image: thumbnail.jpg
Data Science Hangout | Rebecca Hadi, Lyn Health | Transparent & Visible Work
At 19:11, Rebecca shared her approach to communication with transparent and visible work.
Here are 4 things they do on their team to create transparent & visible work:
1. Starting off with an open Data Science channel (different from their team channel) that’s open to the company.
For example, if we do an analysis for the sales team - we will post it in that channel and tag the sales team. That way anyone in the company can see the work. We keep conversations in the channel, rather than DMs. A lot of times someone else may be interested in that work.
2. We also have a Google sheet that we call our insight repository. We have that pinned to our channel so when we post analysis, we also put them in our insight repository with one row per analysis.
For example: in-patient mortality model, a headline, the link to the PDF, and the GitHub link. The insight repository has been very well received as a place that someone can start when they have a question.
3. I also host a weekly data science prioritization meeting and we use Monday as our project management tool.
This has been really helpful because in the past I’ve had some challenges with balancing requests from our clinical team versus sales team and understanding the priority. It can be hard to be that person in the middle. It’s a lot more productive when we can all have that conversation live. The sales team might say, “I need this for XYZ meeting” and you might have to say, “Ok, you know that’s going to push off these additions to the clinical dashboard.” Then the clinical team can weigh in, “We’re okay with that because of XYZ reason”
4. I’m also a really big advocate of a monthly all-employee call and I try to get air time there pretty frequently.
Initially it was level setting on “this is what data science is and how my team operates.” That has been successful for setting this idea of, “you are going to ask me something and I’m going to ask you the why behind it.” This helped set the precedent ahead of time where I’m not challenging an idea, but it will help my team produce better work if we understand why you care about this and how you are going to use it because that gives us context in decisions we need to make throughout the process.
A few resources shared too:
Javier shared: A resource that I just came across last week (def not something I do daily). if you're wanting to collect / manipulate JSON data with R, Tom Mock has a super-easy-to-follow writeup with several different approaches: https://lnkd.in/gwGxJpcW
Luke shared: Data is Plural newsletter is a decent off-the-beaten-path dredger of data sources: https://lnkd.in/grRGqQRA
Rebecca shared this book on management, The Hands-Off Manager: https://lnkd.in/gg2TU2mB
Where to find more?
► Subscribe to Our Channel Here: https://bit.ly/2TzgcOu
► Data Science Hangout site: rstudio.com/data-science-hangout
► Add the Data Science Hangout to your calendar: rstd.io/datasciencehangout
Follow Us Here:
Website: https://www.rstudio.com
LinkedIn:https://www.linkedin.com/company/rstudio
Twitter: https://twitter.com/rstudio
Transcript#
This transcript was generated automatically and may contain errors.
Hi, everybody. Welcome to the Data Science Hangout. If you're joining for the first time, I'm Rachel and it's great to meet you. Today actually marks a full year of the Data Science Hangout. And I just want to say thank you to everyone for making it what it is and for hanging out with us. From that first one, which was kind of an experiment of seeing how things were going to go to what it is now. This is always a favorite part of my week now.
We are going to take a break from hangouts in July as we prepare for the RStudio conference and some big announcements. So let us know in the chat too if you'll be there at the conference. We'd love to see you all in person as well. But we'll be right back with normal hangouts right after the conference on Thursday, August 4th as well.
If this is your first hangout, this is an open space for the whole data science community to connect and chat about data science leadership, questions you're facing, and what's going on in the world of data science. These sessions are recorded and shared to YouTube as well as the Data Science Hangout site. So you can always go back and rewatch or find helpful resources there too. We also have a LinkedIn group I want to make sure I mention if you ever want to continue certain discussions there. And Hannah can share it in the chat as well.
But before we get started here, there's three ways that you can ask questions. You can jump in by raising your hand on Zoom. You can put questions into the Zoom chat and feel free to just put a little star next to it if you want me to read it. Or I can call you to introduce yourself and add some context too. But lastly, we also have a Slido link where you can ask questions anonymously as well. And Hannah will share that as well.
I think you just shared it right when I said it. But we always want to create spaces where everybody can participate and we can hear from everyone, no matter your level of experience or area of work as well. But with all that, thank you again for joining us today. I'm so excited to be joined by my co-host today, Rebecca Hottie, Head of Data Science at Lynn Health. And Rebecca, I'd love to kick things off with having you introduce yourself and sharing a little bit about your work and introducing your cat here as well.
Yeah, absolutely. Thanks so much for having both myself and my cat Jinx. She was really excited for the hangout today, clearly. So thanks so much for having us. Yeah, so my name is Rebecca Hottie and I'm the Head of Data Science at Lynn Health. Lynn Health is a early stage startup and we're focusing on serving members who have multiple chronic conditions and building a health care experience that really supports the whole person.
Buckets of data science work at Lyn Health
So from a data science perspective, my role and the work my team does kind of buckets into kind of three or four different areas. So at a high level, we're a very small team of a data engineer in my team, and we work together to build out our data warehouse. And that's where we ingest data from both our clients, our own electronic health record system and make it kind of easier for access and also trying to build out self-service capabilities for other other teams. So we aren't quite so much the bottleneck. So that's one area.
Another area is sorry, my cat's like all over my girl. Another area is trying to support identification of members who might be a good fit for our program. And we do that by building algorithms to measure something called impactability. So in health care, you can imagine that, you know, you hear these horror stories of, you know, hospital stays that are thousands of dollars and very expensive medications and things like that. But not all of those things are like impactable, you know, like some of those things you couldn't really have done anything about. So we're trying to take the transition from just someone's risk, like how like, you know, kind of traditional health care space, it's like a risk score, you know, like what is your predicted spend? But we're kind of trying to shift the perspective to what's your impactable spend and using that as sort of a main signal for who do we think is a good candidate for our program who has the most spend that using our methods, we think that we can impact. So that's another vertical.
