responsible and you support their groups. And there's no other way to build those products than having the people that understand those individuals. So if you are a data scientist, if you don't have, think about a group that doesn't have any minority groups, you know, white males, nothing bad, white males are amazing. But think about like there's no diversity at all. So those individuals, they're not going to know exactly how a woman thing or Latinos or black, and it's going to be very hard for them to understand nuances and the specific things that they are very important for them. And they're not going to understand that maybe language is critical, you know, something that's funny. They're not going to know that probably data is biased or some of them, right? Like people building computer vision system that doesn't use a balanced data set and doesn't correctly train the models to identify properly black people, they're going to have a problem or people doing speech recognition, building those system to transcribe a speech into text and doing a lot of solutions based on that. If they don't understand that people might have a different accent, different, you know, way of talking, they're not going to potentially use the right data sets to train the models right and, and therefore not, not working properly for those, for those groups. And those groups are going to notice that and they're not going to be happy with that. And they're not going to buy your products. There was a public cases about that, you know, there's a lot of with recruiting with Amazon and another insurance company with claim when you needed to, you know, record a video about your claim and what happened. And it was a biased or some groups and people didn't like that. And they're going to, people are very sensitive and they demand a lot, especially after COVID. So if you don't treat them right, they are going to end with social media things, you know, get shared very quickly. So if you don't treat them right, they are not going to buy your products and they're going to switch very fast. So again, diversity helps with all of that and, and, and how you, you can help, be an ally. If you, if you, if you are a leader, hire those individuals, you are going to have more

powerful, better teams. If you are an individual contributor, ask people, mentor them, and help them. And you will learn to, you will learn a lot from that, from those interactions.

Thank you, Emi. Andy, I see you asked a question in the chat a little bit earlier. Do you want to jump in next? Sure. Thanks, Rachel. Emi, can you share about the open source, our adoption path for your team? I kind of have a few questions here. So that open source, our adoption, and if you, did you get any resistance from IT in supporting getting R into production? And if so, kind of what lessons can you share about onboarding and supporting folks new to R and open source tools?

Yeah, it's definitely, of course, security and data protection is important. So when we have this dedicated environment, we call it model factory, that is, is, is protected, it has all the security controls, and we have R and Python embedded into that. And every time there's a new package, it gets tested, the versions, you know, you cannot go and get the latest version or any versions that you think is best. We have an approved version for each of those tools. We, again, we test the packages, we give a very, we have very long, like a lengthy procedures of how to use it, what to do, what not to do. And we do peer reviews when people use, again, whatever, whatever package or whatever, you know, code you write, it gets reviewed by one of your peers. And then before move to production, there's another team that implement that, that validates your own incurring to any security issues, you know, like a credential are managed in a specific way, you know, you need to follow that protocol.

Now, there's a lot of things that I cannot share publicly, but there's a lot of controls around open source systems. So I think, again, we have a lot of, we are big fans, you know, it's cheaper, it's easier to have the talent knowing those tools, and people like to learn those tools, but definitely a lot of controls to protect the data and protect the code.

So I'm hearing guardrails are important, and then just like peer support. Yeah. Yeah, thank you.

Thank you. Alan, I see you also just asked a question in the chat. Do you want to jump in? Yeah, I had to walk back over to the computer. I'm pacing around in my office because it's cold in here. I'm just curious, I mean, as you described this great big team of hundreds of people, what kind of coordination happens across those teams? And like, is that a challenge for leaders of various teams to figure out how to collaborate? Are you, do you have to have agreement about things like architecture, or about practices and processes and stuff? Like, just wondering how you sort of organize across all of those people, and to what extent that's even like critical, or if they're pretty independent?

Cross-team collaboration and leadership

Yeah. Again, like any other big organization, there's different groups and different part of the different organization across the companies, like our data scientists are in finance, or data people are in technology. You have the data architecture team, and the security team, and within even within IT, or within finance, you have different groups. I mean, business. So I'm not going to, you know, I don't want to lie, and I don't want to ignore that having different organization, different leaders creates some difficulties. And I think that's a challenge.

