Unlocking data team success: hiring, resources, and AI

Summary

Highlights

• Barry, CEO of Hex, highlights the selfish initiation of Hex to solve data problems they previously encountered. 🎤
• Toin Goos from Coinbase shares insights on balancing foundational metrics with business growth signals. 📈
• The panel discusses the role of stakeholders in advocating for data team resources. 📢
• AI's role in data work is explored, with emphasis on enhancing rather than replacing human efforts. 🤖
• The importance of clean data systems is discussed as foundational to harnessing AI's full potential. 🛠️

Key Takeaways

• Hiring the right people is crucial for impactful data teams. Look for curiosity and problem-solving skills! 🔍
• AI is not a replacement but an enhancer for data work. It helps tackle more complex issues by handling repetitive tasks. 🤖
• Stakeholder relationships are key. Having allies advocate for the data team boosts its perceived value. 🤝
• Clean and organized data infrastructure is essential for leveraging AI and achieving reliable results. 📊
• The data team's ability to frame business problems is evolving in importance, alongside technical execution. 🧠

Overview

Chapters

• 00:00 - 00:30: Introduction and Event Overview The chapter titled 'Introduction and Event Overview' begins with a warm welcome to the audience. The speaker expresses gratitude for the attendees and indicates that the size of the audience is perfect for the location. The speaker appreciates the familiar and new faces in attendance and introduces the event as part of a series. The speaker humorously mentions being grateful for the guests, setting a lighthearted tone before introducing the special guests for the event.
• 00:30 - 01:00: Barry's Introduction and Company Background The chapter titled 'Barry's Introduction and Company Background' opens with Barry, who introduces himself as the CEO and co-founder of Hex. He humorously describes Hex as a 'selfish company,' which he explains is common among capitalist endeavors. Barry shares that the motivation for founding Hex was driven by personal experiences and challenges faced by him and his team as builders and users of data tools. Previously, at their last company, they encountered specific problems related to data, which Hex aims to address.
• 01:00 - 02:00: Value of Data Teams The chapter titled 'Value of Data Teams' explores the challenges faced by data teams in providing tangible impact within a company despite having intelligent team members and modern tools. It discusses how the company Hex emerged as a solution to increase the effectiveness and influence of data teams, partly addressing these challenges. However, it acknowledges that achieving the perfect solution ('Nirvana') is not about acquiring more enterprise tools but involves a deeper understanding and execution of impactful strategies.
• 02:00 - 03:00: Introduction of Panelists The chapter introduces the panelists who are leading data teams. The speaker expresses excitement about working with various data teams and leaders, and is accompanied by three of them. The format includes a Q&A session followed by open-floor discussions. The panelists are prompted to introduce themselves.
• 03:00 - 05:00: Value and Impact of Data Teams In this chapter, the focus is on the introduction and setup for a discussion involving a data science team at Coinbase. The speaker, Toin Goos, who leads the platform data science team at Coinbase, expresses excitement about being part of the conversation hosted by Barry and the Hex team. Although the transcript only captures the beginning, it sets the stage for a deeper dive into how data teams operate and their significance within organizations. The initial exchange indicates a collaborative and positive atmosphere for the forthcoming discussions.
• 05:00 - 06:00: Challenges in Measuring Data Team Impact The chapter titled "Challenges in Measuring Data Team Impact" discusses the organization of a company into four different product groups, with a particular focus on the platform group, which is the second largest. The platform group is responsible for providing common tooling and capabilities that support all products and services of the company, specifically mentioning Coinbase. The speaker, who is part of this team, humorously notes their role in introducing a new tool, Hex, to Coinbase, hinting at the challenges and dubious roles often associated with measuring the impact and contribution of data teams within large organizations.
• 06:00 - 08:00: Stakeholder Dynamics and Data Roles The chapter 'Stakeholder Dynamics and Data Roles' delves into the structure and roles of a data team within an organization. It starts with a leader discussing their oversight responsibilities over data engineering, data science, and data analytics, emphasizing their focus on internal use cases such as developing churn models and company reporting. The aim is to ensure reliability and effectiveness in these engineering tasks. The excerpt concludes with a brief introduction from another person named Julia King.
• 08:00 - 10:00: Challenges in Justifying Data Team Investments The chapter 'Challenges in Justifying Data Team Investments' discusses the centralized data team at Gusto, which works across various functions within the organization. Key challenges include prioritizing data quality and availability, and addressing a range of similar problems encountered. The speaker indicates that they've been with Gusto for six months, implying an ongoing process of adjustment and evaluation in addressing these data-related issues.
• 10:00 - 12:00: Stakeholder Relationships and Justifying Headcount The chapter discusses the challenging question of understanding the value and impact of data teams within a company. It reflects on the importance of measuring ROI and how different organizations approach this issue.
• 12:00 - 15:00: AI's Impact on Data Roles The chapter titled 'AI's Impact on Data Roles' opens with a discussion between stakeholders about the perception of the value data roles provide within a business. It highlights the importance of understanding and navigating these perceptions within different business contexts. The conversation provides insights from an individual's experience transitioning from working at Lyft to Coinbase, illustrating how the value of data roles can vary significantly depending on the industry and specific company needs. Lyft, as an example, is mentioned to illustrate how data roles played out in a transportation network company setting.
• 15:00 - 17:00: Improving Data Team Productivity with Tools The chapter discusses improving productivity for data teams using tools. It starts by exploring the role of data in optimizing supply and demand, calculating estimated times of arrival (ETAs), and routing drivers, highlighting the data-centric nature of these tasks. It emphasizes the importance of data engineers and data scientists in such companies, noting that their core product revolves around data. A comparison is drawn between two companies, Lyft and Coinbase, illustrating different business structures and the underlying significance of data in their operations.
• 17:00 - 20:00: Audience Q&A - Data Team Motivation This chapter focuses on a Q&A session discussing the factors influencing the motivation of a data team within an organization, particularly in the context of external factors like macroeconomic trends, crypto sentiment, regulations, and political changes. The conversation touches on the role of data science (DS) and its value, with a specific mention of Coinbase's journey, highlighting the external elements that drive business and the implications for data-driven optimization.
• 20:00 - 24:00: Audience Q&A - AI and Strategic Investment The chapter, titled 'Audience Q&A - AI and Strategic Investment,' discusses the rapid growth of a company following its IPO in 2021. Initially, the company focused on building dashboards to track and visualize its burgeoning growth, marked by consistently rising metrics. However, by 2022 and 2023, the company faced challenges due to a macroeconomic downturn often referred to as a 'crypto winter.' This shift in conditions prompted a strategic pivot towards optimizing revenue streams amid the changing economic landscape.
