Data Science Hangout featuring all of us! | Building up a data science team in a small org

Welcome to the Data Science Hangout. Hope everybody's having a great week.

If we haven't met yet, I'm Rachel Dempsey. I lead our Pro Community at Posit. So this is our open space to chat about data science leadership, questions everyone's facing, and getting to hear what's going on in the world of data across different industries.

And if this is your first time here, or if you're watching this recording at some point in the future, we're here every Thursday at the same time, same place. I should start saying now we are going to take a little break for July for a little summer vacation time, and there's a lot of people traveling. But normally we are every Thursday, same time, same place.

So together, we're all dedicated to making this a welcoming environment for everybody. We love hearing from you all, no matter your level of experience or industry or area of work. And so it's totally okay to just listen in here too, but you can also jump in the conversation and ask questions or provide your own perspective on certain topics in a variety of ways. So you could jump in by raising your hand on Zoom. You can put questions in the Zoom chat, and feel free to put a little star next to it, of course, if you want me to read it instead, if you're in a coffee shop or something.

And then we also have a Slido link where you can ask questions anonymously too.

And so this will be very important today because we are doing things a little bit different here. So if you were on a few, maybe this is a few months ago now, where we had a random opening because one of our featured leaders was sick, we realized that we really liked this occasional format of us all being able to just jump in and be the featured leader. So we are all the featured leaders today.

And so I wanted to also point out that because of that, if you wouldn't mind, when you do jump in or ask a question, to just introduce yourself again too. I know some of us know each other pretty well now from coming to the Hangouts, so it might be nice to just do some introductions again too when you jump in and ask a question.

Building a data science team from scratch

But I can kick things off here with something I'd love to ask you all about. And this is something that's come up in a few conversations with people who have reached out on LinkedIn. A lot of times we'll feature different leaders here that are from larger companies who have like established data science teams. And at the same time, there's a lot of people who might be in the audience who are building up a data team for the very first time. And I was just curious to kind of open the discussion up with that to learn from you all who are in that position of building up a data science team, what does that process look like for you and how might that be different than working in one of these like large Fortune 500 organizations?

Anybody like in the process of just like starting to build up a data science team or you're thinking about building one?

Hey Rachel, this is George. Hey, yeah, you know this is actually like the perfect question for me because I started a job about a month ago and first data scientist in HR. It's a brand new role and so kind of tasked with like starting from scratch with everything. So building out, you know, metrics, defining like defining metrics and doing data governance, learning a new system in HR, you know, there's the HRIS like the main source and then from there kind of like developing a team. So it's like starting from like a even before I think what you're talking about, which is like starting with a team, it's like even before that, but having the mindset of like this will turn into that, you know. So I think right now it's like just looking at the different tools that are out there, understanding how we might manage a data lake and trying to find the right. I think right now we're kind of just looking at tools and kind of focused on that.

Yeah, how do you approach that like knowing what to focus on first? So far my approach has been to have discussions with our HR managers. So those are the primary stakeholders are the ones that are going to be leveraging the data. And communicating it to either like higher up or to the different managers within the organization. And also thinking about security and because with HR data it's highly, you know, secure. The information that we share, we don't always want to share demographics. So also thinking about security policies. But just a lot of, yeah, just a lot of conversations with people about what their business needs are.

Definitely. I see Eugene, you had a few thoughts on this topic. Do you want to jump in here?

Sure, I'll say hi. I had to build a team, yeah, at a previous position. I'm currently a team of one and hopefully we'll be building a team soon. I think there are very few unicorns, which I put three things up there, like there's a use case and a necessity. Sometimes you need somebody to be able to build things as fast as possible. So you got to go for a programmer or people who really know programming, who can build a pipeline, who feel they understand production level stuff, have dabbled in multiple tools, et cetera, et cetera. I can't necessarily rely on somebody like that to be the math genius, someone who understands integration and probability distributions and difference between a Poisson and an exponential and la, la, la, la. That person also has a lot of value and everyone can learn from them too. And sometimes if you're in a place where you need very sophisticated modeling to sort of take a leap up, you need a person like that probably who can bring in that extra skillset.

And then there's the contextual experts and that I'm currently in a position where I'm not a contextual expert. I am in bioinformatics, which is my first foray into the medical field of data. The data is more disgusting in the medical field than pretty much anywhere I've ever seen, by the way. But the contextual expert probably in this role is more important than anyone because they're going to be spending a lot of time trying to figure out, is this a generic drug or is this a name of a brand or what is an encounter? Why is this all event level data and not something where it's at a patient level where I can do my analytics? How do I convert things from event level to patient level? A contextual expert would be very, very probably the most useful person I would dive in on if I had to hire someone in a data science position to underneath me in a team in bioinformatics.