Difficulties, we try to build a pod, a group of people. So we want to make sure that there's leadership support for the project. You cannot start without that. There can be some tests and learn that gets, you know, done more naturally. But, you know, you have a project, you have the support, you create a new business case, you get the funding, and then you say, okay, in order to do this, I will need the data scientist, the data engineer. So a lot of this, yeah, you can, you have support, you have the funding, but also you, you know, that was a relationship, and building a network is important, because you need to go with those leaders and get buy-in, right? Like, even though I might hear, I might say, hey, my sale leader, you know, the head of sales for my business unit, he really wants to do this. But if I go, I say, you gotta help me, because I'm telling you, you know, I say so, you know, that commitment might not feel that strong. So you need to go with them and say, hey, we're doing this because of this, and, you know, what you think about that, and, and yeah, you might need to fund the team or get, you know, like some resources for that. But then, but that convincing is important, right? The why is important. And then you go to this group and do the same.

And all you have these people, one thing I like to do is, I'm not going to tell them what to do. I'm going to, you know, I have this philosophy that you need to be very tight on what, what is the outcome that you want, what is the, the objective, and what is that, and you can, it will be needed. So be very clear on that, and hopefully they truly understand that, and, and you, you describe it, but then you wait for them to describe it, and you reiterate until it kind of sounds, you know, it's clear to them. I think that's the best way, right? Repeat, repeat what we're trying to do, and if they are still loose or not very clear, you gotta keep that message until they are clear. So be tight on that, but then be loose on that, how they get there. They are way smarter than me, so I will never use this algorithm or do it this way, because you, you, that, that controlling is, is no good. It will inhibit, you know, innovation, and, and, and they are experts. So be loose on that, you know, let them define what is that, what is the way, the how. The why and the what is important to be tight, the how, you know, they'll figure out, they're smart, but then be tight on the results, right? Like, go back and be tight on, are we achieving what we want, and be very, very specific and simple in that metric.

I hear someone or saw someone talking about KPI, so I want to increase sales. I want to, you know, reach this market share. I want to make that metric that is simple to understand for the data scientists how to get there, but simple for the business to, to understand that, you know, we're achieving what we committed to do, committed to achieve. So, so that, I think that's, that, that's, that, that's, that's important for us. So, you know, tie, loose, tie, give them the money, give them the resources, give them the people that is needed, let them run with that. So, we, we are, we're starting getting very big into the, the project to product, not being very specific on the projects and how all the waterfall and all that, but be more on the product, you know, here is the product that we're trying to, to get out of there, and let them, you know, let them figure out, you know, the details of how to get to that product and use the money and resources as they, as they wish.

I'm not going to tell them what to do. I'm going to, you know, I have this philosophy that you need to be very tight on what, what is the outcome that you want, what is the, the objective, and what is that, and you can, it will be needed. So be tight on that, but then be loose on that, how they get there.

That's great. Thank you. I really like that description of, like, the articulation of where to be really tight and focused versus where to kind of let people run, and that can be really, really successful when you've got, like, skilled people and platforms to work from that they can use well, and, you know, you can let them be successful in the ways that they're really, really strong at, and know that you're going to get to a well-defined outcome that, that you're helping steer. Yeah, I appreciate hearing all that. Thank you. No, thank you. Thanks.

Building a story for the C-suite

I see a great conversation in the chat around, like, metrics, around dashboards, a number of things to show in them, and I think it's Yinyi, and sorry if I pronounce your name incorrectly. Would you want to jump in and ask the question you put there? Yeah, thank you.

Actually, the question is about how to build a story for the systems, because actually, we are building dashboards that are our management system, so the dashboard should be kind of fixed just for a bunch of metrics on that dashboard, but when we make a plan, you know, design the dashboard, well, if we only put several, maybe four or five metrics on the top, and give a quick context, this might not be enough for a big company, since the story might be fixed, but sometimes our C-suite might want to know other side or another story, so we are thinking about to put maybe 20 or more index on that dashboard, but of course, the C-suite will say, oh, it's too much. We do not have time to, you know, read all that long reports.

Yeah, no, great question, and definitely, you know, storytelling is critical. I go back to, you know, the little kids, right, and you need to speak their language. They are not going to speak your language. Go back to the kids. You need to talk to the kids in their language. You talk about the Shakespearean language to a kid. They are not going to understand you, and they don't care. They are going to lose them, so listen to your leaders. Listen to the earning calls or whatever, you know, ways that they present to the organization or whomever is the leader. Try to listen to previous conversations and pay attention to the words they use. Pay attention to how they describe things. Pay attention to the things that they, the metrics that they share or the metrics that they talk about it, and use those words. Words are important, right? Like, you can say the same thing with different words, or you might have the way you describe something, use their words.