And then the third one is building internal risk adjustment models to help us normalize comparisons between different physicians. So we're focused on looking at outcomes. So one example is inpatient mortality. So we built a risk adjustment model that is able to sort of predict when an inpatient mortality will happen with a high degree of accuracy. And we can use that to adjust for providers case mix to say, you know, just because they see a sicker group of patients, we don't want to penalize them for that. But use that to be able to detect statistically significant variation in the in their inpatient mortality rates as an example. And we use that to inform who we want to refer to. So we want to send patients to doctors who have statistically better outcomes.
And then the last one is more of a kind of sales support focus. And we support our sales team. And one of the favorite things that we did was we built a R Shiny interactive sales tool for them. We got some, it was initially built as an Excel model, but we got some feedback that, you know, we'd love to kind of have a slider that we can do live in meetings. And R Shiny was a great tool to achieve that. So we have a like shiny server instance that we host that tool on. So I've just talked a lot, but those are, that's a little bit about the different buckets of work under my scope at Linhelm.
Awesome. That's really helpful. And, and what is your data team look like today?
Yeah, it's a pretty small data team. So it's actually myself and one data engineer. So my role is, you know, it's a lot of hands on keys, individual contributor work, in addition to supporting, you know, the growth and career of my data engineer and making sure that, you know, he's able to learn different things and build relationships with the business. And then also that, that strategy component of, you know, what, what strategically will help the business? What does our roadmap look like?
Being one of the first data hires
Well, we, yeah, we've talked about that before on Data Science Hangouts, how, when you're the first data person, there can be, there's a lot of decisions that you have to make, and sometimes can be a little scary. But how do you handle being one of the first data people there?
Yeah, it was definitely a bit of a learning curve. I think that I think one of the biggest challenges I faced was initially just not having any of our own data. So for a while, when I first started, we didn't have any sort of products or systems that were generating data. So a big thing that I had to do was kind of search for what publicly available data could we leverage to answer some of the questions that we have. So that was definitely an initial challenge. And then also kind of more of an opportunity as a first data person was just helping the organization see the value of data science and really being an advocate for, you know, using the work that myself slash what grew into my team produces to really drive decision making. So but, you know, to be honest, it can be a little bit isolating. And I've worked previously at larger organizations where I got Nordstrom, for instance, there was a large center of excellence. So I was a part of a community with, you know, I think it was around 100 people of different data science analysts, engineers, so it can be a little bit isolating. But that's why I love events like this, and other things in the community where I can still have those technical relationships, and you know, kind of talk to other people. So that's a little bit about what it's been as a first day to hire.
Absolutely. Thank you. Ishwar, I see you put a question to the chat. Do you want to jump in and ask that?
Hey, hi, everyone. This is Ishwar. I currently work at Nissan supply chain as a data scientist. So yeah, I'm really excited to, you know, know about the how data science journey of Rebecca Harris has started in her organization.
And I was in the same situation when I started at Nissan, specifically in my department, my team. So I had, I had a lot of challenges would be a lot of unstructured data, like XML format and super unorganized, and then not having anyone to help for on the data engineering task and getting those data from, you know, multiple places into the organized structure, and then at least data, scraping from XML to structure tables, and don't even know where to store the data. So those were the initial phases. And I was, I didn't even have a data engineer.
To start, I was just, you know, dead end that I don't have anywhere, anywhere to go, kind of stuff. So, yeah, I totally, you know, feel your, how you might have started your data science journey. So just trying to understanding even more, like, how did you at least kickstart your journey, at least to prove the, you know, proof of concept to the organization that, hey, data science, if we grow from one data scientist and one data engineer to the larger data science whole team, then definitely the organization may benefit from that team. So what was the challenge that you faced to go and get started with your proof of concept? Just wanted to, you know, understand in high level your journey to grow up and then make, creating the trust within the organization.
Yeah, absolutely. So you're trying to understand, you know, how did I sort of convince the business about the value of data science, that it's sort of worth the investment? Of course. Yeah. Okay. Yeah. That's a great question. I would say that I'm still very much on that journey. You know, just given with the economy sort of being what it is, we've had to make some pauses in some of the hiring that I was hoping to do. So, but it's, I think it's less about, I feel like I've done the work to sort of convince the business that it's worth the investment. And now it's more just like, okay, when's the right time to sort of make that investment just based on some of the economic conditions that have shifted.
But I think in terms of like how I built that trust, I would say probably the first project that I did developing our pricing model was very well received by the business. And that was because it was sort of this vague question of, you know, how, like, how are we going to demonstrate like, how do we, like, where do we think that we're going to have the most impact? And also how do we do that without having any of our own data in-house? So it was a pretty significant literature review that I did trying to understand, you know, the AHRQ has a lot of really great research papers about the challenges of people who have multiple chronic conditions, they're over-indexing and inpatient admissions and ED visits as well. So there's that like utilization component, but then there's also just frankly a price component. There's significant variation in price of different hospitals. And so how I was able to kind of convince the business of kind of the worth of my team was building something out of nothing, right? And so it was sort of like, if I can take these very distinct kind of publicly available, like research papers and build a cohesive story where it has to be based in reality and it has to be reasonable, but building that into an actual savings model that I, and I think a big part of it, you know, kind of jumping around in my answer, but I think communication is a really big part of it, honestly, and driving adoption of the work where, you know, I could have taken the approach where it's sort of like, you know, I'm going to go off and build this savings model and I'm going to come back and yeah, there you go. But I kind of took a different approach where, you know, I brought them along throughout the process where, you know, I met with them like every week or so, and kind of said, you know, this is where I'm at. This is what I'm seeing. This is what it's looking like. And then getting their feedback at that point, and then integrating that feedback as I went along. And then the final delivery of the product where, you know, I walked them through it at a very detailed level and kind of answered questions. So I would say it's kind of about that, like bringing them along throughout the process and creating that visibility into the work kind of rather than, you know, I kind of sometimes have the urge to like put on my headphones and go in a cave for two weeks and like just kind of bang something out. But I have to challenge myself to really like bring my stakeholders along that helps them develop ownership. So I would say that that's, that worked really well for me.