So to me, the first question is really like what are going to be the highest impact projects and who of those three major skillsets would I dive in on to try and at least create my first employee and then round out based on their weaknesses. But I would aim for trying to cover those three bases. If you're really going to build a whole team of four, five, six people, you really need people who are, in my opinion, really good experts at one of those three things because it's very rare to find somebody who's an expert in all three and you'd be paying out a ton of money if you do find someone who's an expert in all three.

If you're really going to build a whole team of four, five, six people, you really need people who are, in my opinion, really good experts at one of those three things because it's very rare to find somebody who's an expert in all three and you'd be paying out a ton of money if you do find someone who's an expert in all three.

I think that kind of answers your question you had asked, Libby, too, but do you want to ask that to a few others here in the group? Sure, yeah, that's the sort of experience that I have with talking with people and interviewing people, interviewing with people, is that when they're building a team, they're looking for somebody to hit the ground running and they always use that term, we need somebody to hit the ground running. A lot of times what I feel like they're saying is we need somebody to do everything and we also need somebody to tell us what to do because we don't know what to do and that feels really scary because I often think, well, I'm not that person. I'm a stats person, an applied stats person, that's my side of the game. I can't be everything for everybody, so as a person who is looking at the industry, that freaks me out. It sounds like you guys are also saying that person doesn't really exist and my sort of take on that is if you have somebody who does everything and knows everything, they are now your single point of failure and if they leave, something happens to them, all of your expertise goes in one person and instead of needing to replace the one person who was maybe a SME or maybe the programming person, you now have nothing. So I'm curious about people's thoughts on building a team of people who are kind of lower level and not those unicorns and how do you make that work? What do you look for?

I feel like a lot of times looking for people who can hit the ground running is like corporate code for we are not going to onboard you or train you at all. That's really something that I think when you are building a data science team, you know, like they were talking about before, you need to say like these are the people that I need and the skill sets that they're going to have and then like build your onboarding around what that team is going to look like. One of the things that we've been doing is trying to get away from oh well this one person knows everything and so like for six months all we did was write documentation about our processes and transcribe email chains from three years ago into things like that so that when we do bring on new people, we can onboard them not only on the technical aspect of the job but also on the context of the data, especially because we're in the medical field as well where you are bringing on people who may not have had any medical experience before.

Definitely. Colter, I see that you had shared when I first said about like starting your starting the first data science team that you started a data team and I'm curious like did you have to build up the business case for that or were you hired on to start the the data science team?

Yeah, hey everyone. Kind of almost both. So I work for a local government public health department and there was it was essentially uncharted territory when it comes to data science. Everything is it was done in Excel and I think I kind of tried to get buy-in on this from a selfish point of view because I was doing all this data and I was like this it shouldn't take this long. We were spending a ton of time trying to do all this so luckily I had great leadership that said you know what this is that's right we should start something and started was able to hire someone and immediately just trying to convert everything to R, automate as many things as possible. And I love I guess I'll preface this next part I love hand washing dishes because the catharsis from it and the the first time that we were able to show someone hey you know this thing that used to take you 16 hours it takes 30 seconds now and the joy in their face I will never forget that and I think that really showed leadership this is something worth investing in even further.

So that was that was about two and a half years ago and you know like I said we're able to hire one person to join and now we're a team of eight and it's been super awesome. Everything everybody's already said has been bits and pieces in our experience as well either from security or just scaling just things in general so it's been fun still got a long way to go a lot of cool stuff to learn.

That's awesome you go from starting the team to eight now how did you figure out like what those roles should be like those eight people? Oh great question I originally kind of designed it and someone had just mentioned this as you know hey this person maybe focuses on specifically visualizations maybe pipelines that get to a visualization and then realized for continuity of operation purposes that wasn't going to be feasible so switch to a different method of almost everybody's kind of trained on everything and can kind of do stuff if needed but no one is really expected to be an expert at every other person's job because that's that's asking a lot from people and I don't think as a leader I don't think anyone should expect every person on the team to be that expert at everything.

Absolutely. Okay let me let me double check and Curtis and Hannah are helping me organize questions but anybody else want to comment on on this topic too? Oh Alan I see your hand.