• 24:00 - 27:00: Audience Q&A - Strategic Planning with Data In the chapter titled 'Audience Q&A - Strategic Planning with Data,' the discussion revolves around the evolving role and value of data in business. The speakers highlight the changing perspectives on data as it relates to business costs and macro contexts. One speaker opts to pause their thoughts to invite input from others, implying a collaborative environment. Another speaker, likely Sumit or Julia, is prompted to continue, acknowledging the fundamental role of data teams, who often enter discussions after problems have already begun. This exchange underscores the strategic importance of data in business planning.
• 27:00 - 27:30: Conclusion and Closing Remarks In the conclusion of the discussion, the focus shifts to the self-reliance of business professionals in using data to make decisions, even when a data specialist isn't available. The importance of data professionals becomes evident, especially when they can simplify complicated tasks, such as using dashboards for decision-making or strategizing on customer upselling. Their ability to streamline these processes establishes the undeniable value they bring to the table, highlighting the irreplaceable support they can provide to non-data specialists within a business.

Unlocking data team success: hiring, resources, and AI Transcription

• 00:00 - 00:30 hello everyone um thank you for coming out tonight uh it is lovely to see everyone this is like the perfect amount of people for this space so uh I'm glad this number of people signed up and attended and not one more this is perfect um we uh have been doing a bunch of these events recently and it's always great to see people friends new and old uh and I'm very grateful to be joined by three awesome people who we will introduce in just a moment um when I I joke some times oh
• 00:30 - 01:00 and by the way I'm Barry I'm the CEO and co-founder of hex um I joke sometimes that this is a very selfish company I guess most are because it's like a you know capitalist Endeavor at some level but um uh it's a very selfish company because we we started hex just to solve the problems we had had uh and had suffered with for a while as like Builders and users of data tools and um the specific thing at the last company we were at was like the the data
• 01:00 - 01:30 team itself we had these really smart people and we ostensibly had these modern tools and we didn't feel very impactful we didn't feel like we were actually like driving things for the company and hex was kind of an answer to the riddle for why um how to solve that or a piece of it uh as we've built the product and built the company I've realized and found new ways in which that's true but I've also realized and found new ways in which Nirvana is not another Enterprise tool contract away and that it really matter matters like
• 01:30 - 02:00 who you have leading your teams how you're organized how you think about things and uh we're very fortunate that we get to work with all sorts of great uh data teams and data leaders and see the way that they do it and uh I'm very fortunate to be joined by three of them tonight today so I'm gonna ask you guys some questions uh yeah you look you look scared I am a little scared we're gonna ask some questions have a conversation and then I'll open it up to the floor um so I'm going to start by asking you all to um briefly introduce yourselves and
• 02:00 - 02:30 then we'll get into the good stuff how's it sound sounds good does it work it works you sound great sweet we'll start I'll spare you we'll start this way and come this way so you're ready to go all right this works too right yeah yeah you sound wonderful yeah uh well thanks for having me here uh Barry and the hex team uh really excited to be here um I'm toin Goos I lead up the platform data science team at uh coinbase so um coinbase is
• 02:30 - 03:00 organized into you know four different product groups so platform is probably the second largest product group and a lot of what platform does is like you know like the common tooling the common capabilities that power all of like the coinbase products and services that you're aware of so that's my team uh dubious distinction of also introducing hex to coinbase dubious come on man hey everyone my name is samit uh I work at BR worked there for around five
• 03:00 - 03:30 years uh I lead the data team including data engineering data science data analytics and we really Ser kind of like the internal use cases so we're building turn models and building out the company reporting and all the other things that come with the engineering ter in terms of making those things reliable so it's been fun can I go now yes hi everybody I must okay I'm Julia King uh I'm from G I've
• 03:30 - 04:00 only been there for six months so you know we'll see how much I can talk about Gusto but um we are also a centralized data team so I'm part of that team um and we have all of the different functions and work with every team within gustau a lot of priorities on data quality data availability so a lot of similar problems um that we're trying to solve as well thank you um so I I mentioned a moment ago like this thing we thought a
• 04:00 - 04:30 lot about when we started the company about the value of data impact of data teams and um I've come to believe this is kind of like one of the hardest questions the thing that fundamentally on data in the data World we're we're sort of forced to Grapple with and I'd be curious to hear you guys uh reflect on on this a little bit like how do you think about that within your organization like you know you you're you're running these data teams how do you think about the value you provide how do you think about measuring Roi how do you think about
• 04:30 - 05:00 the the perception stakeholders have of the value you provide how much do you think about that how do you how do you navigate that to him I or I'll ask you to go first yeah for sure um I can go first um so you know the so the thing is that like it really sort of depends on the business right what I mean by that is I was at Lyft before I came to coinbase um and Lyft was interesting because it was primarily if you think about lyt you know you're
• 05:00 - 05:30 trying to figure out how to optimize supply and demand and figure out etas and how to route drivers Etc so everything is kind of like data related like you know the core product is a data product right you know it's everything is data adjacent and nobody sort of questions like what the value of data is I mean that company won't exist if like you know they weren't a bunch of data engineers and data scientists and then I came to coinbase and it's a very different business it's uh it's probably a stronger business than Lyft but a lot
• 05:30 - 06:00 of it is driven by what happens in the macro right you know what the crypto sentiment is what the regulations are who gets elected all of that right you know so and that drives the business there isn't as much for like sort of DS or D to optimize I mean we do our stuff and that's where like this question becomes more pertinent is that what's the value of data right and so um coinbase has also had this sort of interesting jour Journey so coinbase
• 06:00 - 06:30 ipoed in 2021 and it was growing so rapidly in 2021 that like you know the point of data was he just buil good dashboard so that we know how fast we're growing see that number go up yeah exactly everything going up and to the right and that's what we want you to do um and that changed in 2022 and 2023 when we went through you know a macro downturn we called it a crypto winter Etc and that's when we were like oh man we need to optimize this you know we need to optimize our Revenue we need to
• 06:30 - 07:00 sort of think about costs Etc and that's where so like you know just like the what data does and what the value of data is uh changed quite a bit like like I said the context is the business and what like the sort of the macro context is I'll actually stop there and like you know maybe hear what Sumit has to say or Julia has yeah I think you're touching on something that's kind of fundamental to like data teams where we kind of come in after sometimes on a lot of problems
• 07:00 - 07:30 where the business itself is using data or someone in the business using data to make a decision like they don't always have a data person next to them helping them they're just doing things either by looking at an existing dashboard or trying to find a way to upsell a customer and then all of a sudden it becomes really clear on the value of a data person because we can come in and make that job so much easier right and I think having kind of that sometimes follow on like approach can make it UND disputed that we can help sometimes it's
• 07:30 - 08:00 hard when you're kind of coming in it's like I have this headcount I need to develop them a bunch of projects and find a way to measure that as high Roi it's ends up kind of with a recipe of your creating work that maybe doesn't need to exist and you're trying to upscale people that maybe don't even have a use right now for data because they're doing quite fine without it um despite everyone wanting to be data driven I think that it's something we've learned pretty clearly it's like when that need starts to become like a