I think communication is a really big part of it, honestly, and driving adoption of the work kind of rather than, you know, I kind of sometimes have the urge to like put on my headphones and go in a cave for two weeks and like just kind of bang something out. But I have to challenge myself to really like bring my stakeholders along that helps them develop ownership.
Yeah. Did you have any challenge? Like I always face this challenge from the, you know, IT infrastructure team, letting a lot of security restrictions on tool sets. Like I want to use the Anaconda Navigator, you know, whole set of RStudio and Spyder and Jupyter Notebook and so on. And then another challenge is not having any pre-established like data sources, let's say Oracle, a lot of companies have Snowflake, AWS, whatever that is. But me not having those infrastructure in the beginning, did you encounter like those kind of issues? Like you were trying to convey the IS team and then it's hard then to explain why do I need it? And then like it was the major
challenges. Did you have to face anything in short of that or you're in the beginning phase of it?
Yeah. Yeah, I would say that I haven't really faced challenges from our IT department in that respect. It was a while before we onboarded any sort of like detailed, you know, like information that contained PHI or protected health information or personal information. So it was a while before. And kind of at that point, the IT organization was kind of pursuing on this like SOC 2 kind of audit track. And so I think maybe the reason why I didn't have that challenge was because I have a really great relationship like with our CTO, with our IT person and kind of just, again, kind of goes back to communication, like kind of telling them what I anticipate like the needs will be. And so they were fine with us kind of standing up a Redshift instance.
I would say the one challenge that we had is that we were initially planning to use Tableau online as a data visualization, like kind of hosting it. But we had some confusion on whether or not that was HIPAA compliant. We had done some research that, you know, Tableau online was HIPAA compliant. But then when we actually went to sort of go push data up there, we realized, well, no, actually, because it's like Tableau hosting it, like their backend wasn't HIPAA compliant. Whereas like, but we didn't have the IT resources to stand up our own Tableau like server instance. So we had to pivot to using Amazon QuickSight just for kind of like this dashboard that we were building. So that was definitely a pivot where we didn't have the budget to or the like resourcing to stand up like a Tableau server, even though for me, Tableau is a little bit easier to use the QuickSight.
I see Libby kind of had a question along those lines as well, around HIPAA. Libby, do you want to jump in? Sure, I just realized my camera was covered and not on. Yeah, you see, you kind of touched on it a little bit. My question immediately was like, I also work in a heavily regulated industry, financial services, retail banking, insurance, stuff like that. So I know that sometimes those regulations can cause a lot of barriers to just being able to do the work.
So what have you experienced around that? But it sounds like you guys are still in some of the early stages where like maybe you haven't hit some of those barriers at scale yet.
Yeah, I would say we haven't really hit that scale. I think the main barrier that we hit was the Tableau one that I just talked about. But it is something that we're very mindful of when it comes to installing different Python libraries that we use.
It's kind of on our roadmap. I think that's called SNYK. I'm really terrible at pronouncing things. There are kind of tools that we know about that can kind of help with that security audit from a library perspective. So I think that in the future will be helpful to us. But we're just kind of, it's always present in our mind. We don't really take that for granted. We're always thinking about how is this going to impact security if I use this kind of like random thing or push data in a certain place. So I would say we're very cautious about that. But we haven't, kind of to your point, we're a little bit on the earlier side to experience those challenges at scale. Awesome. Thank you.
Transparent and visible work
Great question. Rebecca, I love the point you made about communication too. And just like even when you want to go work for two weeks in your cave with your headphones on, making sure to have regular check-ins with the team. But I was curious what that actually looks like in practice. Do you just keep sharing with the team or just have regular check-ins scheduled?
Yeah, I would say that my approach has evolved over time. We're a pretty like SOC-based company and I'm a big proponent of transparent work and visible work. So a couple of things that I do, we have a cross, like a channel, a data science channel that's open to the company. So we have like our team channel that's more about like, I'm going to be late today or whatever. But we have an open channel where anytime that we, like let's say that we do an analysis for the sales team, we post it in that channel and tag the sales team. So like anyone in the company can see that work. And we kind of try to keep conversations in that channel versus kind of DMs. Like we'll route people back to there, like let's have that conversation. So because a lot of times someone else might be interested in that work. So that's one way. And then the other thing that we've done is we've created just like a Google sheet that we call our insight repository. So, and we have that pinned to our channel. And so when we post those analyses, we also put them in our insight repository where it's just like one row per analysis. That's like the analysis wise, like inpatient mortality model. Here's a link to the PDF. And then here's like kind of the insight or like the headline and then a link to the GitHub, but that's just more for our team. So that insight repository has been really well received as a place that's like, okay, if you have a question, like start here and you might just be already to have that.
And then kind of the third, I guess like maybe there's two more things that I do. I host a weekly like data science prioritization meeting. We use Monday as our project management tool. I've used air in the past, but I actually like Monday. It was kind of new for me to use, but it's really snazzy. And there that's been really helpful from a prioritization perspective, because I've had some challenges in the past, struggling requests from like our clinical team versus our sales team, like what's the priority. And it can be kind of hard to be that middle person that's like, sales team won this time. So like, sorry, clinical team, I'm not going to be able to do that kind of, not like a shoot the messenger situation, but it's a lot more productive when we can all have that conversation live and say this, sales team might say, I need X, Y, Z for a meeting. And we might say like, okay, that's going to push off these additions to this clinical dashboard. And then the clinical team is like, okay, we're okay with that because of like X, Y, Z reason. So it's just helpful to have that like kind of prioritization meeting where there's the list of work and people can see it. So that's really helpful.