I think so sort of reflecting on the whole sort of circuit from from what Libby was saying to take Catherine to Coulter. I was thinking about I think I've had a really experience interesting experience with my team of saying for a number of years like we need cross-support ability for the things that we're doing and finding that as the team grows just the demand grows. We have more stakeholder teams that we're connected with and people end up being kind of de facto kind of drawn with dotted lines to those stakeholder teams and getting kind of specialized in those lanes versus being able to sort of share and cross-support and that's been a real challenge.

Because for years I've wanted to help you know certain folks who support really high intensity stuff to have more help and I think probably I've had a combination of like really good luck getting good people on the team who come in and are eager to to help to do that but I think we hit a little bit of a critical mass with the team where there were enough people and the team is now six and I have enough people who are curious and who are skilled who can help without getting sort of completely pulled into the pool of being able to to cross-support. And so whether it's like disciplinary or subject matter expertise I really have felt like in the last year there's been like a tipping point of my team is big enough now that they can help one another whereas when it was two they were just like head down focused on getting stuff done and then when it was four there was a little more room but still pretty head down and now with six there's enough of a of a cohesive team there that they can share they can collaborate without me driving all of the collaboration. It feels like that size has really opened up a bunch of new opportunity that I don't think I could have created with two or three or four people.

And it's got a whole new set of you know like working and meeting challenges but their ability to run has really taken off in such a cool way in the last six months as they've they've kind of got up to speed with one another so I don't know if that leads to a you know like a concluding point but it's been really neat to see that transition over time as the team has grown and they've gotten to know one another develop their skills and develop their subject matter expertise it's been really really fun and rewarding to see them get good at what they do and get really good at helping one another it's really cool.

That's great thank you Alan. Catherine I see your hand is raised you want to jump in? Yeah I love that from from Alan because that sort of collaborative work really facilitates the rest of the data science. Um it's something that I've I've struggled with is you know making sure that people get credit for doing that where it it frequently becomes you know unseen work you're not emailing a spreadsheet to someone you're not you know doing that but fostering that relationship sending you know being active in your your chat be like sending out those emails to those teams building a business use case around something um is is so important to staying connected to the rest of those teams. Um and so we have like a strategy for getting credit for that but like if anyone else has anything where how do you one manage that unseen work but also make sure that the people who are doing it still get credit for it I'd love to hear about that.

Can I ask you Catherine what's your team strategy for that? Um we do something every quarter called um take a bow where we basically have our members write up and they do it in an hour markdown um but they write out everything that they think that they did throughout the quarter um and then they send that to me and my sister team manager and we like aggregate that information and we escalate it up our reporting pipeline so it's like it's got the hard details of the stuff that we did um you know here's the GitLab repo here's you know the business value that this created etc but also I did all these other things like I was active on my committee board I like you know anything else that you're a part of that you want to make sure that you get credit for you can escalate that so I did kind of like outsource the like management of that credit but it still makes sure that that's it's at least seen by like one and two levels above the person who's doing that work but it's a lot of manual stuff.

Yeah somebody had it recommended to me in the and then our ladies slack group to like every week send yourself an email of all the things that you did that week so then at the end of the year or whenever you're showing your leadership team what you did you remember all those little things and I think a lot of what I do is sometimes that glue work that I sometimes forget like oh that was really important that I did that and I connected these teams across the company but if I didn't write it down for myself I wouldn't maybe have remembered it or thought it was as important.

Managing backlogs and descriptive analytics

I see Jeffrey had asked a great question in the chat I know you don't have a mic at the moment so I'll read it but the question was at what point and how do you decide to stop focusing on the non-analytics work that has filled your backlog and push for more of the advanced analytics a lot of our issue is requests come in and they are dashboards that should be done by more of a business analyst and not a data engineer or data scientist but for us it's difficult to dig out of that hole or find a stopping point.

I see a few comments here and the thread of that so I'll I'll start with you Jared if you if you feel like jumping in here. Yeah no I think it's just kind of like he was talking about further down like chicken or egg if you're getting to the point where the workload in the backlog is just becoming insurmountable how do you do a more efficient job of of organizing and kind of disseminating that to push the needle forward instead of just kind of being drowned under the work so do you take the approach of trying to split and split the work and split the the idea of having a focus on analytics versus non-analytics teams or are you trying to hire more of those you know unicorns that can do both sides of that to uh to help and you know how do you kind of manage or approach that you know that's I think that's the big question for me anyway.

