Sean Collins: 0:00

Hey, this is Sean Collins.

Ibby Syed: 0:01

I'm A. B. Syed.

Sean Collins: 0:02

This is, Numbers Narrative. Today we've got Selena Wong, the CEO of Dataculture. We're going to talk about everything from embedding teams, building trust, and making sure you do the work before you start doing data analysis.

Ibby Syed: 0:16

Yeah, it was an awesome, awesome conversation. I think you guys are going to really, really enjoy it.

Sean Collins: 0:20

Selena, if you were to summarize our conversation in you know, like a couple, couple sentences, what did we talk about today?

Celina Wong: 0:27

I think we talked about trust. So number one is building trust between the business and the data team. I think there's also an element of us disagreeing on whether we have centralized or embedded data teams. And I think that we also, you know, the third takeaway there is. As data people or technical people to not forget what business impact you're driving, because if you can't, if you can't, you know, start with that before you dig into data, I promise you, you're going to not be satisfied.

Ibby Syed: 1:03

And again, check Selena out at datacult. com. Enjoy the episode.

1:08

JUMPSCARE

Ibby Syed: 1:13

So Selena, thank you so much for coming on today. You want to go ahead and like briefly introduce what you're up to, what you were doing before data culture. Give us, give us, give us the spiel.

Celina Wong: 1:25

Absolutely. First of all, thanks for having me on here. I'm excited about what we're about to talk about. I'm Selena, CEO of Data Culture, which is a data consultancy based in New York City. We offer both data engineering, services as well as data visualization. And if you don't think about data visualization, it's It's the, when you see the New York times with their funny animations or Spotify's, you know, year end wrap, we do that type of animation. So I think that, you know, data engineering alone doesn't tell the end to end data story. Once you pull data, what does it mean? How does it tell a narrative? How are you telling the end user what to think about this data set? So that's why I think it's beautiful that we have both data engineering as well as data visualization as a part of our services. Before my time at Data Culture, I was actually head of data at ULA Skincare. And so for the listeners out there who don't know ULA Skincare, it's a really fun probiotics skincare company. That was actually acquired by Proctor and Gamble in early 2022. I was about to say last year, but realized we're now in 24. And so one of the really cool things that, is just a highlight of my career was when Proctor and Gamble were. They were looking at Tula and saw our data tech stack and our strategy and called out how impressed they were, which for any data person out there is just glorious. The fact that business leaders are recognizing the data work you do. Tula was actually my third startup. So I've been head of data at two other startups.

Ibby Syed: 2:55

Two things to come out of that. One, do you mind explaining data engineering a little bit? Absolutely. Yeah. I mean, I think from, from the very start of it, think about how, if you're a Shopify shop out there and you've got GA4 and you're using some sort of, you know, ESP for mailing or SMS, I bet that your data is Actually sitting in those different sources in silo. And everyone starts from somewhere where you're asked for reporting. Suddenly you're manually downloading things and doing all those VLOOKUPs and pivots over and over again. How many times have you banged your head against the table? Like I wish someone automate this. Well, data engineering is essentially that, right? Like how do you automate the call for that data without having to manually go in and download as well as create that VLOOKUP logic. So that an end user can push a button and get the report they need versus spending five hours pulling data. So that's at the very basic level of data engineering. I can go into much more that, you know, I might lose everyone and you might fall asleep on me. I

Sean Collins: 4:00

have so many questions. I'm trying to figure out which one I want to start with first. I think I'm going to go with the data visualization side. I'm fascinated so when you're doing these visualization projects, what is sort of the workflow and process for, for this designer, essentially, to, to figure out what is the right, right way to display this? What are the insights we're really trying to take away from the data? is that how collaborative is that with the client? Is it a team of people who are working through? So what does that look like?

Celina Wong: 4:31

Yeah, that it is such a wide range of people that we're working with. As you can imagine, sometimes you're working with someone who's the head of creative design, so they don't know anything about the data, but they can give you a vision for what they want to see out of it. All the way down to people who are technical, right? They've built visualizations, but perhaps they don't have more of that, I call it, creative juice of how do you make your visuals, pop to the end user. So everything starts with, phase one, which is scoping. And a lot of the scoping has to do with what does the business want the user or consumer or data consumer to take away from this, right? And I think starting with that, what is your vision and what is your want from this is harder than anyone realizes because when you're working with a creative person, and if you've worked with people like that, they love to just throw things on the board, right? It's just like whatever sticks. And our job is to hone in on all these different ideas you gave us. And marry that with the data set that you provide us to come back to the drawing board and say, this is what could be viable. Is that acceptable? Is this the route that you were thinking about? Perhaps this is a new idea that you didn't think about. And that's the beauty of bringing in someone like data culture or to just have someone who understands how to make visualization stand out to the data consumer.

