The Data Stack Show - 200: Data Team Struggles: Telling Stakeholders the Truth vs. What They Want to Hear (How to Tell The Truth, Tactfully)

Episode Date: July 31, 2024

Highlights from this week’s conversation include:Lightning Round Discussion (1:21)Data Team's Truthfulness (2:21)Culture as a Blocker (9:10)Misconceptions about Data Jobs (10:32)Cultural and Technological Influences (11:51)Challenges in Data Science Projects (15:19)Embracing the Process (17:23)Barriers to Entry (19:36)Hiring Data Leaders (22:06)Challenges of Data Leadership (25:38)Evolving Hiring Criteria (27:30)The Data Stack Show is a weekly podcast powered by RudderStack, the CDP for developers. Each week we’ll talk to data engineers, analysts, and data scientists about their experience around building and maintaining data infrastructure, delivering data and data products, and driving better outcomes across their businesses with data.RudderStack helps businesses make the most out of their customer data while ensuring data privacy and security. To learn more about RudderStack visit rudderstack.com.

Hi, I'm Eric Dotz. And I'm John Wessel. Welcome to the Data Stack Show. The Data Stack Show is a podcast where we talk about the technical, business, and human challenges involved in data work. Join our casual conversations with innovators and data professionals to learn about new data technologies and how data teams are run at top companies. Welcome back to The Data Show.

This is our monthly show with the Cynical Data Guy. If you didn't hear the introductory show last month, definitely go back and check it out. For those of you tuning in for the first time or hearing about Cynical Data Guy for the first time, here's a little overview. So every month, I pick out some really nice LinkedIn posts, and we get Matt Kelleher-Gibson, who's on the Rudderstack team, who came to Rudderstack from the bowels of jaded corporate data America, straight to Rudderstack. And he gives us his hot takes on these LinkedIn posts. So cynical data guy,

welcome back to the show. Thanks. So I'm back. Yeah. And of course, John's agency is Agreeable Data. So we do try to pit them against each other. All right, this is lightning rounds. So we'll get cynical data guys take. John, you get a chance to respond. Some of these LinkedIn posts, the authors will remain anonymous for obvious reasons.

Others will share. And I will just- For their safety and security. For their safety and security. For their safety and security. And their emotional well-being. Yeah, right, right. You can't expose everyone to the cynical data guy. That's fair. Okay, are we ready? Let's go. Yep. All right, first post, actually,

Clint Dunn, who we had on the show recently, this is a great post. I'm going to read some excerpts here. It starts out with a zinger. Your data team is lying to you. Data people are supposed to be truth seekers. Why would they lie? The reason is actually very simple. Data teams tell you what you want to hear. That's why your head of data rarely gives you unfiltered bad news. Instead, they'll hint at an issue and try to redirect you towards a better goal. It's a lie and pretty packaging. Yeah, that's very true.

But, so, yeah, they're lying to you. And they're mostly lying to you because you want them to lie to you. This becomes the thing where everyone says, well, I want Data to be a partner. It's like, well, you say that, but you treat them as to be a partner. It's like, well, you say that, but you treat them as a junior service partner.

So that's what they're going to act like. So, okay, give us a scenario. Don't name any names, but give us a scenario where, you know, I mean, lying, I guess, I don't want to put you on the spot and ask you when you've lied. But put him on the stand, ask him when he's lied. Take us to corporate America as a data leader, and you're in a meeting, and you face this situation.

Okay, so there's different types of versions of this I can think of. The one that immediately comes to mind is the one where you've been reporting something you've now gone through and realized oh man there was like duplication or a wrong source or like something has caused it and the number is about to get very off and there's off as in like from what it was before but more accurate right you're going to be now delivering accurate results, but they're not consistent with what you've been saying, right? Or what the company has been saying for a while now. So there's one form of lying, which is we just report it the old way and just don't even, and we come up with like version 2.0 of whatever the metric is and massage around exactly why, well, this is what we want to transition to and things like that.