I thought something that was interesting on the like topic of of dashboards is I think it was um I'm blanking on who had said this but if you stopped creating any of those dashboards like which ones would people really notice that were missing and like which ones are we just creating for the sake of of creating them and updating them. I was I was remembering that too from yeah a couple weeks ago and even going back to Libby's comment in the chat of like what's going to happen if you don't do that um and if the answer is you know four billion dollars of critical business processes stop then they probably need to hire someone to make that dashboard and if the answer is like one guy gets a little upset then like maybe you don't make it exactly.

I was saying that I get frustrated because uh descriptive analytics has a bad name now nobody wants to come into our field to do descriptive analytics or reporting but it's actually still very important um and you just can encourage yourself if if the descriptive stuff you're doing is very important like it's okay and it's not always easy to put together either. Uh but if you're trying to make that switch you may have to first convince people of the value not to say oh we want to do advanced analytics they don't understand what that means but like what's the outcome that you want to achieve that you need to go do advanced analytics to get to. And some of what I've done internally is you're talking to to stakeholders in the company what are their goals what do they want to know what are questions that they haven't asked what are questions they um don't ask because they don't want to be disappointed that there's not an answer. Uh so I will ask people to like think without um think without any restrictions like don't worry about what we can or can't do what would you like to see and then if you can get people who would who would sponsor yeah these are the these are the outcomes we would like to get then you have a little more weight behind that request to try to shift your time away from descriptive analytics to something else.

Thank you um Gerardo I see you have your hand raised here to jump in on this topic. Yeah I think what um what you just said is is is critical because getting access to the data is usually one of the harder things to do so if you have an executive who wants a report or a dashboard or whatever you can break that that deadlock in getting access to the data and once you have it you know you can you can do the more advanced analytics and and start to build that business case of like if we look at it this way if we do more we can give you more value. But in the end getting that data and and breaking up that process of getting the data processing it and analyzing it in three different steps it helps in in a variety of things it helps in building your team slowly because those are sometimes three different expertises and you can potentially um give the getting the data to it so that you don't have to do it. So you know it it opens up time for um for your own analysis um it also makes it easier to go back all the way to the beginning of the discussion about hitting the ground running um you have people who are good at these three different things and I think when they describe that they're looking for people who are hitting the ground running are people that can answer that final question that give that report to the executive and you know give that final result. Um so if you can use that uh from a report from a from a straightforward dashboard if you can use that to break the um the deadlocking in getting access to the data and start slowly building some analytics I think that that's a good way to start. You have to be very careful that to to also what what people were saying that it doesn't become your day job and your only day job you know.

I kind of view I was thinking there's like this classic three or four step thing it's like you have descriptive and then diagnostic and then predictive and then prescriptive that was like something maybe eight ten years ago that made the rounds. I've seen that too Chris what I think people when they say analytics they're talking about everything to the right of descriptive on that visual okay the script isn't it's like what is what is going on right now yeah how many of the plot we've this is what is happening or has happened um but diagnostic is kind of like what has happened I think it's just yeah this is what it is cool thank you.

Ralph I see your hands raised here by the way you just can't do any of those other three things good at all if you don't do descriptive analytics. Yeah that's what I found with working with clients is helpful is I actually made an image of a pyramid where the first layer is descriptive analytics and then predictive is smaller than prescriptive is on the top and so with the idea that majority of your time running your business and most of your efforts as a data team should be in that descriptive side because that's the basis of anything to predict the future or making large changes in your business so that's been helpful and then I also started adding a basement of data management. Yeah there's a data quality layer under there somewhere I call it the basement even though I don't know if pyramids have basements.

That's great George I see your hands raised. Oh yeah I just wanted to chime in on the descriptive and um yeah to kind of reinforce what was said about you know it's just as important and crucial and fundamental and made me think of actually this morning I cracked open my Edward Tufte book and you know just like in awe over the descriptive stats that are in there but the way that they're presented like they're presented in a way that solves a problem right like there's the classic like cholera outbreak map that was done and it's like that's descriptive analytics but it's just in a way that's telling a story it's in a way that's that resonates with people so it's not necessarily like oh it's just so simple um and you know maybe not not worth our time but it's more like how do you how do you convey those descriptive stats it could be as simple as a bar chart but is that really answering the question is that really getting to it could be it could be descriptive stats that's all they need but again like how it's conveyed is you know vital.