• 08:00 - 08:30 burning fire like all of a sudden no one has that question um they don't care about the data value they just want it to be better and want it to be faster and all of a sudden you can get more predictability out of in all the things that come with adding like a data scientist or dat engineer to a particular problem I think to me that's like it's a lot easier sometimes to be in that support role and then develop the partnership versus trying to kind of go around finding every single possible place but data
• 08:30 - 09:00 in I agree I think one of the big challenges that we have and I've had in Gusto and private companies too is also separating the value of building the data foundation and making that solution more scalable versus saying we can build you a model the model is a lot easier to measure you know there's an Roi on you know like we were just talking about risk model the other day of how much money we were able to save because we prevented that much fraud right like it's a lot easier to put the number to that problem and what the impact is
• 09:00 - 09:30 versus what does the model build on top of which is a ton of other work that has to happen for this to even become a reality and so I think the challenge becomes um not even always the value of a whole data team it's a lot more about the value of the the like the foundational data teams that allow for this to happen so how do you how do you navigate that because you know you were mentioning like a lift or like a risk model there there's these use cases where it's like hey we are on the data team that the deliverable we have is something that's like either part of the IP of the product
• 09:30 - 10:00 like how we're doing um Supply demand optimization or it's measurable in terms of like you can you know we had this much fraud and now we have this much fraud or whatever it is right but like I us the vast majority of data stuff isn't like that either as you said because it's more foundational or it's just like we're supporting this decision with an Insight it's very hard to measure I I want to really pin you down on this like the rubber meets the road you're going and advocating for headcount or tools or whatever you know budgetary thing H how do you navigate what have you learned
• 10:00 - 10:30 about having those conversations and like how to justify you know the the wonderful and expensive people you have and the tools you have and all that stuff like laughing because one of our senior managers is right here and we just had thisday she like I'm just going to drink um it's really hard we've been trying to use a lot of examples and trying to explain how like in the real world in the real data world if we don't do these
• 10:30 - 11:00 things if we don't hire these people to build the foundation or in in you know invest in data quality or invest in um building the kind of the more scalable Solutions um that we need like how long would it take for us to build that risk model um and so like that's one way to do it do do the powers to be believe us when we do that you know dbd sometimes it works sometimes it works not as good um but I do think es especially I'm
• 11:00 - 11:30 going to say I'm going to say it AI um AI is I thought we'd make it a little further n um I think it's AI has been an interesting uh kind of big U not wrinkle to all of this because obviously there's a lot of hype and I'm sure we're going to talk about it later but um I think there's becoming more and more of a realization of how accurate data is is important for any of the tooling to work including models right so I think when we had humans building models a lot of the flexity was sort of hidden because
• 11:30 - 12:00 they figured out they would figure out how to build this solution but if you put a tool on on top of it something like AI where at at first it looks like it's doing all the great things and then it breaks because the data is not clean or we don't have the right tables or the names are confusing um and it gives you wrong answers as a as a user of that um whether you're a leader in the company or product manager whatever you start realizing that that value of that data foundation and good quality data becomes uh is important so I think in a way it
• 12:00 - 12:30 hurts us in the short term but maybe that's another way of saying but see this is why this is important doesn't matter if a user builds a model or an AI agent is trying to build a model like that if we don't have that Foundation it's not going to work you guys have any insights I think in terms of like justifying headcount or like advocating for kind of a new area it's like for me it's like it should be so painful that like you don't have a capability that
• 12:30 - 13:00 you know data science can own or data engineering can own and like you've also maybe done approv of concept like I think that's part of what the prioritization is for any team it's like you almost want to like show before you get that that it's actually going to work and it's going to be worth it because it is a risk right like adding a new headcount having a whole two and a half quarter work stream stood up and start running it's like maybe we can start going quickly to figure that out um or like really have many projects
• 13:00 - 13:30 stacked up back to back to back to make sure that it's not just like oh I have a new project I need someone for this it's like well what's going to happen next quarter and that was how in the zerp era people would get head count it's like oh like this new team farmed like they don't have a data person let's add a data person to that and then all of a sudden another team forms and another PM gets hired and another data person gets hired and you have like 80 data people for you know 80 PMs and it's like that was unsustainable every every PM needs their pet of of course yeah I mean
• 13:30 - 14:00 especially how are they going to PM without it you know I remember during review season especially I'd get like thousand ping mean like was this product release impactful tell me exactly in how many dollars amount I made uh it's like can the chart can the chart look like this yeah can you just yeah I need to put this in my uh in my review you thought about building a mode in HEX that's like it detects whether the Y AIS is positive or negative sentiment and it just kind of gives you a little boost in the chart whichever direction you know I think that'd be a popular feature you know a funny a funny experience I
• 14:00 - 14:30 have in my life is I still hold on to this identification I have as like a data practitioner having been on like sort of one side of it and helped support people on one side of it and then now I'm like I'm the PO sppy I'm the CEO like the headcount requests yeah how big is your D come to me uh our our D team's five people now but um it I will say that I find it interesting to like self-reflect on this that the most persuasive headcount
• 14:30 - 15:00 arguments don't come from our wonderful incredible transcendently amazing head of data they're they're coming from like the the partners that the data team is working with it's like when the when our when our uh transcendently wonderful marketing leader who's here is saying like I need I need I need like data support for this it's like that's when it gets really visceral it's not like the data team and there something interesting about that like stakeholder Dynamic of being in data that it's like it's tough to talk about sometimes because you're almost acknowledging that
• 15:00 - 15:30 you're a bit of a service org but I'm curious to hear you guys talk about this a little bit like to and at coinbase like how do you think about that Dynamic of it because there's like there's like the hey I want to hire great data people who are going to build analyses that are right at some level that although it like really matters that like the people you're working with are very happy and like how how much of that are you thinking about as you're working with your team so let me actually take a second to sort of answer um just like comment on
• 15:30 - 16:00 like yeah it's all connected yeah um so I have relied on uh in the past three years because like I said coinbase is very different than lift at lft I never had to or none none of us had to ever sort of fight for data Investments right but coinbase I think like the the the rough heris stick that we think about is you know what sort of like pm to DS and I'll specifically talk about like product I mean marketing also needs DS but like you know what kind of like sort of PM to data ratios make sense right
• 16:00 - 16:30 and so U what we would do is like we take numbers from You Know Places which have done well like you know meta Netflix Etc which are mature businesses we know that data did well there and like you know that rough ratio is somewhere between 1 is to one or 1 is to two DS topm ratios at Netflix and meta so if the area of the product something like Risk which we know requires metrics requires trade-offs requires a lot of sophisticated experimentation and it's like a really sort of nuanced balance