The other version of it is where you talk about how the numbers are going to be a little different, but it's okay and it's going to be more accurate and it's still directionally the same. Oh, directionally. Directionally the same. Directionally accurate. Anything we've done in the past, we just, you know,

there was a little hiccup and now we fixed that that and you know and it wasn't our fault it was clearly someone else's fault not that they care lots of hand waving happened lots of hand waving yeah so the reaction to this change i mean well the reaction or what everyone fears the reaction is going to be everyone's fears this is like the we're all getting fired type moment like right that's just the way it is without naming anything i i remember working on one team where i had a boss that anytime we got a problem the first thing he said is what did we do wrong and we had to spend the next 30 minutes explaining to him how it was not our fault on any of this stuff. And it was business users or downstream, upstream, all those types of things. Yep.

The reaction, I think, from people who are, if you're not in that fearful state, is kind of the like, oh, all right, here we go. Especially if it's something where you figured out that there was like a previous iteration of this was done incorrectly. Or there was, you know, a lot of the like, oh, there was Gary and he did all of this was done incorrectly, or there was, you know, a lot of the like, oh, there was Gary and he did all of this and there was this mistake in his code. And so like Gary's code mistake, it was a Gary process that's what it is. Gary DB. And and it's you want to fix it, but you also know that you're going to get into this thing of like, well, those aren't my numbers and it's's like, well, I know they're not your numbers. You can't say because your numbers have been wrong and every decision you've made is off of long term numbers. That doesn't go over well. So you have to come up with these very, well, it's not that it was wrong or all these types of things to try to move around that. That's why I say, that's when you end up with just like,

you know, sales 2.0, because we can't change sales 1.0, even though it's wrong. Agreeable data guy? My first thought on this was actually in marketing in Eric's territory. So there's this really popular book in marketing called All Marketers Are Liars. Like, do you know that one?

Yeah. Seth Godin book, yeah. So, and his point in point in the book and he actually there's actually two different covers the originals all marketers are liars and then the new cover is all marketers are liars and there's like a cross through with like all marketers tell stories which is interesting fun fact gosh you know i was winning wow a point for one point for cynical data guy one point for cynical data guy i mean we may just need to move on okay actually which yeah the point the point being that whenever you're taking complicated data and simplifying it into a story a that's really hard right and then b like how do you do that such that you keep the

right level of fidelity or accuracy to communicate what you're trying to communicate and not get lost everybody get lost in the details yep it's like i guess on one level like you could call i mean i wouldn't call it lying but like on one level like when you roll up enough it's like you miss out on so many details that like yeah like you're not covering all of the things it's inherently reductionist when you're doing that stuff right like you're gonna leave things out right really just you know it's taking it from a book to a movie i can't include every plot in there i gotta make decisions of what i'm gonna put in and data people are the book is always better people generally right right? Because there's more context and, you know, whatever. And that's how you end up with the emails to a VP when they say, what was sales last month?

It's like, well, here's a chart with all of our sales products for the last full, you know, every day of the month. And they're like, am I supposed to add up every one of these numbers to get my answer? Well, yeah, it's in the chart. I just wanted one number. Right, right, right. Cynical data guy, we need to move on from this lightning round, but last word. In general, your business

stakeholders want the truth or don't want the truth? If it disagrees with them, they don't want the truth. Your face was unbelievable. The look of like, are you serious? I should have taken a picture. Okay, lightning round one done.

Moving on. The author of this post will remain anonymous. I did a poll last year asking data and AI leaders what the biggest blocker was to delivering successful initiatives. The overwhelming answer was dot culture. No surprise, really. Changing culture in an organization is incredibly hard. But might generative AI be the answer to solving all sorts of cultural adoption issues?

Join me live on the 16th of July to find out more. No. Next question. I'm going to go with the no next question as well. Oh, also hashtag culture, hashtag generative A. Hashtag wishing. Okay. That one was too good not to include.