Definitely thank you I'm loving this lively discussion here Mike I see your hand raised. Hi yeah I was just going to say that I heard a great talk from a data science leader I think working for one of the big supermarkets in the UK who described descriptive analytics as kind of looking in the rear view mirror so you know what has happened so maybe that's a that's also good analogy.

Dashboard governance and decommissioning

I see on the topic of like getting rid of some dashboards mark you had a great question earlier do you want to jump in here. Well the question was does anyone have experience or suggestions for getting folks to agree to deleting dashboards mark said our group has so many dashboards we maintain with only a few people so I'm wondering if you could to deleting dashboards mark said our group has so many dashboards we maintain with only slightly different views of the same data set and I have thus far been unable to get buy-in even though the technical debt we've incurred is astounding.

I have a few comments on this um we just so I'm at Centene along with Catherine who's mentioned you know who's talked a bit this morning but we have like 75,000 employees or so and I actually just helped run a hackathon for anyone that wanted to participate at the company and the whole purpose the whole problem statement was effectively what can we do about our dashboard craze situation and believe it or not like outside of this hackathon since January 1st there have been about 25,000 dashboards that have been deleted so the idea is like what is going on with so many dashboards and reports that are now all on the cloud.

And so yeah it's a serious problem and I don't think that for larger corporations at least I don't I kind of feel like the thought has been just create a dashboard if it's not used like don't even really think about it you know if it if like yeah if we need to delete this later we can no big deal but there's costs associated with all this and the resources that are used to just keep those online um so yeah this this this whole hackathon was done to really try and drill in on like can we put some strategy behind a rationalization approach or some kind of decommissioning approach or strategy to all these different um you know reports and dashboards that the company has and it's not just I'm not just talking like 25,000 Shiny apps would be incredible I'm talking more like um you know reports and dashboards across all of our platforms Power BI Tableau Micro Strategy etc. But hackathon sounds really cool too yeah the winners did an excellent job coming out with kind of a solution that had more of that like prescriptive um insights for leadership um so I'm really excited to see what we do about maybe productionizing that as kind of a Shiny app or something like this for leadership to to have a clear visibility into like how many how many reports you know should be on a decommissioning path or etc.

Anas I see your hands raised you want to jump? Sure so I guess one gentle way to um to test the waters for governance for some of these dashboards is to um you know if it gets updated on a constant basis try one period to not update it or one month not to update it and see what reactions you get and if it's a live feed try to cut it off and see what reactions you get now if no one says anything and they're not really paying attention to it then that's a good sign and you can say okay this is probably not being used um as much as it should as it should be or as much as expected but obviously if you hear people uh express their dissent and say no we we rely on this data then that's a good indicator to say that the dashboard is still relevant and uh you know the consumer it's stakeholders still need it to be in place and active.

Thank you I see um Abigail you have your hand raised here too you want to jump in? Yeah so we monitor our usage of dashboards I mean like that's really the start like we Tableau has APIs um I don't know what other tools have but like you can see who's looking when and so being able to go in with like nobody's looked at this in three months like that's a really strong argument like literally nobody's using this um and even then nobody wanted anything deleted but we were able to get permission to hide basically hide dashboards nobody was using except for people who are actually building the dashboards write some documentation that says what all the ones that are in hibernation do so like in the future if somebody wants to bring it back they can read that and say oh I needed this but it's not something that we're having to worry about maintaining on a regular basis. Do you think sometimes it's that like the dashboard or whatever visualization was really helpful for someone when they first saw it but then they don't know where to ever find it again? No because nobody's looked at in three months I think people just really get freaked out by the idea that like somebody spent time working on building this thing and you're gonna get rid of it and so like we're not we promise we're not getting rid of it it will still exist if you ever need it I think like the safe thing even though really some of these are never are never coming back.

George I see you have your hand raised here too you want to jump in? Yeah um we actually did that uh with the dashboard because we actually it was a Power BI dashboard and tells you when it gives us reports that no one's using it and so we decided to stop updating it and found that okay they actually did need it but not as frequently and so I think like just going back to the requirements for building it and understanding you know the frequency of of that and so just having a discussion about like what were the expectations so realizing oh they really only need it like once a quarter but you know we built it in a way that could be done like every day like you know but we didn't need to update the data every day um so you know sometimes um it's just understanding the uh the requirements and not necessarily going to think well they don't they don't ever use it well when they use it it's super important it's just not frequent.

Keeping track of code and upskilling

There was a Slido question someone had asked anonymously I try to save code tips that I can see uses for in later projects but I feel like it's so messy how do you all keep track of frequently used code notes and tips.