Sean Collins: 5:56

That's really cool. It reminds me, you know, maybe not actually talking this morning about dashboards and reports and how, if you put a number in front of. a business stakeholder, how like they will become kind of addicted to that number, right? Like, I think the line I said was like, data people are drug dealers. Because like, you know, if I show you a traffic number or something like that, like If I put that on the graph or I put that on the dashboard, like they aren't going to hear the words I'm saying, they're going to focus on these numbers, and they'll become obsessed with optimizing towards these numbers. I think there's as much value in what you don't display as what you, what you do display. Yeah,

Celina Wong: 6:33

Yeah, I number one, the drug dealer analogy is hilarious. I'm like, has someone been watching? Breaking Bad or something. But the addiction is real, I will say, and it begs the question of making sure there's data quality before you show it, because somebody will be addicted to that number, like you said. Whether it's right or wrong, suddenly they're addicted to that's the number they saw. And the second piece is, less is more. I think even early on in my career, and I see both of you nodding, that everyone just wants to give you everything, right? Like, oh my god, look at all these insights. I actually want to just give you everything because I had to dig into everything. But it's really about, it goes back to that. That like scoping, it goes back to defining your problem and goal before you even do anything, or else the next thing you know, you're left with five different charts that you're putting in front of some business stakeholder executive and, you know, comparing it back to like universal analytics or GA4. What are you trying to give someone that they can hone in on? Because I've watched someone look at a dashboard and when there are too many tiles. I don't know if you've seen this, but I've seen the eyes just glaze over.

Sean Collins: 7:43

You see them pull out the cell phone and start like texting to you, like, okay, I already lost you now.

Celina Wong: 7:48

Oh yeah, or if you see, especially nowadays that we're in hybrid remote situations, if you see their eyes just shifting, I'm like, are you reading a slack and not paying attention to what we're telling you?

Ibby Syed: 8:00

You see like the face color change where it's like very clear that they've like moved from the video to like a white screen and it's like, oh, they're like looking at their email now.

Celina Wong: 8:08

Exactly. So that's why I think less is more, right? Because One, you don't want to lose their attention to, I think that the, the visualization as well as if you're presenting the analysis back to them, it is how tangible or how many analogies you're giving the end user that make it more accessible to them or more approachable to them. Right. Like if I was talking to you about your numbers and you're like, tell me about the number of users we've acquired on a specific product. And that's the question you started with. Why would I be providing you other charts if you didn't ask for it yet? So sometimes what I've done to train, you know, more junior folks is when you think about presentation visualizations or in a live presentation, think of it as storytelling, right? Think of it as a narrative. You have chapter one, which is. What did they ask you for? And this is the answer to it. And then chapter two is, guess what, Sean's probably going to ask you a follow up question. Let him engage and ask the follow up question to pull up something else. Or if you think about the visualization, place it in a different, or like have a break in your dashboard. And place it below the fold, if you're thinking about opening up, you know, a dashboard on the desktop, you want people to engage and feel like speaking of like, you know, episodes or TV shows, think about the best TV shows out there, it leaves you on a cliffhanger, and you have more questions, and you're going to keep digging in and ask for more

Ibby Syed: 9:40

yeah, that's super, super interesting. How do you think about, you mentioned something super interesting there, which was, training junior folks. Because someone comes in to a particular disadvantage if they don't know very much about the business, right? And as a consultancy I am sure that you guys run into tons and tons of different businesses and you have to sort of figure out both what's going to matter to them from a KPI and, and, and business standpoint to where you can actually surface the right data How do you, how do you think about that interplay between, executives or leadership and the data team,