Okay. This post, the author of this post, we will disclose, it's actually the cynical data guy himself. Okay, and this post went viral. So, I mean, hundreds of thousands of impressions, I don't know, 80 comments, however many. Not that you would be obsessively looking at the stats. No.

Okay, I'll read it. One of the worst things to happen to data was the sexiest job of the 21st century article. It caused two major problems. One, it attracted people to data and data science that were looking for high status jobs.

Two, it gave the impression that businesses had already bought into data-driven culture and had the tech ready for cool work. The reality was and is most of the job is hard, unseen work. There is very little splashy work. And second, data is a support function in most businesses and the infrastructure is nowhere near ready to just walk in and do quote unquote, or quote, cool work. I'll stop there.

It's a really good poster. Wow. Wow. Should I respond? It's his post. So maybe I should respond. You go first. First off, I have to derail us for a second.

I was going to look up an example on company culture and the top SEO spot for company culture examples is WeWork right now. All right. We should do a whole episode. Yeah, that's worth a whole episode. Yeah, that's worth a whole episode.

And that's why we're all going to hell. Yeah. Yeah, so the data science. Do you know around when that article came out? We're talking 15 years ago? It was like 2011-ish, 2010-ish. Yeah, yeah data science. Do you know around when that article came out? We're talking 15 years ago? It was like 2011-ish, 2010-ish. I'm probably off by a year or so in some direction.

I've actually been thinking through this a little bit personally because that is around when I started in data. And I think within a couple of years of when you started in data too. And it pretty much corresponds to um which i think was an influencer not the main influencer but a influencer is moneyball right so that was around that time too so i feel like there's all these like some cultural things and like some technological innovations it's a little downwind from moneyball because moneyball was like kind of 2003 or the movie yeah the movie and the book i mean

who reads book yeah i mean it's not like no you're right moneyball is like 2000 it's not like i read 30 books a year or anything yeah right um anyways all this to say there were like lots of like cultural wins and like data and like like moneyball the movie and sure other things like wow this and then you know i don't even know who wrote the article some you know forbes or whoever from like but coming off of that and then even the reality then was what percentage of the people were actually doing like cool data work there like that was i don't know dreamy then like that's that was what people wanted it to be right right and it was I'll just make up a number maybe 10% maybe less than 10% we're doing that and then yeah let's call it like 1% so then like you're going from that like oh that's so great like

that's going to be the future and then when somebody declares the future enough years go by and then like the obvious reaction to it is like oh okay like we're about in the same spot that we were yeah yeah so and there's a more tangible thing that came out of that because i remember when i first started having a hire and the number of people that if you put out a post that said data scientist on it, they would go to it and had outsized expectations for salary. And like we had one, when I first started as a manager, we were, so we're going to hire a junior data scientist, like someone, you know, out of school or something like that. We got about two weeks in and we were like, this does not work.

And we downgraded the role to a data analyst because we were like, this is just crazy. And this was also at the beginning of when all of these data science master degrees started coming out. Oh, yeah, yeah, yeah. Of varying levels of quality, much of it dubious. And we went to people who were in the pipeline

and we're like, well, we're changing the role. We want to let you know that, give you a chance if you want to go there. And we had a couple people that were like, I just really feel like I'm just a modeler and that's like really what I want to let you know that give you a chance if you want to go there and we had a couple people that were like i just really feel like i'm just a modeler and that's like really what i want to do i and i saw the same thing you know work at other places and there was one they were not they weren't paying market wages right right well we're going to make all these positions a data scientist wrong like don't do that because i'm like you're gonna get people who just want the title and they're gonna leave in a year when they have a year's worth of

experience to go somewhere to make a lot more well i was gonna be fair they want the title to make the money that's why i want the title i mean there's no status thing too but a lot of it's the money because the hr does like the indexes against market and they see the title and you get more money like it's a fairly simple equation at a mid to large company. And they had a lot of turnover issues because people would stay for 12 to 18 months and they'd go work at a bank as a VP because everyone at a bank is a VP and, you know, get a 30, 40% pay increase.