And Arcos I see your hand raised here so I didn't think you'd see it in time before someone else started talking um I so I have to I work in Python almost exclusively like at work and I think I've I think I mentioned that before um and you know we do a lot of Jupyter notebooks for interactive things which I know people have a lot of feelings about but I am determined to put as much code as I can into GitHub so we have a we've got repositories for our main like production things that are controlled we have our team has a big team a big pricing team has a a repository that's basically scripts and some of the scripts can be things that we run automatically but some of them are just like it's just a repository that anyone can push to and so if I run anything that I think is cool and I want to use later and I think anyone else will have any benefit of I will put it into like a utility scripts folder in there and it's version controlled and people I'll tell people about it and then other people sometimes will contribute to it. And then I also have my own little secret private Marcus repo where if something isn't quite baked enough for to share with everyone I will save into my own repository and so version control um and Git and GitHub if you have access to those tools are a really good way to not lose uh stuff into some one-off file or some one-off Jupyter notebook whatever. And so my strategy is to write as much as I can inside an editor and save it into a Git control code file and then if I need to use it in a notebook I will import it into my notebook which you can do with Python pretty easily um that I'm not exactly sure how the R folks um uh think about this um since I have when I used R I was terrible at writing functions um but um uh that's kind of my strategy um and so if you haven't learned about Git and version control highly recommend um either chatting with other folks here or um or just uh yeah trying to learn more about it.

Definitely I uh one resources I know it's like the happy Git with R um book if anybody wanted to share that in the chat um but Eugene I see you have your your hand raised. Yeah functions functions functions functions just functionalize it give it a descriptive name put it in your in somewhere that I created a company uh R repo you know or R package um just titled it the company name everybody had a copy of it everybody could make edits everybody could update those edits and every single function was it just became more and more generalized which is the best thing about writing functions and forcing that everybody writes functions and you can't finish writing your function unless you actually create a manual page and which is the best part about it so I know it takes you know an extra 10 minutes and uh it's such a hassle you know to add that three sentences of documentation that describe the arguments and what you want to use it for but I mean the overall build of productivity over time is just you know nobody has to resolve the same problems we all have to convert dates we all have to convert zip codes we all have to add leading zeros we all convert things from factors to strings or strings to factors etc count the number of factors like just build functions that do this stuff just build them and leave them there and anytime anybody has a problem like this you know in your team meeting you say hey what new functions were added this past week it's very simple to get that sort of it's exponential productivity that you'll get across the the team.

Catherine I see your hand raised here. Yeah I was just gonna comment that like even if you're not on like a full data science team in an organization that has GitLab and our professional tools like before I was at Centene I was the only R user and a data science team of one and I basically made a setup script for myself that had some of my most helpful functions in it and in my little profile I said put all these in my session at startup and so it was it was a function for dealing with dates in the format that I knew that they came from my database and things like that like even if you're just doing it for you it's still a good practice and then if you do onboard more people and you start share you can share that script with them and then eventually make that business case like hey we do need a GitLab license so that we can share these things as our team grows and as our organization grows like and if it does still just be you you've made your own life better.

Of course yeah thank you all for the the thoughts here too before leaving this topic I was kind of reminding myself I think Mike Smith you talked about this by using like snippets too. Yeah you know snippets are fabulous right because you you kind of give a little shortcut and then it gives you a skeleton or framework where you can jump into certain parts within that framework and kind of almost auto-complete certain sections you can you can share snippets there's use this allows you you know has a way of opening up the snippets so that you can edit them but you know so for things that you do frequently and where you actually have that framework they're really really helpful but I think I'm leaning more towards what's just been said about right right functions right because it's like you can also then document those test them you can roll them up into a package for some reason writing a script is a little bit writing a function it just sounds scarier to me than the word snippet. Well it does because writing a function implies that you're going to write the documentation then you're going to maintain it forever right whereas a snippet well it's a bit more throwaway right.

Yeah I think just a point to add to this this is something that I try to do my company is also sharing that you've done this with others making sure it's on your knowledge-based solution because then you find out oh this has already been done by somebody else and I didn't have to put all this effort if they'd only share it so I think contributing to the community is important internally just as well as it is externally.