And then the last thing is just kind of having, I'm a really big advocate. We have a monthly like all employee call and I try to get airtime there like pretty frequently to just kind of talk about, you know, initially like a level set of like, this is what data science is and how my team operates. And a big part of that that's been successful for me is kind of setting this idea of like, if you ask me something, I'm going to ask you like the why behind it. And I've had instances in my career where people can get kind of defensive when I ask that question, you know, like I'm, and maybe it was just my historical wording of it, but it really helped to kind of set that precedent ahead of time where it's, I'm not like challenging you or this idea, but it will help my team produce better work. If we understand like, why do you care about this? How are you like, how are you going to use it? Because we can, that gives us context and, you know, decisions that we might need to make throughout the process. So that was a really long answer, but those are some of the things that I do to kind of help with communication and building relationships.
I'm not like challenging you or this idea, but it will help my team produce better work. If we understand like, why do you care about this? How are you like, how are you going to use it? Because we can, that gives us context and, you know, decisions that we might need to make throughout the process.
Building and structuring the data science team
That's incredibly helpful. Thank you. I see, David, you just asked a question and Libby plus one'd it right away. So David, do you want to ask that?
Sure. Thank you, Rachel. So Rebecca, my question is really around actually building your team. Do you plan or have you thought through whether you're going to have one centralized group of data scientists that support the entire organization or your group, or do you think about maybe embedding data scientists in those functional groups so they get an idea of that, just that domain and work with it like that? Just your thoughts on that. Yeah, that's a great question.
My current thought about it, I'm planning on starting with a centralized data science team, mainly to just establish kind of like standards and processes around, you know, kind of like we talked about, like from a security perspective and some different kinds of methods that we use. So I'm planning on starting with a centralized team, but then my goal would be that as we kind of continue to build our self-service tools and up-level the scale of our business partners to kind of do some of their own like initial analyses that in the future. I really like the idea of the embedded model where the person sits under the data science team and can benefit from that technical mentorship and, you know, knowledge sharing resources, but then are aligned to a specific vertical because I think domain knowledge is just so important and that it can be hard when you're working on a bunch of different things to really get deep in that. So once we grow, I'm a big fan of the embedded model for just building that, like developing those team-specific relationships and that team, kind of team-specific domain knowledge. All right, thank you.
I see Luke just said huge hub and spoke advocate as well.
Oh, I'm sorry, I finally figured out how to use my microphone again. Yeah, especially in healthcare, like there is a big, there's a big need for like a hub and spoke model, I think.
I know personally I'm the spoke for my department and it's almost like you need that person to be like the liaison between the central group and, you know, what and you can be as the spoke also needs in the department. So it's one of those in healthcare, I would really advocate for that.
I see that Blake, you had asked the question a little bit earlier about Redshift. Do you want to jump in?
Yeah, let's see if I can get this, I don't know if my video is going, but can you hear me on the audio check? Yes, I can. We can see you too. I suppose it's a little bit more technical than the managerial and the team structuring and some of the organizational hurdles, but I noticed you mentioned that you went Redshift. So are you in a specific cloud environment like AWS right now and is that what led you to Redshift? And I was just, I don't have a lot of experience with Redshift. I was wondering what some of the trade-offs were with that versus other database technologies and whether or not you have a cloud hybrid or if that had something to do with it. Over. Yeah, that's a great question.
The reason that we chose Redshift was because both myself and my data engineer came from Nordstrom and in Nordstrom, the database that we had used was Redshift in an AWS environment. So that's not to say that it's because it's just what we used before, but the benefit of that was speed and not having to learn sort of like a new SQL flavor or we've already had a lot of experience with like EC2 instances and moving data between S3 and Redshift and those kind of different types of processes. So I would say that the kind of main deciding factor was just that kind of speed and we were already familiar with it. So the learning curve, there wasn't really a learning curve and we could get to the work more quickly. If that hadn't been the case, I think we probably would have done a bit more exploration on different tools, but that was really kind of like the driving factor of that decision.
Cool. If you're okay with that, I have one piggyback question. On your initial like upfront investments, so with getting your company to buy into the cost of the cloud environment, was that a challenge and was that an upfront investment or did you have to kind of prove your worth before they decided to hop on those costs? Over. Yeah. Our CTO is very much on board with the cloud technology, so that wasn't so much of a hard sell, but we actually were able to take advantage of like an AWS Redshift trial. So we got the first two months of our instance free and in that time we were able to produce some different sorts of analyses that really sort of drove value for the business. So then after that, you know, it was kind of an easy sell to say, you know, we need to continue to invest in this. And I also think that recently we were approved, applied for and approved as part of like an AWS startup that has, you know, like a certain amount of AWS credit. And so kind of taking advantage of those resources also helped to build the case, if that answers your question. Absolutely. Thank you so much.
Thank you. Sep, I see you put a question into the Zoom chat too, if you want to jump in next.