By the way, can I just call out that the agreeable data guy just had such a cynical take. Well, you did have like the number two comment on it all I know yeah really

yeah I didn't even look at that I know I know right and I pulled the post to go back and correct the record

Moneyball 2003 okay for the book and the movie was 2011 okay yeah so it's about right yeah

wow okay how many projects have you worked on how many data let's get specific because you've done tons of types of data works That's about right, yeah. Wow. Okay. How many projects have you worked on? How many data... Let's get specific. Because you've done tons of types of data works. Tons of types of data work.

How many data science projects have you worked on that you would describe as sexy? Oh, I love this question. Well, this is all on a sliding scale because we're all data people here. So, one or two. Wait, out of how many? Hundreds?

Yeah. Or at least dozens, right? Oh, at least dozens. Yeah. Yeah. Well, yeah, because you also have to remember, I've worked at several places where it was like, I spent a year going like, all right, we're going to fix this so we can actually do data.

And then I realized it's probably actually just the stand. That's everywhere. The analogy that I used is at the time when we first started like trying to hire people was I said, it would be like if you went into a law firm, right? And especially if you were like, I don't want to go to law school, I want to, but I want to work in law firm. And so you got a master's in lawyering. And it was all about how to be a trial lawyer and how to pick a jury and how to do a closing statement and all that. And then you walked into a law firm and you're like, I'm ready to go.

And they would look at you and go, that's like 2% of our work. And we avoid that whenever possible. Like that was part of my job as a manager was kind of like crushing data scientists dreams when they're like we could build a model or we could take the average the dream crusher i got i got to decide if i agreeable more agreeable take on this but i think we're pretty aligned on this one the only like i don't even know if it's a caveat is is that some some of it is i don't know that like some of it's a culture

problem sure like we laugh at that which ai obviously is gonna fix obviously so we can consider that a solve right but but i think's, and this maybe is kind of a culture problem, but I do think there's a component we haven't talked about of people really like to understand how things work. Yeah. And models are inherently opaque

unless you are pretty good at math. And then if it's AI, then basically no one knows. Except for Sam Altman. Yeah, just Sam. Well, I will say one thing about this, that like, cause there was some comments in that post too.

And it was like, I used in the post, I compared it to like being like a professional athlete where it's like, you gotta love to practice, right? Like there's a lot in this and there were a decent number of people who have gotten into this

and maybe they got into it because of that. And then they went in and found like, oh, okay, I'm gonna have to do a lot of sequel work. There's a lot of this stuff I'm going to have to do. And they, you know, for lack of a better term, fell in love with the process. So there is that out there.

You know, it wasn't all bad. It was just, you get to sift through a lot of other people to find them. And it makes the job, when I was a manager, harder. Like I put a lot of emphasis on like training culture and specifically on things like how to think because i was like i would much rather take someone i mean one of the best data analysts i ever hired a 50 year old woman who would come back into the workforce after raising her kids and the only reason i knew she was 50

years old was because that was the first thing she told me when i said ask me about yourself she was one of the she's one of the best analysts i ever worked with years old was because that was the first thing she told me when I said, ask me about yourself. She was one of the best analysts I ever worked with, and it was because she really embraced, like, I want to learn how to do this better. I want to learn how to think. She was really great at that. And she had super strong SQL skills, too. You just got to

work through it a lot of times. That sounded almost positive. I know. I think we're i think we're reversing today no my sense yeah my only other take on this though is i don't think this is unique to data like i remember talking to having this conversation with somebody about being successful in general like it were and it was you're going to have to spend an inordinate amount of time like around the things you're actually doing to be able to execute on what you're doing and in any organization whether it's politics whether it's like i need access to this thing