• 16:30 - 17:00 hey we should be looking at one aspirationally we should be at one is to one right if you're launching a new product and all you care about is like sort of like tracking usage and like sort of weekly active users maybe you can start with like one is to four right you know and then as we figure out that there are needs to do like I said more like tradeoff decisions more experimentation like you know is there like opportunities for actually sort of automating stuff through ml models Etc we should aspirationally move closer to
• 17:00 - 17:30 that one is to one ratio so that's the way we think about like are people are people bought into that like when you when you bring that to your Finance team or who whoever you're sort of like presenting that to are they're like yeah numbers make sense um like how you how are you kind of backing that up those ratios up so you know I think uh we have been on a journey this was uh so so it depends on like you know what where people come from right so somebody who comes from meta like they will buy this argument for sure right you know so somebody who's Maybe had like you know their career in startups you know like
• 17:30 - 18:00 you know they've seen like different patterns and different models it's a different sort of conversation I think we've gotten to a point culturally where everybody is product leadership engineering leadership is roughly aligned now we do have certain areas where like you know product is not like the main stakeholder it's engineering right you know so for example and that may be the case with you know uh one of our areas is we have to manage liquidity right you know how liquid should our assets be that's not a product thing it's an engineering thing and so again
• 18:00 - 18:30 there you have to sort of like the rubric that you use is do you just need foundational metrics and that's it or do you need to make trade-off decisions there's a certain sort of unique cognitive repertoire that like data scientists and data people have can that contribute to the business if that is the case we should be sort of improving that ratio um that's the way we model most of our investments uh you know into it that's interesting and so then shifting that toward like the the consumers of that like when you
• 18:30 - 19:00 think about the stakeholders you're partnering with like you know I don't know who the stakeholders for the liquidity decisions are the I don't know head of liquidity or something like but like but like you know how how do you think about that Dynamic of the job like I I think some people call this like politics or there's like other words for it but like how how much of that are you like working and coaching your team on or how much of your day is spent on like the sort of like delivery and buyin and like NPS for lack of a better term like
• 19:00 - 19:30 of the stakeholders you're partnered with yeah so um so I don't know if I'll have um a very structured response to that but I think the sort of different pieces of that when I think about it is so you are meeting with your stakeholders regularly you're also hiring a team which is very curious and passionate about like the business problem that you're solving and that is something that we've actively tried to do because you know there were people who are very happy being in a support function being Consultants just doing reporting but then you bring in people
• 19:30 - 20:00 who are like you know this liquidity problem that I was talking about right you know so we have to decide how much of our assets need to be liquid because that exposes us to risks like the DS who was on that project came to me and said you know I think we're doing it wrong right you know I really think that there should be some optimization of how much we keep in hot storage and cold storage and these are crypto terms but um and so it's my job as a DS leader uh to then go to the engineering manager because you know they're the on who are running this and basically say look we've done some
• 20:00 - 20:30 offline simulations we can save X millions of dollars if we do this right and that's very tangible right I don't always have that in my pocket in every conversation in this case it was very tangible to hey let's further this partnership I'll put a more senior person on this project but we are looking at potentially x million dollar in sort of savings with uh proper liquidity management so that's the kind of like sort of conversation you have but like like I said I also want to be sort of a little bit you know candid
• 20:30 - 21:00 here that like not every engagement works like that like some of these engagements are hey we launching this new if if folks remember coinbase launched launched a nft marketplace right you know when nfts were a thing um then uh remember that that was crazy yeah yeah exactly wild like AP uh smoking cigars and all of that that was kind of cool um go go so um when they were launching that so I was uh the data team supporting
• 21:00 - 21:30 that launch I was like you know what are we going to do with this right you know it's like we just need to sort of figure out how much traction we are getting uh so there my point is like okay you need like some amount of reporting some amount of like you know sort of either regulatory reporting or like you know just financial reporting and those are foundational needs and this is my estimate of what is needed to like you know just support those basic Regulatory and financially and then we figure it out if thiss business picks up and you know there's opportunities for like you
• 21:30 - 22:00 know doing things getting sophisticated with the metrics looking at like sort of experiments like you know doing some tradeoffs we'll invest more how about you guys yeah I think at some point brex is not hiring at all or like very little and the question to the folks who are asking for data people are like you want to convert a headcount and like that was a good test because ultimate limus test right yeah like you're trying to hire another marketing person or engineer like how important is data to you like do you want to make
• 22:00 - 22:30 that person a data engineer or data scientist that I think really helped how did that go uh what what was the answer not a lot of head count on the data team but I think like how did that make you feel I think it I think it helped clarify though like it is tight and like it should be an expectation everyone in their role has some aspect that they can selfs serve hex was a great tool for that for a lot of our Engineers like why am I hiring a data analyst to do a dashboard for this launch like I built the tables I know the data I know all
• 22:30 - 23:00 the joins I can build the dashboard and I think that kind of like helped alleviate some of that support um and also made it clear like that's not going to be like a role that we're going to saff for every single launch at this point like we're going to be more choosy we're going to have that level of headcount like threshold of like would you convert an engineer like would you kind of do that and so I think it helped communicate that message of like this is the new state of world like everyone
• 23:00 - 23:30 talks about D being data driven everyone wants to selfs serve it's like now's your chance like come on put up your shut up baby some sequel let's go just to just to to go a little bit because I I I talked to so many people about this type of thing like today or as you were going through that process like you was like do you want to convert a headcount um people would you know sort of hem and ha were there where where are the places or what what do you think are the the shape of it where that partner is like I
• 23:30 - 24:00 would convert a head counter like or maybe it's just the current data people that are supporting them they're like you know I need these like well how do you think about that in your role are there things you do with your team to try to like maximize that sense of essentialness I think like one area that was recently like this is on the finance team for us like we were just seeing a lot of spreadsheets passed around a lot of numbers that were like should be in a table somewhere a lot of like calculations happening in those
• 24:00 - 24:30 spreadsheets at like at some point it's like are we going to hire another financial analyst when we should add add a data engineer to this and like put things into tables automate a lot of these manual decisions that are happening at least so that when we do come around and maybe build a model we can trace it through and actually maybe model some of that stuff so I think it is like that point of like you start seeing that exhaust of the non data person's work and like it just accumulates and accumulates and I think
• 24:30 - 25:00 at that point our our CFO Ben and I talked and it was like obvious I think I think it has to get to that obvious Point um but there's a lot of people that like yeah they get they they think they need it and then you challenge them and it's it's like yeah well we can kind of get away with not having it or okay how about we just get 20% of this person's time and it becomes not a head count so I think that that is kind of how things ends up shaking out you need some amount of data work in that space that like could use a data person I