I go back to, you know, the little kids, right, and you need to speak their language. They are not going to speak your language. Go back to the kids. You need to talk to the kids in their language. You talk about the Shakespearean language to a kid. They are not going to understand you, and they don't care.

So, you know, if you can talk about, like, we talk about the data foundation, but if you say data integrity, you might say, well, data integrity, data foundation is the same, but if they use the word data foundation, use the, you know, same data foundation. If they say self-intelligent, you can use another word. Use their words so they can understand you and keep it consistent. You know, if you keep changing the words or the way, you know, the views, think about, like, in the old times, or even now, Google, but think about, like, probably the newspaper or the format, any website, but think about, like, the old-fashioned newspaper. New York Times, they always look the same. They have the same structure, the same format, the articles, the layout is similar. So, every day you go and read that newspaper or, you know, your online news, it looks the same thing. Imagine if they will tweak things and move things around, you will spend a lot of energy and time trying to figure out, and that's mental thing that your brain is taking energy to figuring out the format and the structure, and you don't care about that. You want to read the news. So, New York Times, Washington Post, they keep the same structure. They keep it the same. So, all your energy is to understand the content of the news that you're trying to read. So, the same thing with the storytelling and the dashboard, find a way that the words, the structure, the metrics that they use and they're familiar with, so they don't spend time and energy trying to understand your dashboard, your story. That's something they are familiar. The only thing they're trying to understand is the story you are telling, right? Like, okay, the metric that you're tracking is going this way, is going this other way. We might need to take action on this, and be very specific.

Don't say a lot of things. Less is better. Be clear and concise, but be very specific on the things that you really care. Repeat throughout your session and throughout your dashboard, and be very specific about the action plan, what you need about them. And silence doesn't mean agree. You know, silence can be, I acknowledge that, but it doesn't mean that they agree. So, watch them, you know, look at their expression, and next time you go, repeat.

And the one thing that we, another mistake we make is that we go back, you know, let's say you presented something last month, and you're going this month to present the same thing. Recap, it said, last month we talked about this, now this is happening, and now I need you to know this, or do this for us, right? We need your support. And don't go there thinking that they will, they remember everything that happened last month, they know everything that is happening, you know, in your work, and assuming they know that, because they don't. Whatever happened last month, there's been so many things, like the kids, right? So many things have happened in their life that they're not going to remember, you know, what you talked about last time. Don't feel offended. It's normal. Go back to that and say, last time we spoke, we did this, now we're doing this, now let's bring it to the present.

Oh, yes, thank you. Oh, yeah, we need to do some recaps in the system. Thank you so much.

No, thank you.

Yeah, thank you, Aimee. I love that. I have a tendency to think like, oh, we're going to remember what we talked about last month there. I think that, I love to stay on this topic too, I love to stay on this topic too, because I see Nessa, you had some great thoughts that kind of tack on to this idea as well. I didn't know if you wanted to jump in as well.

And I'm, I don't mean to be putting people on the spot. I love when people just share ideas in the chat too. But Marlene, feel free to jump in with your thoughts on the topic too.

Sorry, which thoughts of those?

Well, no, I just see, like, I'm trying to read at the same time the conversation happening in the chat around this, like, user experience for the dashboards and thoughts that you and Nessa are sharing back and forth. So, I'm seeing if you wanted to share that.

I think, so, there's like two components, right? What you show and what you say. And Aimee's hitting, you're, I'm just so appreciating, I think, I think these things, but hearing them out loud from somebody else with an organization that has their stuff way more together than mine. That's not a knock to my organization. It's just a hundred years old and they do things like a hundred-year-old organization. But the doing, saying the things you need to say is very important. So, Aimee's hitting that right on the head. I struggle with, I want to do everything and do everything well. And I had to realize a couple years ago, I can't do everything well, right? So, I have to lean on people to help me with the visualization bit. And so, that was the bit I put in the chat. So, two people that can really help you is usually somebody in the marketing department who really have an eye for user experience, how something's laid out visually and how it flows. I get that guy to look at it. It could be a woman in your organization or somebody, you know, non-binary, but the marketing guy, the UX guy, I ping him at least two, three times during a project to make sure, like, does this need major overhaul? Is this okay? Does it get the story across? And then, I get somebody with finance experience. Doesn't have to be a big wing in finance, even somebody who's just in the accounting department or finance. They could be different departments in your organization, but somebody who can kind of see the numbers beyond the data, right? So, this trend, this upward trend, well, what does that mean to you? Like, oh, well, yeah, I could see that we might want to spend money there or we need to cut costs there, right? So, if they can extrapolate that easily from what you've done, I think you've done more than half the job. The rest is just the speaking. And then, following, I want to say, like, Emmy's kind of book of, you know, keep it simple, stupid, right? Just keep the bullet points or, for those of you, I might date myself here, dragnet, right? Facts, ma'am, just the facts. So, just giving the facts and the takeaways, I think, really, really helped. So, teaching a bit of a master class here, Emmy, like a TED Talk. I kind of want to just bottle you up and take you everywhere. Thanks for your insights.