or i need this like you're going to spend a ton of time on that not just it or data marketing is the same way and and there's just a lot of like reps and discipline and like yeah you know relationship building all these things around that to be successful at the one thing and even like a time percentage breakdown like you yeah especially as you progress in an organization you spend even more time on the like auxiliary things and even less time on like the core like what needs to be done yeah i think the one thing you'll notice in that is that a lot of other professions where you they're like let's call them like high prestige or where like a lot of people would like to do them or there's a lot of money there's a lot of barriers to entry yeah and especially 10 years ago there weren't a ton of yeah that was the like you knew a certain amount of like

sequel skill or excel or just skills yeah you could get in there or i mean you know there's people that were just like i just love data which i always those people want to write off not hiring you i don't care about how passionate you say you are about this because this is not an emotional gig i love digging ditches right it's like yeah okay yeah good but like there just wasn't a lot of that there i think we've seen a lot of differences now just in the fact that like, you know, if you want to really work in like algorithms, let's say like,

you better go get a PhD now. There's no jumping into that. Yeah. Yeah, yeah. There's a, is it okay for me to insert a comment here? Yeah.

Is that allowed? I guess I actually make the rules. Yeah, you do. Wow. Yeah. Okay. You think you make the rules. Yeah, we're just guests

on this show. You're the hosts. I'm driving the car and asking you what exit I should get off at. Yeah. You think you may. We're just guests on this show. You're the hoax. I'm driving the car and asking you what exit I should get off at. Okay. One brief comment from the moderator. There's a really good paper for those listeners or for listeners who'd be interested. It's, I think, a college commencement speech called The Inner Ring. And it talks about how people try to.

Oh, that's a C.S. Lewis. Exactly. It's a great one. That's a great one. Yeah, it's really good. And basically the summary is there are people who try to get in the inner ring or chase a title with prestige or whatever

in any context of life, whether it's I want to date a size title or whatever. And the whole point of it is if you just focus on getting extremely good at a craft, you'll realize that you have entered an inner ring of craftspeople who, to your point, love the process, right? I just want to hone my skills in this craft. Okay, we have time for one more.

Since the culture AI one was brutally and swiftly disposed of, We're going to try to squeeze one more in. Excellent. Okay. The author of this post will remain anonymous. What makes a great data leader? This may be controversial, but most of the time I go off gut feeling when I determine if a data professional

is right for the job. Or if you're the cynical data guy, if they say, I just love... I'm just passionate about this data. Obviously, this gut feeling is produced by what the candidate and their references tell me, but I've been doing this job so long, it's a well-rehearsed process

in knowing who's right and who's wrong for a role. When I ask to speak to data professionals, I know that a good leader in data will always hold a mixture of characteristics likely to be a storyteller, strategic thinker, skills in cross-functional collaboration, people management, and be a risk taker. I think those are all true. And I look forward to the day when companies start hiring for those data leaders. That's probably when I'll be a data leader again. What do they... Wait, what did you say?

Wow, that's not what I expected. That's a cool data guy. I know. Okay, so they're not hiring for those. What are they hiring for? Technical skills. When was the last time you coded a neural net? That was literally a question I had

in a director's interview like three, four years ago. And I was explaining to the person how like well i haven't done that because i've been managing but i can talk to you about how i got this you know in a regulated industry we got this model into production and over here it's like yeah we're looking for someone who's coded more recently than you have even though that's not part of the job description? Well, it shouldn't be, but I think for a lot of people, they think of director as super individual contributor.

Yeah. Agreeable data guy? I think it's funny how it starts out. Where basically it's like, I use my intuition to make decisions rather than data so that's fascinating right for me i have other thoughts about that but they're not relevant to this right so we're gonna yeah but yeah okay beyond that i got a little hung up on that honestly yeah beyond that yeah like i understand that's why we picked the post yeah i mean i don't think people like hiring data leaders, period.