• 25:00 - 25:30 think I think that is a good sign if it's completely started from scratch it's like maybe you don't need aone oh yeah we're doing great uh I think what we are coming to is that um I'm gon to go back to data Foundation my favorite topic but um we would love to unlock a lot of the self service value and and the capability ities but we we
• 25:30 - 26:00 have to fix some of the data that's coming into to the data warehouse and and that and that and thankfully we're not the only ones who have this realization so engineering partners and product Partners have the same and so really focusing on how do we staff the right team to go fix it so we can allow for for these types of solutions to exist versus today where we have to staff it with data people because the data is so messy that we like it just takes a lot of time to figure out what the right answer answer is um and so up
• 26:00 - 26:30 till now I think and I say this because I wasn't you know I haven't been in Guster for as long but we definitely fixed some of it with putting people on the problem and we know we all agree that that's not the right solution long term um and that's how we're thinking about it going forward and and we actually going through annual planning right now and talking about how we're going to take a step back and think about how do we clean up our data um coming from our source systems coming from our Salesforce instances coming from any other you know external uh uh
• 26:30 - 27:00 tools that we're using as well so that that we can allow for not allow but feel more confident that somebody going in and building their own version of a dashboard or looking at the data and building their own um joints is actually going to get the right answer or hopefully we'll get a better answer than what they can do today less wrong less wrong directionally correct directionally correct yes that's right what more can you hope for but and how much how much of your time when you're
• 27:00 - 27:30 talking to the people on your team you're coaching them when you're thinking about who you're hiring how important is it that you're bringing on people who have you know they're not just doing a good job on the analysis or the data pipelining but like the sort of stakeholder almost like the sales ship of feeling impactful as a data team is that something you you think about or we do for certain roles the way we've been thinking about it is that um we have certain roles like data analytics that are is our embedded team they're
• 27:30 - 28:00 the the customer facing part of the data team and so we really think about them as the ones who represent data as an organization and the impact that we can provide regardless of whether they're the ones who are actually doing the work or if they bring it back and a decision scientist has to build the model or data engineer has to build like a solution and so they're the that's where we kind of put the a bit more of that expectation and pressure we talk about like internal PR and how do we think about what you know what Solutions we building and how do we communicate that
• 28:00 - 28:30 um a little bit less on some of the more technical indoor specific roles like I might not care as much about my risk modeler if they can go out and Advocate on on the team's behalf so like we started to really separate what those expectations are depending on what kind of work we expect them to do interesting internal PR you said is there is there like is that like a term you made up I've been using it for a long time where's my Square people what's your other the thing internal PR it was a thing well how do
• 28:30 - 29:00 you how do you do internal oh God we're we're bad at external PR we're PR in general was really hard um I actually literally yesterday I think asked my our CMO about if I talk about governance most people just start yawning do you have a better term for me he said he'll get back to me he hasn't figured it out exctly second there I know we're like we know it's important but like we need to Rebrand it so well people don't care about until the numbers are wrong yeah which are all
• 29:00 - 29:30 the time so I don't know why they don't care about it yeah IPO allowed to say that you didn't say that I know but like um but I I I I'm I'm obviously very interested in this because it's it's um to me it's like this Paradox on data teams it's like you a bunch of really smart people they're working really hard they're doing work that you know is like correct and right you often think it's moving the need all in the business then you get to annual planning or you get to headcount planning or CP winter hits or whatever you know happens in the
• 29:30 - 30:00 business where all of a sudden it's like hey why do we have this many data people or why do we need more data people and like I I've come to this belief I was alluding to this earlier that it really comes down to like the people around the data team are are advocating for it but that does strike me is it requires a level of PR adroitness that is not necessarily what is like selected for in data team interviews true I we do talk about it a lot uh and do we interview for that probably not or
• 30:00 - 30:30 maybe we should in certain cases we do talk about data St storytelling which is not sort of same but you know PR internal PR is probably a little bit lighter than that um but we do talk about rep how do we represent our work um I'm very lucky that that team at Gusto has been really amazing and brainstorming and and thinking about like how do we change it and knowing that it's something that we need to continue to improve um and so we have some ideas of what we're going to try to
• 30:30 - 31:00 do um I don't know if I have a perfect solution but we've tried many things we've tried to run like at Square we tried to run qbrs which were a lot of work um and they were they were pretty good um makes sometimes better than others but like trying to figure out how do we surface our work the problem for data team too is that a lot of our work ends up being surface through other teams right like if you're partnering with a liquidity you know we have actually a similar model which is so interesting but like if you partner with somebody who leads that team they're the
• 31:00 - 31:30 ones who actually going to present the the overall result of it so presenting it again as data work becomes kind of challenging and so how do you find similar but different inside that has been presented already do not allow do not look at the numbers until we present them that's right only the data people can do that um okay I got a couple more questions then then we'll open it up so I want to build on this topic on people interviewing skills I made a new guys I made a new friend tonight where's Allan I was asking there is um I
• 31:30 - 32:00 was asking about what he's interested in he talking about um the idea that data rols are going more full stack I think there's other people who it's an interesting debate I've talked to other people who think that data roll is going to get more Specialized or narrow there's a lot we haven't really talked about this much surprisingly about AI So like um and how that you tried you did um um that's going to change data roles I'd be
• 32:00 - 32:30 curious just to get you guys take on this as data leaders like what do you what are you interview for what are you thinking about and especially I'm especially interested in how that's changing are there things you guys are looking for now that you weren't before are there things you think will be more or less useful like I my personal thing is like you know it used to be that it really mattered that you had like a really thorough knowledge of the pandas syntax if you're like a data scientist it's not clear that that's like the most important thing anymore but like does it that you know what it means at all like probably like so it's like this stuff's
• 32:30 - 33:00 evolving and you know AI can write SQL now does that mean all the data jobs go away yeah uh I I roll my eyes as well I like but like I I'd love to hear just your guys point of view on like what's going to happen with data roles specifically like what's this look like over the next few years I'll toss it to you first yeah I definitely think like it should conceivably get easier to do any of the data roles if you have like basically the base knowledge of being
• 33:00 - 33:30 able to SQL and like SQL really really well I think that you mean easier if like the technical bar goes down because yeah because of AI right like you can ask it to do random forest and like if you have a clean data set and you've done that work which is not easy work something AI probably can't do and we talked about that a little bit in the pre interview thing but I think like it is really important to get that technical knowledge of how data goes in and out of a system system how tables relate how to find like the right
• 33:30 - 34:00 information how to feature engineer that's probably what I would care the most about is like the Curiosity of like and honestly like passion for data because like if you don't have that part like the sequel the AI stuff like it's not going to click in the same way that that role will be a data role like you can do all that stuff as a non-data person but if you don't really have that like I don't know passion understanding and also the technical skill on the sequel side probably the most at this point like um along with understanding