Oh, no, thank you. Thank you. Thanks for the kind words. And you're right, you know, nobody knows everything. You need to check with people. Just be careful because you want to get it perfect and you want to check with everyone. You might get a paralysis, right? Because you, well, I need to check with the user design, to check with this, I need to check. And then, you keep checking with things and trying to make sure everything is perfect. And then, you know, you never get it out. So, I used to have my, I was an entrepreneur for a little bit, had my own company. And I think that helped because when you're an entrepreneur, you don't have money. You need to get things out and start making money. So, that pushed you to, well, I need to get something else. I can bring income. So, think about the same as a data person, a data scientist. You want to get something else. So, start producing money for the company. You're going to make it better because you need to have the right diligence in terms of security and accuracy and all that. But think about like, you know, yeah, you don't want to rush things too much, but you got to get the money. So, check with people, make sure that everybody thinks the same, but keep it moving. Prototype it, get it out, ask for some feedback and keep making it better.

check with people, make sure that everybody thinks the same, but keep it moving. Prototype it, get it out, ask for some feedback and keep making it better.

Development methodologies for data science teams

Thank you. AJ, I see you had asked a question earlier on different tools. Do you want to jump in? Yeah, I was just curious for your development process, whether you use like a traditional Kanban way of producing content, such as analytics dashboard, or you follow all agile approach where you treat like one dashboard as its own digital product and continuously, you know, every PI make improvements to it. Because I've seen both in the past, their pros and cons. I was just curious if you had opinions on what one works better than the other.

Yeah, I think at one point we tend to go like a full, you know, agile and follow the methodologies used by IT. I don't think that was the best. I think that it has to be, we follow CRIBS DM methodology, but we realized that data scientists and the teams, they don't like the daily standoffs and the full, you know, agile concept. You know, they need time to sit down, think through the problem, you know, work on that. And so we give them that space so they can focus and spend quality time on the project. We try not to bug you a lot and kind of protect them. So it's again, CRIBS DM, a little bit of DevOps. You know, of course, when we implement the solution, we work with IT and we follow their methodologies, their agile DevOps and all that. We use DataOps. We started, you know, we started implementing DataOps for the data part. And now we have MLOps for all the model monitoring is in production.

So I think, you know, I will go with this, building those parts and giving them the freedom to work together. And that has proven to be the best, not try to go all in with agile and traditional IT technologies. Of course, the things that probably they hate the most is all the waterfall and PowerPoint status updates and things like that. They really don't like that. Of course, we need to share with management updates and know what's going on, but we don't we don't force them or we don't make them build this beautiful PowerPoint presentation to keep management happy. It's a lot of, part of my job is trying to protect them, to give them the freedom and the time to focus on what they do best. That is not PowerPoints or, you know, having, you know, apps and things like that, but to sit down and build algorithms.

Career advice for transitioning into data science

Thank you. I see you have a fellow Nationwider on the call too. Francisco, do you want to jump in with your question? Yeah. Can you hear me OK?

Yes. All right. Hi, Emi. My question was just a basic, like general advice for transitioning into like a data science, analytics focused role, specifically for someone coming from somewhere else. Like personally, in my experience, I'm going, I actually just found out I'm moving towards like that data and analytics role and like a recent team shift where I'm coming from like operational support. And I don't really have much experience within analytics and data science, studying operations in college, where my internships were mostly focused on that. So I was wondering from your perspective as a leader, what advice and insight do you have for someone who's brand new like myself?