Or Sep, I can read your question if you want to add more context after. I didn't have my microphone on. You could tell that I was animated. I knew you were here. Yeah, exactly. Thanks, Rebecca, for joining and taking the time to answer questions. I personally get excited because it's someone that's in a similar field. And so my question is a little bit more probably specific, but I was just curious, you know, you guys have built your own EMR. How do you deal with like the structured versus unstructured data that's in your EMR and, you know, if possible, like changing scope over time, if you've gotten to encounter that yet? Yeah. So we use Athena Health as our EMR. And one of the services that they offer is called Dataview, which is a snowflake, like database instance that they manage. And so that's actually made it like pretty, like easy for us to access that data because they're kind of like build like a database structure. And so we're still, oh yeah, sorry. Yeah. Electronic medical record. So the data that we've accessed from that so far has mostly been structured data. And we actually, one of the kind of tools that we built was what we call like a PRM or patient relationship management. So you can kind of imagine like
the concept of a CRM, but, you know, CRMs are pretty expensive. And so we built this sort of homegrown solution, leveraging data from Dataview and Snowflake has really nice Python connectors. And we were able to automate that. We use Daxter as our sort of like workflow automation tool. So that was sort of like a really big win to be able to automate like the pulling of appointment data and patient data sort of like into this kind of spreadsheet that then our clinical team can use. Whereas before they were having to manually export CSVs every night and copy and paste it. And so I would say that we haven't really dealt too much with unstructured data from the EMR, but that's something we're still very much exploring because like for the note specifically, like there's a lot of good information there that we're going to want to extract around like social history questions and things like that.
I would say when kind of earlier on, we explored some different partners like Particle Health and like Redox that were more about leveraging data from like the network of electronic health record systems. So like if I have a patient, you know, how can I get data from providers that they've already seen before in an automated way? And that's where I dealt with more of that like unstructured data because kind of the trial, we just worked with like a trial. We ended up not pursuing that at that point, but that was more of like the JSON like nested structure. And so that was definitely a learning curve for me kind of like working with that and trying to pull out what I needed, but we haven't had a like very big use case for working with the unstructured data yet.
Cool. Thanks for answering that.
Another cat sighting, Sep. Where are my cats? Patrick, I see you have your hand raised. You want to jump in? Sure. Thank you, Rachel. I appreciate that. And Rebecca, thank you for taking the time to talk to us about this. Similar to Sep, I'm coming from a similar content area, and so I'm going to ask a bit of a nitty-gritty question. I'd appreciate an answer, but I don't want to take up too much time with this because I don't think it's probably all that applicable to everybody else. I'm wondering about how y'all handle your record linkage and deduplication for patients across EMRs, across data sources, specifically about checking the quality of those things. We're working with a couple of health information exchanges, and we need to verify quality of their patient identifiers, and we've tried a bunch of different things.
I'm sure you know there are myriad problems with any single approach to this, and so I'm just wondering what y'all do and how you handle that particular problem. Yeah, that's a great question. I would say that we're not at the point of maturity yet where we're really leveraging data from the health information exchange, so we haven't had to do a ton of mapping. The main mapping exercise that we've done is making the connection between the claims data that we receive from our clients, which has a member ID, like a health plan member ID, and the patient ID that we have from our Athena instance. We're using logic that's first name, last name, date of birth, that kind of matching, but we haven't had to deal with the challenge of leveraging data from the health information exchange, so that's probably not super helpful.
In a past life working at Nordstrom, customer identity management was a huge problem that required a lot of work and skepticism about the customer ID, and different databases use different customer IDs, so if you ask someone how many customers do we have, and if you used a cust key, you got a certain answer, and if you used a wallet, you got a different answer, and so that was a really big challenge, mainly from a stakeholder management perspective, and it was kind of building that data literacy of, you know, if I'm asking about customers, like, do I care about customers that we can actually, like, send an email to, and we have high likelihood that we know who they are, or am I okay with a more probabilistic assignment to kind of have, you know, maybe capture people who we aren't able to target them directly, but, you know, we think that they're a distinct customer, so my answer is probably not super helpful to your question. No, thank you. It is just good to know. Sep, thank you for the recommendation. I did not know Datavon, but I will look into that, and yeah, Rebecca, when y'all venture more into that, and I'm sure that is coming somewhere in your future, I wish you much luck, and I'd be happy to talk. It is a challenge for sure, but I appreciate just sort of the feeling that y'all are thinking about this, and that I guess we can explore, like you said, Nordstrom customer management is a similar kind of problem, so maybe we should look a little bit more broadly than we have been, so thank you. Yeah, great question. Awesome, thank you.
Advice for breaking into data science
I see there's a few anonymous questions coming in on Slido, too, so just a reminder, if anybody wants to ask there, we'll share it in the chat again, too, but the question was, I'm a recent graduate, and I'm hoping to become a data scientist in the pharmaceutical space. Do you have any tips on how I could help prepare for interviews? Yeah, that's a great question. The best advice that was ever given to me was that data science is a contact sport, and what they mean by that is that the best way to learn is by doing, and you know, my recommendation for trying to kind of break into the field would be, especially if you know which domain, you know, find a data set that's on the internet. Kaggle is a really great resource for public data sets, and just do something with it, right, like it can be as simple as an exploratory data analysis. You can take it further and do some kind of like predictive model or like an inference type of task, and then put that on a GitHub profile, and if you kind of want to go above and beyond, like I have a personal website where I, you know, it's just a Google site, so it's not any fancy web development, but I write about the work that I do, and I link it to my GitHub, and I have that on my resume, so I found that that really helped to, has helped me in my career, and it gives me something to talk about in an interview, and so that would be my recommendation is just find a data set that you're interested in, whether it's related to pharmacy that you're interested in breaking into, or you know, there's lots of like, there's like a, where I live, there's like a pet license database, and you can just do like different sorts of fun things like that, so that would be my biggest recommendation. Yeah, that's really helpful. I see Deepsha also recommended participate in Tidy Tuesdays. Yeah, exactly. That's a great idea.
The best advice that was ever given to me was that data science is a contact sport, and what they mean by that is that the best way to learn is by doing.