• 34:00 - 34:30 the actual mathematical Concepts behind some of the modeling techniques and and how to create data infrastructure um it's kind of you're you're never going to get there I feel like well on that last point though you you said like yeah you know the the llm can do a random Force for you now sure it can but like it's I'm not sure that I would want a data scientists on our team or an MLG whatever the title is like being like do build me a random force and they build a random force and it could actually be like the code runs it's good but like do
• 34:30 - 35:00 you know what this is doing is this the right model to select can you interpret the results can you update it can you maintain it like it's it's an interesting question like would you would you want to hire someone who like didn't know how to make a build a random Force but is like just able to invoke it via magic I guess like you would hopefully have enough of the background on all of those things that like that part of like General in not just that model but like six other versions you might be
• 35:00 - 35:30 considering and then evaluating it like the threshold in terms of knowing maybe also comes down but I would hope that like I think probably that's the answer to the question like if you're looking for what you should learn as a new data scientist is all these conceptual Concepts in a really deep way and at least one thing technically deep uh sounds hard honestly uh but like that's because because you don't really have to know all the steps between A and B anymore cuz you can hopefully get an AI
• 35:30 - 36:00 to do it if you know at least how to pick like the right result and this result and interpret those results you can probably throw out that random force it doesn't make sense I would I would think that that's some of the work we still have to do right like you know you might try to do all this pre-work to determine what the right model is you might be down to two it's like okay usually it takes a lot of effort to build one maybe there's now a possibility to do two faster still got to pick so yeah that decision Point becomes a little bit faster and earlier in the process which
• 36:00 - 36:30 is kind of exciting it also kind of makes that like Choice a little bit easier May from the data scientists to what are you how are you thinking about data rolls what's going to happen what are you guys looking for and not looking for in the roles you're yeah there's um there's a lot of different threads in my head but I let me see like the ones that I want to talk about is what does it do for data hiring um so this is kind of like I'm going to use the you know the often used metaphor of freeways in LA
• 36:30 - 37:00 right you know so you broaden the freeways does the traffic go down no like there's is more traffic right reduce demand yeah so the thing is that like most data teams are overwhelmed it's not like there isn't enough data stuff to do right so there is this myth that like you know AI will come and then like all the data problems will sort of like resolve themselves and you'll need um fewer data scientists I kind of feel like this's like you know we can now actually take on more of the backlog right because we just like so
• 37:00 - 37:30 overwhelmed handling like ad hoc stuff coming from our stakeholders all of that will probably be handled by Ai and we'll be doing more higher value ad stuff I agree I think it's I think it's funny when people talk about it that way because it's like I I I very rarely encounter a data team where they're like yep we can get the answer back to the stakeholder quickly on time there's not a line there's like it's like so so you guys are behind you you can't possibly keep up with the demand oh no absolutely not and you're worried AI is going to it's
• 37:30 - 38:00 like it's that's right strikes me as a long way off that but right so so I think you know what my hope is aspir aspirationally what I want for my team is that we'll be doing more specialized stuff so because like lot of like the ad hoc stuff we tend I mean like you know I think that's the story of maybe every other data team um we tend to get overwhelmed by a lot of like the ad hoc stuff right uh we are in a very fast moving sort of business and there's a lot of like regulatory stuff happening that's true probably most inexs um and
• 38:00 - 38:30 so my hope is that you know we'll be doing there's a lot of like you know sort of things like experimentation and causal inference and like you know offline policies that like we should be paying attention to that we don't pay enough attention to and so going back to your other question of are we hiring differently we absolutely are hiring differently we are and that sort of depends on like the role so most uh companies uh sort of coin size companies have data scientists senior data
• 38:30 - 39:00 scientists staff data scientists uh at the data scientist level you're looking at sort of like technical execution abilities and that's kind seems tends to be the core focus when I'm looking at an L6 staff data scientist I'm looking at somebody who can frame problems right so that wasn't the case three years back it used to be hey can you handle a lot of stakeholders and like take in like a million requests at the same time no but the emphasis is now changing on like there's a lot of stuff like a lot of the boiler plate stuff can be taken care of but can you really embed yourself into
• 39:00 - 39:30 the business problem and frame the right business questions that need to be answered right and we are looking for that we trying to sort of like design our interview Loops so that we are testing for that during the inter Loops that's really interesting Julia are AI gonna replace your whole team no but I do I mean I I agree that I I we do hope that some of the more like operational repetitive sort of metrics that we have to pull over and over again
• 39:30 - 40:00 if we can automate ourselves out of that I don't think anybody will complain because there's so much Insight that we can find from our data that we're not able to even get to today um and we think that there's a lot of interesting nuggets and problems hiding in there that we we can't even get to because we are inundated with ad hoc requests all the time so like I hope AI gets to the point where we can trusted enough to go do that so we can spend our time elsewhere and I think in terms of hiring and I don't know if that's changed necessarily I think it's always been
• 40:00 - 40:30 something that we try to hire for but you said there Curiosity I think technical skill you can learn or maybe you come with it but I think the the biggest value that the data person can bring to a team or project is curiosity and asking questions of being open-minded and if you can figure that out during interviews how curious that person is are they asking the right questions are they hearing a request but thinking about what the underlying problem problem is that somebody's trying to solve like that is an regardless of
• 40:30 - 41:00 whether your decision data scientist data analyst data engineer if you have that ability and and like um the the trait that's going to make you like llm is going to be you're gonna be fine with any sort of technology that comes out that's what makes a great data person I think I think it makes an incredible data person yeah if I if I may just add one other point is that we talk about AI quite a bit U but also the tooling now is so much better and this like you know
• 41:00 - 41:30 appropriate moment to for a plug for hex is that I think thank you you but send you we'll get you your bottle of wine you know AI is still in its sort of you know I won't say in its infancy but it's still evolving and we don't exactly know you know I'm very optimistic about what shape it'll take but it's not quite there yet but what has happened over the last five or six years or maybe a decade is that you had this like sort of like sort of camean explosion of tools and then like there was some tools which one
• 41:30 - 42:00 out Etc and we were I mean frankly even three four years back we were using like sort of bi tools which were leading to this they were making this problem worse right so ad hoc request because Self Serve wasn't possible it took like data scientists and data Engineers forever to turn requests around so I also want to say that like you know a lot of this change is also coming because the tooling frankly is getting much better and there's also like I think think there's clear winners at least in my mind I already named one but like you
• 42:00 - 42:30 know there are other sort of clear winners in each category which uh I sort of think about it's it's EAS easier for a data leader to make those decisions today and that seriously like you know that has improved the productivity of the team significantly well what a wonderful not and you have something nice to say yeah no not about hex uh no I just like about data uh what a cool job like everyone comes to you every day asking for like help like that's like not that bad like