Celina Wong: 10:15

'yeah, I I think that it depends on your stakeholder. So if we're working directly with the leadership team versus if you're working with a data team, I always teach my team, whether it's internal or at the consultancy, do you understand how the business operates? Do you know how the P& L works for this business, right? Because every business operates in their industry. They all have to make money. They all have to spend money, right? And I think that one of the traps or, you know, black holes that data folks fall into is I'm just going to get access to all the data and start building tables because we need a dashboard, right? And that's my data product, air quotes. But I start from, and maybe this is the finance background of me, right, which is challenging people to think about, do you understand how the business makes money and how it spends money? And then what are the goals? Because sometimes when you're, you know, I think Sean, you're probably Very familiar with this, but if you're in growth phase and you don't care as much about reeling in every penny, then it's the goals are all around growing revenue. And then, well, what's the data question being asked and how does that tie back to the business goal? Is what I challenge people to think about because. It goes back to the idea of data teams are seen often as a cost center. We're like one plus, if not two degrees away from actually being the ones executing on driving revenue, which is why it's so hard at the end of the year. I'm sure people out there are sitting there going. Oh, I'm looking for a promotion. I'm looking to measure, you know, what I've done and it feels so hard because you assisted somebody, right? But if you started your analysis with, I'm helping the business drive 15 percent growth in revenue or in product adoption, how did your data analysis drive that difference? So I think to get back to the full question, it's, it depends on who you're talking to, and then it also, from that, it depends on what are their initiatives. And I say depends who you're talking to, because sometimes when you talk to the data team of a project, they're not even sure about the business's initiatives.

Sean Collins: 12:33

Yeah, that's, that's really cool. I mean, I think one of the biggest challenges I've experienced as being someone who's like sort of data literate has been data should really be a very collaborative process, right? Like, it can't just be like, hey, hey, Ibi, here's a ton of data, find me three things I should do. I think a lot of people are intimidated because we start throwing up Greek letters or we start throwing up explanations of things that are just like way beyond any mathematical course that I've taken. So how do you, how do you build that, that team empathy and collaborative nature? How do you build that culture that understands that this is a back and forth, this is a team effort and not one where it's like, okay. If he's going to go do data stuff, he's going to come back to me with a product, a product, I'm going to go send some emails and we're going to be good to go. Yeah.

Celina Wong: 13:18

Yeah. I, I think a lot about building data culture because the amount of times that someone says their company is data driven. And then you look under the hood or you peel back the onion and everyone realizes, wait, how are you data driven?

Ibby Syed: 13:32

they've got an Excel sheet that gets updated every day.

Celina Wong: 13:41

Yeah. And maybe that's how they define data driven at this point. Right. But in order to be, you know, to even catch up to average or be at the forefront, that's not the definition, but I, I think of it as the what and the how. When you think about, you know, how to drive data culture or making that sort of collaborative effort on a project where you have business context, but then IBBY has the data, right, I think the, what is pushing people like IBBY where if they're working on something, make sure that there's from a leadership perspective, make sure there are consistent check ins, right. When I say leadership, this is the data team's leader. This could be from the growth side, right? Being a business partner where you're reeling in that analysis and keeping that focus on. What was the question we're trying to answer first, all the other things are nice to have. It's like, what is the must have delivered by when that's what you do. And a little bit of how I think a big part of the how, and what I've seen success in is make data approachable. And I mean that in a way of. Approaching people with empathy. Because sometimes what I hear, unfortunately, is, you know, the data team or tech team just saying the business doesn't get it. We have so much work to do. They don't understand the amount of cleanup I have to do. I can't believe they don't appreciate what I do. On the flip side, the business is sitting there going, this is taking so long. I have no idea what they're doing. I just need a number. Right. And how many times have you felt that friction if you didn't have a good relationship with each other? So I think I, when I think about empathy and approachability, what I really mean is, have you built a relationship with your stakeholder from a data perspective and vice versa? I challenge the business to build a relationship with the data team, right? Like, don't treat them just as your data monkey, right? Like, treat them as your true partner. What do you think about this initiative? And they might bring more insights than you realize

Ibby Syed: 15:46

Yeah, that's, that's, I think, a lot of what you talked about, I've seen in old jobs as well And so one thing that I also wanted to bring up is, is actionability. I feel like one of the places that the data world sort of falls apart is we're still at this stage where We talk so much about, Hey, how do we like analyze this? How do we get our, how do we get our numbers in the right way? And, and, and one of the things, one of the places where we fall apart is really that actionability. Once we know what we have to know, how do we use it to, to, to drive action?