Just want to check over on, oh, a follow-up to that question anonymously was, do you think it's important to like know all the different pharma keywords, or know the keywords that people use in certain industries when interviewing for a role? I would take my opinion on this with a grain of salt. I think it depends on what they're like looking for in that role, like if they're looking for someone to hit the ground running, or to kind of like mentor others in that specific domain, then building some of that domain knowledge will probably be helpful, but my hiring practice is that I look more about like aptitude and like curiosity, so it's like, in my opinion, like it's okay if you don't come in with that expertise, but maybe if you like researched it a little bit, and you could kind of turn that back on the interviewer, like, you know, I saw that there was a certain term, you know, like is applicable to your work, and kind of more of that like curiosity, and like maybe you researched it a little bit, but you don't have to, in my opinion, come in sort of like knowing everything about the domain. It's really helpful, thank you.
Switching industries and embracing uncertainty
I was curious, Rebecca, what it was like for you moving from a retail space at Nordstrom into your role now, and what was, yeah, what was that experience like switching industries?
Yeah, so I actually kind of have a bit of like a healthcare sandwich in my career, because I started out working in healthcare at a couple different health insurance companies, and so I had a lot of experience working with like medical claims data, pharmacy claims data, like enrollment data. I had worked on never building a risk adjustment model of my own, but like leveraging software, and kind of tangential to your question, but like one of my favorite memories of like early in my career was we used this risk adjustment software, I think it was called like DXCG, and I was asking questions to people in the department about, you know, how exactly does this work, you know, what it, like what inputs, like kind of might change the result, and no one could give me sort of like a firm answer, and so I realized that no one had read the documentation on this model, and that was actually an opportunity for me, because I printed, you know, printing was big at the time, so I printed out this like 30-page document and read it, and I actually became sort of the subject matter expert, and I was able to partner with IT whenever we had a version upgrade, because I would read the release notes and know like, okay, we need to pick a sample that has like these kind of qualities, because I expect like, you know, they made a significant change to like neonates or something like that, so kind of a tangent to your question, but then transitioning into retail, you know, I knew nothing about marketing at the time, but it's, I'm just a very like continuous learner, and I love to learn about new things, so I am kind of going back into healthcare at somewhere where, you know, I had kind of had some background in it, but a lot had changed in the time since I had left and gone back, so I think my attitude towards it is just being comfortable with uncertainty, and just being not necessarily attaching pride to knowing everything about a domain, and ask, and just being very open about asking questions and clear like, you know, I don't understand that, can you explain that to me, or I haven't seen that before, and just kind of having that courage to ask those questions made both of those transitions really helpful for me, or even if it's something that you know a little bit about, just being clear about like the amount that you know, like I've heard of that, but I don't know much about it, you know, so kind of that made the transition go well for me, I would say.
Balancing learning with a busy schedule
That's really helpful, yeah, and I see Ravonda commented, always be learning is great life advice as well. How do you balance that in like a busy role where you just have so much going on, and so many tasks, and people asking you for things to still make time for learning?
I think I've gotten better at that over time. I, when I worked at Nordstrom, Nordstrom, oh I see, got your cat now, nice, full circle. I'm going to jam to make an appearance here. That was so cute and fluffy, oh my gosh, I love cats, clearly, but at Nordstrom, they were very like upfront about carving out that time for self-development, and so we would actually include that as part of our like JIRA stories, so you know, if we have like, if there's 10 days in a sprint, we would only plan for like eight points worth of work to allow kind of that extra two days for meetings, and then carving out that self-development time, but then it becomes actually doing it, you know, like if I have my self-development time set up for Friday afternoon, do I just like stop working early and not do it, or like actually kind of do it?
So I think for me, it's kind of about like creating like a little bit of a plan about, you know, what is it that I want to learn, and like, how do I say this, I think for me, like sometimes for me, I get this idea that I have to sit down and like completely learn something in that session, and so if I think like, if I don't have three hours to go through this whole thing, then it's just going to be like not a good idea, but over time, I've learned where it's like, it's okay to start something and not finish it, and that's helped me more kind of learn new things where I'm like not setting super high expectations on myself, or I can just kind of play around with something. I would say the other technique that I have is, I kind of did some self-reflection and realized which part of the day that I'm the most productive. I was, I'm kind of, I'm an introvert, and so I was, I used to have a lot of meetings in the morning, and so I would have my meetings from like 8 30 to like 12, and then I'd have the afternoon to work, but I would just be like so drained from that, and then I'm just like, I don't want to do anything, and really have to convince myself, and so once I realized that, I'm, I'm able to get a lot more done in the morning, and so I moved my, I had the ability to move my meetings to the afternoon, so then I would be able to have that concentrated time in the morning where I could be hands-on either doing a project, or kind of exploring something new, and then having my meetings during that time where I've already kind of, you know, I did what I wanted to do today, and now I can focus on more of that communication aspect, so a little bit beyond the scope of just self-development, but more just about how I'm kind of managing meetings with actually kind of getting things done.
it's okay to start something and not finish it, and that's helped me more kind of learn new things where I'm like not setting super high expectations on myself, or I can just kind of play around with something.
I love that, yeah, I realized that too, and I, there was a book that I read called When, it was just all about timing of things in your life, and I realized that about like Monday mornings, no matter what, are never going to be a good time for a meeting. Yeah, I see a lot of great comments in the chat, but I want to make sure I'm not missing questions in there too. Let's see,
Libby, I see you had a comment on, and lots of apps that you use to track stuff, was there anything you wanted to add in? I think this was a follow-up to a question about where to find data, and it's just a source that like I have found in my life. I'm a serial data collector, I collect data, and I track a lot of things, and I have for years and years, so I kind of realized that some of these apps that I use, they have little export options. If you go in the settings, export data, and it'll email you a CSV file of all the stuff you've tracked over the past however many years you've been using that app, and an example for that for me is I used to track my sleep really habitually with a sleep tracking app that listened to your breathing and stuff like that, and I ended up with like two and a half years worth of daily sleep data, and that's stuff
that you can bring in external variables, right, like the temperature, the precipitation, whatever is going on that day, and then I could also sync that with my school schedule, like was I in school that day, what was the day of the week, like all this stuff, just fun little projects, so look at your apps, what are you tracking, if you're a data person, you're probably tracking a lot in your life.