• 42:30 - 43:00 I feel like I think that's like one of the things that you're talking about like what makes a great data person is like okay if everyone every day comes to you asking for the same thing like do you just give them that or do you actually understand their problem do you then provide them a solution that changes the way they work and like you get that opportunity according to our slack channels like eight times a day at least nine maybe more than that today I think that's like exact kind of where the support role of a data person is
• 43:00 - 43:30 opportunity for driving Roi and all the stuff we've been talking about and the day that stops is the day AI has taken over and like I guarantee it's not anytime soon yeah someone someone said one time they they have another data leader said we did one of these events with um they love haate relationship with the questions from the business it's like man it's so annoying we get all these questions but uh the question stopped so cool all right well uh I'm going to open it up I don't if folks have um any questions for our fabulous
• 43:30 - 44:00 panelists who's going to be the first someone told me they were going to be a plant and I forget who it is now Jess hi I'm Jess I work at vran we are hex customers as well um you really no one needs to say that it's okay the logos right there we're plugging ourselves funny cool um I was wondering how you keep your data team
• 44:00 - 44:30 motivated she asked how do you keep your data team motivated Sarah are we M motivated right I'm taking them to dinner next week um occasional dinners there you go Jess we feed them I think I think what you were just saying actually the the the the right if you hire the right people which is you know goes back to how do you do you find
• 44:30 - 45:00 people who are excited about asking questions knowing that you get to solve these uh problems that are vague and open-ended in many cases and seeing that impact that you know maybe you did this body of work or it was a model that you built or a question that you answered that couldn't be answered before it's like it does take the right type of personality and so in some cases it's like it's there's some motivation that we can provide as leaders but at the end of the day it's the type of problems that we can prioritize for the team that will allow
• 45:00 - 45:30 the team to feel like they're contributing and I think the right type of Personality will find satisfaction from that and I think the other part of it is giving the team some freedom to experiment and and drive their own work like how do we balance the the day-to-day needs of the business and in some cases of course there's a lot of work that is like feels very like a drag but how do we give some space and flexibility to play was different tooling to try different options to play with llms to maybe look at some data
• 45:30 - 46:00 sets that we haven't been able to look at before to to kind of feel like they have creativity creative outlet and so I think it's a balance yeah I think everything you said I would agree with uh I think also just learning from other really smart people on the team I think like that's something I get motivated by just kind of see everyone's work it's it's a good thing about the position I'm in I can kind of get a p like a really close look at this like incredible team that like
• 46:00 - 46:30 figures out Ai and how to apply it to this new use case and I think for me like to motivate other folks it's like hey look at this cool thing is kind of how I try to go about it alongside everything else you said uh so when I think about like the personas you know on a data team uh so there's maybe like four basic sort of types of personas right so there's data Engineers analytics engineers decision scientists and applied scientists and so I will speak
• 46:30 - 47:00 more to the latter twoo because that's the world that I'm closer to I think you know what they get motivated by is ultimately business impact right so they want to have bidirectional relationships with their product engineering marketing stakeholders and they want not just to be at the receiving end they want their thoughts heard like they want to be able to sort of influence product strategy Direction Etc now what gets in the way is like one of the things we talked about like there's all this ad hoc stuff so how do I you know I can't do anything like strategic you know so there are ways of
• 47:00 - 47:30 addressing that um but like the other thing is that like just the personality types on data teams like you know so we had some folks on our team like take a personality test and they put you in like these quadrants it's a very homogeneous group everybody is pretty cerebral and like introverted and that's not great for I mean that's good like you know for doing deep work but that's not great for like you know evangelizing and like sort of talking about like so that also gets in the way right so as
• 47:30 - 48:00 data leaders actually I'm also that I also definitely fall into that quadrant but I have to come out of that like sort of you know uh that's outside outside of my comfort zone what we have to do is enable those like sort of bir directional relationships with our stakeholders that I think is like super important to like keeping folks motivated they want to be sort of solving hard problems and all of that but ultimately like you know this feeling that I'm having impact on the business is is a is the best motivator yeah right your work matters it's doing
• 48:00 - 48:30 something someone appreciates it it's true for everyone I guess more questions hi hello I'm J and with Checker um it's a question to get a perspective on how Ai and your strategic how you think strategically about your road map and your Investments they they kind of intersect I think the narrow version of the question is do you think it's still worth a Strategic investment in cleaning a warehouse semantically kind of structuring the data as the AI
• 48:30 - 49:00 can help you figure that out for yourself and more broad version of that question is like what is the kinds of things that you you find now exciting strategically that you're willing to invest given that AI is going to change everything I have a very strong opinion about that let's go um I actually do think that the investment in cleaning up the data and building the semantic layer is what's going to enable AI to be successful I don't think AI can be without doing that and so to us it is the investment that
• 49:00 - 49:30 we making um of how do we you know do it in a contain enough space so we can really prove to ourselves that it's going to work and either way that's not we talk about no regrets work like either way we need to do it to allow self-service capabilities and if AI can then take it even further great but we do know for a fact because we've tried it many many times that if you put the AI on bed data you're going to get really crazy answers and it's really hard to know when it's actually going to give you that answer and so the thing
• 49:30 - 50:00 for humans you put humans on bad data you're going to get really crazy you know what to yell at it's true comp the human I think I think the difference is um the gut check of humans is a little better than AI um and you know I I've been joking about it that we we would see SQL statements that have like extra columns and extra like we Clauses with fields that don't even exist in our underlying model and it's basically AI telling us our data model is not good enough apparently like why don't you have that trying to help it's totally
• 50:00 - 50:30 helpful so it's it's it's interesting where and how it hallucinates which goes back to like yeah it's it's why it decides to do that I I don't know if I know that but you know my my big assumption and I'm I I'm even if I'm not right today I'm hoping AI will catch up by the time we're done AI will catch up with us that having that semantic layer and having that clean data structure will allow for that and will allow for other tools to work better as well so either way it's going to be a win for us yeah I think exactly like we're
• 50:30 - 51:00 trying to do all those things like not for the purpose of AI but with a hope that like eventually that an AI will come and will actually be able to do this stuff it's like a religious every I feel like every time I've yeah preparing for the day in which it will we all know it's coming AGI uh not making a prediction on one uh but yeah I think that is kind of like it has to be table Stakes at this point for a data team to care about that not just for that but also because you know there's a lot of
• 51:00 - 51:30 benefit to doing all those things including like owning your own semantic Clare not having it only being looker and all these other things that I think have now seem archaic if you were to redo things from scratch and and most new companies don't do that I think something that we're thinking about for sure yeah I don't know if I have anything uh significant to add but you know there are sign there are a lot of Investments that we making to in prepare to prepare right you know so um we had