Building community and collaboration

Libby I saw earlier you had a question around collaboration do you want to ask that one and I um I was probably about time like there's a lot of conversation around like we really want people to collaborate and how do we get them to talk to each other and how do we get them to like you know get to know each other collaborate and how do we get them to talk to each other and how do we get them to form a community and be more connected because we're all remote and my question is do they have the time to do that because if they don't you are expecting something magical like it's not going to happen. Um and so the the push that I'm that I'm trying to you know go for is like prioritize timing prioritize chunks of people's time for that stuff if you don't tell them to do it and that you want them to do it give them the time to do it it's probably not going to happen because they're going to be heads down um and feeling too pressured. And I was thinking about there is a Harvard Business Review article that I read recently about capacity about how people do their best work when they're actually about 85 capacity or 80 capacity not 100 and aiming for people to be at 80 or 85 percent does free them up a little bit to have the mental and emotional capacity to connect with other human beings to share stuff and we have all this double work and we have all these silos guess what if people talk to each other those walls start to break down and you start to realize other people are duplicating work and you start to realize what's possible because someone says oh yeah I did that in this other project and maybe it's not the exact same thing but it's going to help so that's my that's my two cents I guess less of a question but are you prioritizing your people having that time.

There is a Harvard Business Review article that I read recently about capacity about how people do their best work when they're actually about 85 capacity or 80 capacity not 100 and aiming for people to be at 80 or 85 percent does free them up a little bit to have the mental and emotional capacity to connect with other human beings to share stuff.

I love that thank you Libby and I think Mike has some thoughts to share on it sorry for being the guy that jumps in here but um you don't have to be sorry the I've I've been the our guy at our company for many many many years and before we really got our our community going everyone came to me oh Michael know the answer to this oh Michael help me to solve out these installation issues oh Michael help me answer this tricky question I've got about how to do this thing so and that then funnels everyone in through me you know and and that you might argue that that's great for my brand but it's you know making me beyond 100 capacity and then it's hard to keep up with all the requests and it means that I have the same question again and again and again and again and again and it gets really tedious. Because even though I bang about it and go you must do it this way everyone's so busy and they're like this and and it never gets through once we have the community in place we have a single source of here's the information on how to install our here's the information on how to you know get the packages you need here's you know we have that place for the the written documentation we have our community sessions that say if anyone's got questions they come in through that community page and anyone can answer that question right so it's absolutely vital that the community gets going because otherwise it all funnels through your regional local friendly our expert and that person just winds up burning out. So you know it's it's kind of you know we've been talking about the ROI on taking time to build community but without it you're gonna have you know someone's gonna burn out because they're at the sharp end of that funnel.

That's great yeah big uh kudos to the the Pfizer team there and all the the different community efforts you lead. Uh Gerard I see your hand went up on this topic yeah so uh when COVID hit and and everybody started working from home what we did was uh I established tea time um a couple times a week for the team to get together and there was a there was a very simple rule you were only asked to uh talk about work related things in the last five minutes um but it really helped to have that interpersonal connection with the people in the team and so you kind of knew if they were stressed if they were doing something fun at home or whatever and then in those last couple of minutes you could quickly ask you know what they were working on that did anyone need any help um you know do we need to make any connections uh between different teams. Um and so to uh to to Libby's point yeah the more you um if you don't force it you know those conversations don't happen and and like because the first couple of weeks of COVID no one talked to each other like everybody was just doing work and and uh that just didn't work um so uh after initiating that tea time uh that at least that forced conversation um really helped in in multiple ways not just socially but also um in in keeping that collaboration and that networking going.

And thinking about how we sometimes we've talked about like the power of automating something that was once manual and how that helps you like exponentially down the road but it's almost the same with the community efforts like you need to take a chance to like step back and look at the big picture and create that community so that it can help everybody later on down the road and that you're not just the one person who everybody's going to for questions and that there's so many others that are supporting each other.

Um Eugene I see your hand is raised too. Yeah I mean this it's an interesting concept that the documentation and the building functions for a team and having your team iterate on the functions um I don't think people recognize that that is its own collaboration it begins literally the collaboration. You I remember when I was hired um at really like my first management gig and I I walked in on the R code that they were using and everybody was on a different version of R and everybody was using different versions of packages and everybody had three or four different templates to do basically a similar version of the exact same analysis uh theirs worked for them they had certain functions that had certain edits for them for one of these particular use cases I mean it was terrible in terms of overall productivity and building models took literally three days and it took me about six months to a year but in a year I had people building models in four hours instead of three days right. And we were all using the exact same R version and we were all using snippets of functions that slowly got better because somebody finally put it put one into a package and when somebody checked out that function and actually did something with it it broke for them because they were using a slightly alternative use case so what did we teach them to do you add a new argument and then having that new argument you add that to the documentation and then on the bottom you have these edits you make that you just add an if-else statement you know based on that new argument and now the function does both things for both people.