That's super cool, I love that.
Ravanda, I see you had a comment, this kind of data being very useful for clinical trials, do you have anything you'd want to share on that? Sure, just there have been more trials where people have been getting personal trackers to track things like their heartbeat, and it kind of helps to, like a lot of times one of the challenges in clinical trials is that people have to go to a certain test site, and that's not always possible, especially within the last three years, right, so having these kind of devices and having a patient being willing to help you with that is really helpful, and I think that's something that so having these kind of devices and having a patient being willing to have that kind of data tracked kind of helps you to check in on them without like being as inconvenient as maybe a drive to a trial site would be.
Definitely, yeah I see a bunch of people sharing different tracking apps and data that we can find as well, yeah maybe we all need to work on this together. I love this conversation y'all, keep going, my favorite topic. I know Frank is pretty big on that with Strava as well, I don't think Frank's on today, but Rebecca on the topic of like learning and building teams, are there any resources or books that you found especially helpful?
Management lessons and resources
I would say maybe like less of building teams and more about just management. You know, to be completely honest, I struggled when I became a manager for the first time. I was a really good IC, like not to toot my own horn, but like I was a really good IC, and when I first became a data science manager at Nordstrom, like it was really hard. Delegations specifically, like I would have an ask come in and you know maybe it was like really quick turnaround for some like marketing fire drill, and I would have the thought like it's just quicker if I do it, like I can just do it, and I thought that I was helping my team by kind of saving them from doing that, but then what I realized is that one, I was actually not giving them an opportunity to shine and build those relationships and kind of have that, you know, like I did this thing that was really impactful, and then two, I got super burnt out.
So once I was realizing that, I started looking at different sorts of like management books, like how can I kind of build that skill set, and one book that I really liked was called like the Hands-Off Manager. I can't remember who it's by, but I can like share a link to it, and that really helped me because I didn't want to be a manager that was very like micromanaging, but I also, you know, you have to hold people accountable and get things done, so I kind of was having that balance between you know checking in and making sure things are moving along with not wanting to be micromanaging, so that book in particular really sort of helped with like changing my perspective on how I can like support my team, and you know like say like one of my team members is struggling or like missing deadlines, it's instead of like why are you doing that, I can't believe you're missing deadlines, it's more about like you know I know that you're capable of this, so like what's going on and kind of having that different change in perspective, so that was one resource that really helped, and then as far as like building data science teams, I took this like Coursera course that I can't remember what it was called, but it was talking about building a data science team, and they really advocated for like your first data science team should be like a data engineer, and that was like a really helpful lesson for me and something that I acted on, and really ended up being a good fit because I had a lot of synergy with the data engineer because you know I had like the data analysis statistics kind of background, but you know I never built a database before and kind of like that was a nice synergy, so I don't have any more specific resources than that.
That's really helpful, I was guessing at what they were and putting them into the chat, but you can tell me later if those are the right ones, but I see Ethan you had a question on management, do you want to jump in?
Hi everyone, just got back from holiday, so I've missed a few in the past couple of weeks, very nice to see you Becca, I just want to ask, because you mentioned about curiosity and we all strive to learn ourselves, but as the team managers, how do you foster, how do you encourage curiosity within the team?
Okay, so your question is how do I encourage curiosity within my team? Yeah, that's a good question. In my current role, I haven't had to really encourage it because the person that I have on my team is naturally curious and I've worked with them for a while, so it's, I would say it's less about encouraging and more just we have casual conversations about you know what are we kind of interested in. I would say maybe the thing that I do with my team that I like is that we have like our team meeting on Friday and we do like different Fs, so we do like F'd up Friday, so like here's something that I F'd up this week to kind of take away, kind of take away that fear of like making mistakes, you know like let's just laugh about it and actually share what you learned from it. Then there's fine work Friday, so it's actually you know celebrating like I really liked that documentation I wrote because it was super awesome and helpful. Then the third one is found out Friday, so like what's something that you learned this week and so having that be kind of part of our like weekly team meeting, I think kind of you know it becomes part of the dialogue, learning becomes part of the dialogue of the team especially on a regular basis.
learning becomes part of the dialogue of the team especially on a regular basis.
In my role at Nordstrom, I worked with, I had a little bit more junior members on my team and there was this one team member that I had in particular who she was super smart and like very like a very good analyst but her coding wasn't the strongest and so I kind of just like told her that I believed in her and that like I knew that she could learn it and kind of just that like belief and encouragement like kind of helped her not be afraid to try it. I feel like there's a lot of times like fear of failure of trying something new, so I'm not sure if I'm answering your question in a very direct way but I would say it's more like building, creating space for people to not know everything and kind of having the team be about you know I learned something new or I learned that something that I thought I knew was incorrect. So yeah that's a little bit about how I do that.
Yeah thank you, that's very interesting to hear about the field of sharing meetings because I've heard about you know meetings where people share about their achievement but not necessarily openly talk about their effort. So very interesting that you know you can encourage people to openly talk about that then everyone can learn from the mistakes. So usually someone maybe we push the wrong code to production then falls out or whatever. It's between you and the manager to solve it and for example for a few people but not necessarily something you talk openly about in a meeting.
Yeah I'm kind of sorry please go ahead I didn't mean to interrupt. No no go ahead. I was just going to say to kind of expand on that like I try to take that culture also to my to my business stakeholders as well. I've had a lot of experiences in the past where there's some campaign or something that like my business stakeholders did that ended up not being a success and then there's a lot of kind of like weird feelings and your data's wrong and you know kind of like push it under the rug and so I try to also encourage that like concept of you know it's like not scaling something that didn't work is like a good thing right like even though like you tried something and this particular experiment wasn't successful like you avoided scaling it to our entire customer base so like you know let's move on to the next experiment kind of encouraging more of that you know failing faster so that we can kind of iterate through things and figure out what does work because most of the time the things that we're going to do like probably aren't going to be successful.