• 51:30 - 52:00 really poor metadata and that was okay because when humans are working with tables Etc they can reach out to other humans and say hey what the hell does this column mean right um so cleaning up metadata cleaning up our data models especially for areas where there's a lot of like you know stuff can be self- served the the blocker today is I mean you know yeah the models are getting better and they're able to reason better Etc but the blocker today is our data model are not there yet we don't have like metadata or governance in place
• 52:00 - 52:30 where you know you can trust the results right so a lot of Investments that need to happen like you know uh just foundational Investments and this is kind of um the the other part is that um the engineering teams for example are using co-pilot cursor Etc um the data teams are not quite there yet there isn't really a cursor for data yet right I mean so we we probably will get
• 52:30 - 53:00 there but I think like part of that again has to do with I think part of it is like thinking about like these self- serve opportunities but also thinking about like the productivity of your data teams themselves how can they you know write this hex magic what I think like you know gets in the way of maybe hex magic being you know as effective as it could be is because you know it just doesn't have like the clean data models and the metadata to work on right totally I mean I think that is the Paradox of AI and data like um you we we obviously think a lot about like I'll
• 53:00 - 53:30 actually give an opinion on this because I think about it a lot um it's like we um like it the the the hard part for AI and data is that like it really is important that you're right and like the the AI features we have today that are basically code complete can be a little bit wrong and still generally useful it's like directional it's the same as software engineering As you move up the stack to to sort of high levels of abstraction like being right's really important so like we're spending you
• 53:30 - 54:00 know you guys think about this a lot actually at coinbase we like talking to you guys about it they're spending a lot of time on like Hey how do we have the AI features be able to rely on a semantic model that we're synced in from that you're controlling that you're updating and have that be accurate and then have it actually tell you what it did and which part of the model it really it's like a lot of this that goes into it that's it's not even just about the like increasing the accuracy but it's also like helping the user understand why did it use this model model what did it do what was the sort of thought process to the extent there
• 54:00 - 54:30 was one and I I think that's just like very hard to get right um cursor cursor Works in software engineering because like vs code already exists there's they're doing a lot of really impressive stuff it's not quite so simple in data and um of course we we think about that a lot as tool Builders but I think it's very interesting for us when we talk to data teams that are s like starting to think ahead on that on both like how we're organizing our data the skills as we just talked about that we're bringing in like how we're um assessing tools it's all It's tricky because it's like this
• 54:30 - 55:00 space is moving so quick too that you feel like you're kind of aiming off into the distance um and it's it's a it's a very interesting moment I'm not sure there's any there's been a moment quite like this in my career at least um on either side of it okay we'll do one more question and then we'll we'll liberate everyone yes hi hi I'm Rebecca and I work for a company called Kippy where a snowflake system integration partner um so I'm on the services side
• 55:00 - 55:30 so I love it when my clients call with ad hoc requests every other week um I get paid for that but you have to deal with these ad hoc requests I'm curious around do you have Frameworks or how do you influence your executive teams it's the beginning of the year how do you say like here's how we're going to go about this this year here's how we're going to you know manage the the business with data like how do you
• 55:30 - 56:00 influence these teams to leverage the data really effectively because I'm trying to help these clients like not just have these knee-jerk reactions when their stock price Falls 10% one day and oh we need to we need to have this whole new dashboard like I can only imagine what you have to deal with internally so tell us about the Frameworks or how you plan strategically for you don't get paid you you don't get paid per
• 56:00 - 56:30 request sounds nice so um I can I can take a step at this um so you know there's a framework that I use um which I don't know I don't want to make it too well known because I don't want everybody to use it but it's basically starts with threats and then obligations and then opportunities right so when you think about what I want to do threats are number one right you know and and in like sort in my world you can
• 56:30 - 57:00 imagine what like the sort of threats could be right you know so hey we're going to get defrauded by like sort of billions of dollars if you don't do that right um and then there's obligations obligations are typically regulatory obligations right uh so threats everybody signs up for obligations yes people are and then opportunities which excites me but like that's the one which requires the most convincing right so that's like the first framework the other framework is you need to have a
• 57:00 - 57:30 very transparent sort of like way in which you prioritize stuff like so even when we triage requests like you know what is this is this like a compliance thing if it's not a compliance thing how urgent is it who's asking for it what's the ROI on it right can I put a number on it like this is like you know how much effort will go into it what's the kind of sort of dollar value of that and and I think like for any new request that comes in like that should be pretty transparent right you know in the sense like everybody knows like why my stuff is not getting addressed while you're
• 57:30 - 58:00 working on like sort of Bar's request right so um you should always be working on my request yeah all right what about you guys what are your Frameworks yeah I guess I think for us at Brax at least it's been a lot about the incentives in place for either those Executives or those teams that like Orient around data I think like it's been really eye opening to see when someone is committed to a number how many questions they have about what that number is this month and why it moved like they want to know exactly how it's calculated and exactly
• 58:00 - 58:30 what drives and how they can impact that I think like it's really clicked at least a lot on our GTM team where we don't have to advocate for them to care about these Financial metrics they're literally getting paid on them and they they want to know exactly what it's going to be at the end of the month and our job is to make that faster better more reliable but also to educate them on the things they can do to make that stuff better so I think like that incentive doesn't exist it's hard to necessarily like completely generate that although we try every single time with some of our epd colleagues uh and I
• 58:30 - 59:00 think that that's kind of something we've learned it's like all right it's like if that's really clear we can definitely justify all the investment and the head count and everything around that U and make a big impact on that so at least in terms of like getting people to care about data and like in advance I think aligning incentives honestly like is the simplest way but takes a big commitment from leadership last word you know it's not going to be one word um
• 59:00 - 59:30 figuratively um I actually I agree with that and I would also say that we do spend quite a bit of time aligning with the leadership of the company and the product leadership and GTM leadership and what the priorities in okrs for the year for the quarter R and that's kind of where we are all going to try to focus and even though there might be and we're not quite there yet this is what we're trying to do so I want to be clear like it's easier to say than do but really trying to allow the data team especially certain parts of the data team to focus uh because that's where we
• 59:30 - 60:00 know the biggest value is going to be and so part of the framework is what is going to be the value of this work what metric can we impact if we do this work and if we can't identify it should we be working on it um but being clear about what we're working not working on what you were saying and like having transparency and how we decide or Andor escalate when we can decide ourselves becomes very important so it's never surprise um to our stakeholders wonderful well that is a
• 60:00 - 60:30 great last word to leave it on stakeholders cool well I want to thank uh you all again for being up here with me thank you for doing this with us thanks everyone for coming out I hope you enjoyed this uh we'll hope you see we'll hope to see you at the next one of our events and have a wonderful rest of your evening and a beautiful week thank you thank you [Applause] it's