Um and and this literally began the collaborative process uh and there were you know people who took two years to finally check in their own brand new function even though they were making edits of functions in the meantime and it was a big you know yay Amanda finally checked in her first function you know and and the package grew and more use cases were handled and the and our entire library of code got better um and our ability to talk to each other hey didn't you build something that did that didn't you build something that you know solved a similar problem um and if it doesn't do the exact same thing you do you know you get to be the unit tester and figure out how it works and then you build experience with git now your resume grows because you have experience working with repos you have experience working with version control you have experience modifying functions automating your work all these things are like you're solving the collaboration problem you're solving a productivity problem you're solving um you know the ability to actually contribute and build these skill sets that every future job is going to need from you. I mean the whole thing is literally like you can solve in my opinion multiple problems just by trying to centralize your code and get everybody to work together on similar codes similar projects.

I mean the whole thing is literally like you can solve in my opinion multiple problems just by trying to centralize your code and get everybody to work together on similar codes similar projects.

Encouraging upskilling

I love that I love how passionate you are about it but also that yes and you can do it too. Um one last question I wanted to ask I see Priyanka asked a question that is it's kind of on the same topic around making time for community but for the team builders and leaders here how do you encourage your teams to keep upskilling while continuing the business as usual. I'm really lucky I have Sam Palmer on the team and he's he contributes to our weekly and so you know it's part of I see that as part of his job too you know it used to be a hobby now it's actually part of his job but Sam collates that information anyway but then internally they produce a little kind of monthly blog post to say you know what are we seeing is the latest greatest and that helps enormously.

That's great any any other thoughts on encouraging your teams to keep upskilling? I'm gonna say the same thing that I did for community stuff which is like put that time into people's reviews and into their weeks like I've had jobs where it's expected that 10 of my time is upskilling and I need to be able to talk about it I need to be able to say like this month I use 10 of my time to do x y and z and if it's a part of the accountability for somebody then they know that they're allowed to do it.

Any other it's hard though right back back to also what um was said earlier that you know your your focus on your day job to get stuff done and off your desk and out the door means that it's hard to assign time to learn. I was just thinking there that we have people who learned base R you know like 15 years ago and they haven't even got to the tidyverse yet and then we've got other people who learned the tidyverse five years ago and aren't aware of the the kind of you know most recent changes to functions and so it's it you know you've got to keep that ball rolling and continuous improvement um and just kind of keep chucking it under people's noses.

I like the ideas of like actually putting it on your calendar I see Tony you mentioned you're encouraged to use two hours per week upskilling.

I've been thinking about this with um I talked with our president yesterday about sometimes I have a tendency to want to focus on a lot of things and a lot of different ideas and making things better but it's also important to like pick what is the most important thing that I want to focus on and trying to like block a whole day for that um just as dedicated focus time. Tony I think do you want to add something there? Uh yes it's um it doesn't always happen but if um if you can have that goal and maybe put two hours on your calendar especially if if you have a remote work day and can avoid having meetings during that time a lot of times it seems hard to do because of the immediate needs the fires that you have to put out um but getting it could pay off in the long run because something could take a lot less time um when you because you've already got that baseline of knowledge and on whatever skill it is you need so um it's it's definitely worth it if you can if you can find a way to fit it into your schedule.

Really uh and George I see your hands raised too. Yeah I wanted to share my latest uh tip or trick for um upskilling I I'm going back to the old days of using RSS feeds in my Outlook and uh something that you know he used to see back in like a long time ago 90s and early 2000s um so now I get our bloggers and a lot of other resources sent directly to my Outlook um application and you know every time you see that little number saying how many unread messages you have I know like okay there's another article I'm going to keep up with and that's that's been really really helpful for me lately.

That's a great tip well I know we are a little bit over on our our time here but thank you all so much for jumping in the conversation today and being the featured leaders here I love to see all of this the chat happening as well and I'm excited to go read that later um but thank you all again so much for for joining us here today. Um and I wanted to let everybody know that next week um Joe Chang our CTO at Posit will be our featured leader Joe's the creator of Shiny and that will be our uh the last hangout before our summer break so we'll be off from hangouts for July but we'll be back the first week of August um so yeah next next week with Joe Chang hope to see you there too bye everybody.