Career paths and the Friday meeting format
That's a great point. Sloan I see you had a question in the chat too do you want to jump in? Sure I guess I have two questions so the first one is did you always know that you wanted to be a manager and then also your Friday's idea is absolutely fantastic and I'd be curious to know more on how you landed on that idea were you inspired by like either you know poor meetings or was there someone that really inspired you that you know you got that meeting idea from or just how how'd you come up with it?
Sorry what was your first question? Yeah they're kind of unrelated the first one was just did you always know that you wanted to be a manager? Oh okay got it yeah I would say no I didn't always know that I wanted to be a manager I would say it kind of like came out like naturally I would find that I'm always that person in class that like is asking questions and like makes people stay longer because they're like asking so many questions and people probably get really annoyed at me for that but it was kind of like I really liked the technical work I was doing but the part that I really liked is like helping others and sharing my knowledge and like building relationships with with stakeholders and so kind of those things kind of put me in a position where you know a reorganization was happening and there was this new manager role being created and I kind of I like put my hat in the ring and I said I you know I want to do this and I think I'm qualified for xyz reasons and kind of Sheryl Sandberg my way into that job so you know just to try it and I believe that career paths are like not necessarily linear so even though you know right now like you know I'm the head of data science like I can see myself in my career being a like principal data scientist at some point so I'm kind of not like making like a firm commitment to a certain path but becoming a manager actually gave me really great perspective on my work as an IC because you know is that like imposter syndrome like does my manager think I'm doing enough like how how am I working how does my work compare to other people's and being a manager really helped me sort of gain that perspective I was okay yeah like I'm pretty good.
And then as far as like the found out uh I can't remember I mean I think it was more just like I'm a big fan of like puns and alliteration and um tasteful use of the f word so I would say that's how it kind of came together and it's just working in a small environment and knowing the people on my team really well I kind of was like okay I can I can get away with this but I can't remember like I can't remember where the idea came from initially so so it's not a very good answer to your question.
Thank you so much Rebecca that's really helpful. Elizabeth over here but I think it's Libby I guess I can call you both Libby I see you had a follow-up to Sloan's question too. Yeah I had a follow-up and kind of like a observation from from my experiences is that there tends to be like a lack of upward mobility for data people that doesn't involve people leadership and I'm wondering if maybe to open this up to everybody like if they work in an organization where there there are options for that to become more of a technical leader or an R&D leader versus a people leader I know that that kind of environment exists like at Metta you can make the same amount of money for instance at the same level as an IEC or a people leader but it's been just something I've been wondering about.
I worked with a person at Nordstrom her name is Gina Schmalze and she's just like one of my favorite humans in the entire world and I think she's a great example of like a very like she's she's a leader but she's not a people manager like just how she presents herself and like kind of helps other people learn and so I think that she she does an amazing job of that at Nordstrom just to kind of share a specific experience but I think that's a really great question yeah and I think for companies who don't have that it's sometimes hard to make the case for that as well and like carving out that leadership role that isn't a people manager.
David do you want to share any thoughts on what that looks like at NASA? Sure um it well we we don't have a whole lot of it but we do have most of the people who are at the higher level management are going to be supervisors also but there are a small group probably about 10 percent and I'm just guessing at the at that number there where myself included I am a a manager but I don't do the personnel things of what's required of a supervisor I manage but I don't necessarily supervise the personnel issues so I don't do time cards I don't do performance evaluations I don't do all of those paperwork things that need to be done for a manager but I participate in providing feedback to that I focus on managing the technical work managing the folks that and how we do the technical work and doing that and I may assign tasks and things like that but it's all around how we handle the technical products that we're trying to deliver.
I see thank you that's really helpful
I know an hour always goes by way too quickly but just want to check to see where there are other questions that I missed and if so if you want to just like raise your hand or unmute
I see Jordan you unmuted were you going to ask something
I have a quick question so since you're the first data scientist right now and you said you have to ask a lot of questions from the stakeholders when they ask you how do you help them understand the difference between interesting questions just like keep on asking because you're the only data person they have to go to versus actually impactful while still maintaining that relationship with them
yeah I would say that that's a challenge that I continue to build my skill set on I would say that I'm not always successful I think I sometimes have to weigh the political capital that comes with doing something that's interesting even though I don't personally think it's the most impactful so kind of doing a couple of those things to build the relationship
I don't know maybe that's not the best way to go about it but kind of once I got to that point it's more about like having a conversation about an ask instead of it being transactional it's more like okay I understand that you wanted to look at this but actually you know like what's the why behind this or just being very direct and saying that you know like what's the business value or like what action are you going to take as a result of this and you know a lot of times I can help them be like oh maybe that's actually like or it's you know kind of a nice to have and like not and that kind of comes a lot with like something that's urgent it's kind of like understanding like is this really urgent um so that's you know a challenge that that I continue to face um
like okay I understand that you wanted to look at this but actually you know like what's the why behind this or just being very direct and saying that you know like what's the business value or like what action are you going to take as a result of this
yeah it's a really good question thank you yeah
thank you so much Rebecca this is great and thank you all for all the great questions too
I just wanted to share one more time before we leave here that we do also have the LinkedIn group if people do want to continue any discussions there mostly it's just me posting things and just like recaps from sessions but would love to have you all feel comfortable to post there too so feel free to to use that um but Rebecca I really appreciate you sharing your experience and insights with us this was awesome yeah thank you so much for having me and I've got to drop so thank you so much for the great conversation