I think if you already have, it ends up being first data person in the door. Like, since nobody, like, the IT people don't really know how to manage data people. Like, the business doesn't either. So, they just keep, like, going down the train or up the ladder. And then they're going to self-promote to be whatever leader there is and when they hire somebody even if they're hiring a for director they're looking for like some sort of technical skills and to offload work because i mean if you are hiring like i i just you know like that's pretty much what happens like oh we're gonna like we want

to keep you like you know about the legacy stuff like much what happens. Like, oh, we're going to like, we want to keep you like, you know, about the legacy stuff. Like we'll promote you and hire in a new guy. I mean, that's typically what happens. I also think it's a bit of a, I mean, you know, John, if you think about what are the profiles of most people who move up to being like VPs and stuff. Yeah, we actually talked about this a lot. Yeah. They come from a specific background they've typically come more from the data engineering or architecture side of thing which is a little more of like it's farther away from the end product right yeah but it's also one where it's a very like deliver type thing and

the people who get promoted are the ones that were there like we need this yesterday and they say okay and they get it done right yep right And then as you go up the ladder, like then they're like, well, we want you to be a strategic partner. Well, it's like, well, you have not been rewarding a strategic partner this entire time. Even rewarding technical skills and project management.

And just get me to get it done fast. Right, yeah. Put out the hottest fire. Put out the hottest fire, get it done quickly. And so like, it's not like it's the person who's in that position's fault necessarily that's what you've been hiring for because it's also because you know you view it a little bit as like a junior service team member right like that's what you want them to be and so that's what you've been hiring for

those skills which i think is also a problem because those are those people are farthest away from kind of like the end of the data pipeline. When you then say, how are we going to use this in a strategic way? They're dealing with infrastructure. They haven't dealt with the actual, what the number is in this. Why would they

know what to do with that? I will say that I do think it depends on who is hiring. Is the CEO hiring? Is another executive VP? Is the CFO hiring? Is the CIO hiring? I think it depends. I would say if you have

an analytical CEO, which there's actually not a lot of those, that is a great hire for this to actually like value a data leader. Because it's their job. But if it's the CFO,

like probably pretty mixed bag. And then because it's their job but if it's the cfo like and like probably pretty mixed bag like and then if it's a technical like cto like pick on my old role like odds are and again it's a cto that came from a developer or ops background they're probably least likely to value like a yeah data leader type skills i think it's like as he as john said you know a lot of the first ones either were like a lot of the first data leaders when the first started becoming popular were either like academic PhDs, which were hired because of how smart they were on the topic, which we saw had issues early on with many of them,

not knowing how to run a budget and the team. Whoops. An academic. Is this a, what is this? Is this a tenured position burn or the other one just being like this was the person who had worked here the longest and they had been the person working in excel and that creates an archetype that yeah think about

it they think well what's the data leader oh well it's gary over there he's been here for 20 years he knows so much about the business which yeah is valuable yeah yeah and then when they go and they look for other people they're not looking for it and it's it'll probably evolve over time but right now it is we're not looking for those skills we're looking for your technical ability that's super nice and for all our listeners named gary yeah, sorry. We love you, Gary. Sorry, Gary. All right. Well, that is a wrap on this month's show. If you're out there and you find a LinkedIn post that you want the Cynical Data Guy to comment on, send it to us.

You can go to datasackshow.com, reach out to us there. Tag us on LinkedIn. Tag us on LinkedIn. I think we're on X too, if anyone uses X anymore. And Cynical Data Guy has a blog now, so just look that up. Go find him on Substack. Subscribe if you haven't. Great shows coming up for you, and we'll catch you next time. The Data Stack Show is brought to you by Rudderstack,

the warehouse-native customer data platform. Rudderstack is purpose-built to help data teams turn customer data into competitive advantage. Learn more at RudderStack.com.