Yeah. Don't try to be a data scientist, you know, from scratch. It requires a lot of expertise, a lot of knowledge and experience and education. A lot of people, they go and take online course, which is great, or they build up a little toy project and they want to go into data scientists and they don't get the job and they get frustrated. So start with something, data analyst, start with something, business intelligence, you know, supporting an area with some dashboard, some analysis, even if it's in Excel, build those projects that can get you started. And then think about like how much passion or where do you want to go? Like if I think about myself, I was passionate about the topic, I've been learning, you know, always trying to learn new things, but I knew myself that I will not be able to compete with the data scientists, people that have PhDs, they're mathematicians, they're really good at what they do. If I would have gone into that track, I would probably have failed and not be as competitive and as good as they are. But then I thought about like, well, what are the things that I think I do best? And the things that, it's like you have a new product, right? Go back to being an entrepreneur, you are your own product, you are your own brand. And think about like, you will go in the market and compete versus Apple with the iPhone, you're not going to make a good deal there. You're not going to be able to beat the Apple and be better than, you know, build something better than the iPhone. So instead of that, I say, what is the gap? What is that thing that is missing? And then you'll say, oh, but this is not enough to do this. Then build that app that will, you know, make iPhone better. So think about that in your organization or in the industry. And what is that, you know, data scientists, yeah, we'll never have enough, but there are a lot of good people, very smart, but maybe think about maybe in my organization, there are not a lot of very good data engineers. And not like that, maybe start being a data engineer because there is the need, you know, there's the market need, again, being yourself a brand, a product, and I don't mean it in a bad way, but feel that need, understand that need and try to fulfill that need. And then continue learning and continue, you know, figuring out what is your career and where do you want to go. So I hope that helps.

Continue learning, don't get desperate, but don't try to, you know, get pushed to something that, you know, might not fit with your career progression or what you can really do in the short term and don't get frustrated that there will be someone or a position that might help you.

Go back to being an entrepreneur, you are your own product, you are your own brand. And think about like, you will go in the market and compete versus Apple with the iPhone, you're not going to make a good deal there.

Yes, thank you so much.

What to look for in new data science hires

I think on this topic as well, there is a kind of related, there was a Slido question. Is there anything that you or your company looks for in new data science hires? So on both the technical and soft skills?

Yeah, and I put it on the chat, my link, and if anyone wants to connect, happy to do so. I think, you know, there's definitely the math, the technical expertise, the coding, but being, you know, being very analytical, like asking questions, be someone that cares about the business, someone that is passionate about that, that asks good questions, think through like problem solving, you know, think about the problems and what is the thing that we really need to solve. So I think it's a combination and some people are better at one thing than the other, so try to fit, you know, again, your profile to whatever whatever the spectrum that can be. If you are very, you know, into the technical thing, the math, you can do that, but if you are more on the analytical or the business translator, do that. But, you know, we are very rigorous into data scientists and we have a very strict, you know, hiring process. So that's, you know, that involves a lot of the technical expertise. And depending on, you know, like if it's an LP, computer vision, you know, whatever that might be, we have a specific process. But again, don't feel discouraged if you don't have all of that, maybe find that role that fits your needs.

Thank you. I know we just hit the top of the hour and a few people have to drop and I want to make sure, do you have a meeting you're running to, Jemi? I don't, but I need to disconnect quickly. So I don't know if any last question, comment. There was only one last question, if that's okay, one that we had missed. And thank you for sharing your LinkedIn there for people to connect and follow up as well. But the question was, with so many data scientists, I assume a polylingual team, and you talked about all the different languages people use, how have you navigated the R-Python dichotomy, especially with respect to deployment solutions?

Yeah, our model factory environment is suited for both and other languages. So we, everything we build try to be, you know, to be flexible with whatever programming language is used, especially R and Python. The people that are experts in R or that's in Python, but a lot of that, you know, we as an insurance company, financial services, we're very heavy on R, but this transition in, you know, probably 70, 80, 20, I think about 50, 50, but moving towards Python. So people are trying to learn both and they prove it to be more successful if they know both. So not easy and difficult to have, you know, expectation to be an expert in both, but if you build expertise in one, but then know how to do at least the

basic on the other language, that can be extremely powerful. Thank you so much. I think we got to everybody's questions. I'm sorry if there was anything that we missed, but thank you all so much for joining. Thank you, Emi, for all your great insights. This was great. And I see, if you can't see now, there's a lot of people in the chat saying thank you right now as well.

Thank you all for joining and I appreciate the time and happy to connect and let me know if there's anything I can do to help. And thank you for having me. Enjoy Florida over there. Absolutely. Thank you. Have a great rest of the day, everybody.

Thank you.