Celina Wong: 16:21

Yeah. I, I think this is the part where, when we talk about data storytelling, but also what connecting that back to what's the business looking for and their initiatives is critical because this is where. And tell me if you both disagree, but this is where I see the biggest gap right now in the data industry, which is we spent a lot of time on building up data tooling, right? And I'm sure even the business side is like, what are all these data tools I'm supposed to know about? You govern something, you have a intelligence tool, you have some data warehouse. ingestion, things are happening. But I think it'd be this goes back to our coffee chat where I strongly believe that the next phase is software is over here, you know, for data tooling out here and people think. That the folks who build it think this is great. Look at our product and end user. Like Sean can just jump in here and find what he needs. And I think there's that gap of, of like, what is this data trying to tell me? And then, but how does this then marry with my business initiative? And by the way, people like, like Sean have 10 other things that are on fire. I remember using one of these BI tools and I gave feedback about something as simple as date filter. And I'm like, why isn't this just the way that a human thinks you're thinking like a software engineer and the feedback I got was, oh, that's because it was built by a software engineer. I'm like, okay, well, guess what? Your users are not software engineers. So create, you know, change the interface to make it user friendly for a CEO or a VP of growth or whoever's trying to use your tool.

Sean Collins: 18:05

Yeah, I mean, I think a lot of times I think you're absolutely right. I think a lot of times the the data experts are only in this academic world where they want. So, so to be so precise with their recommendations. And so, like, this model is perfect. You don't understand. I've spent. Months refining it and me in a very ugly world of digital marketing, where so much is hard, impossible to measure, like truly impossible. I care about how fast can you give me something actionable, right? That speed to insight. And so I think it's on one hand, it's on me to go to the data team and say, listen, I need at least something directional in the next. 36 hours, and then we can refine from there, but like, give me, give me something I can move in a direction because I'm not going to not do marketing while you refine your model, so we can either waste money with with no data insights, or we can have something and then refine it over time. I'm going to make this easy for you. You're in every slide. You have to answer. OAR, Observation Analysis Recommendation. And like, it has to be able to fit on a single slide. And if you can't get to that bite size, like you can't present to a client.

Celina Wong: 19:20

Right. I, by the way, I'm going to steal your framework over there which is also banned. Yeah, excellent band. But I think framework, right? When you think about actionability and giving junior people, or even I would say, I would argue senior people who are not used to looking at data like this, right? Like what framework do they follow? Oh, OAR is a perfect example of how do you hone in on the concept of less is more. And then on the other hand, when I think about actionability, time to insight when is it the right time to just give directionally correct versus absolute, like exactly, you know, refined, is Up front, like when you're talking about the problem you're trying to solve for one of the things I did quite often, and it didn't matter what level you were, what title you had, was when you asked a question, the data team has to ask why, what are you trying to answer, because when you tell us that context, then we are able to assess and ask you, right, like if we're not sure, just ask the person asking, right, can this be directionally correct, or does it have to be directioned this way. Exact because there are consequences for certain situations like in financial reporting where you've got to get it right and then there is a lot of the growth marketing world. I would imagine you're like, I just need something that's. You know, even if you tell me this is a 10 percent variance, I'm, that's excellent to you, right? You're like 10%. That's amazing. Meanwhile, someone could have gone off and said, that's awful, right? Like I'm going to go spend another two weeks of refining my model to give you 2 percent that, that two weeks is a waste of your time. And it's also a waste of time to insight because you could have gone off with that segment ran an experiment.

Sean Collins: 21:03

Yeah, that makes, makes a lot of sense.

Ibby Syed: 21:06

This is all really good. Should we try to hit one more?

Sean Collins: 21:09

Yeah, what else, what else is interesting? We haven't disagreed about anything yet. That's kind of frustrating.

Celina Wong: 21:15

Yeah, okay, let's, let's try to disagree on something. No, I want to, like, how do we make this spicy? Oh, I know. I think I read something, Sean, that you said. About embedding data analysts versus, yeah, central data team.

Sean Collins: 21:32

So we, we, we talk about getting, getting you know, the, the data team closer to, to the product. Right. And so, so they have the ability to, to understand the business problems in a way that if, if you used to have a centralized data team that, that, you know, I submit a ticket to in JIRA or in linear or whatever, you receive it assigned to some, some analyst or data scientist, and they come to you with, insights. I feel like there's going to be always gonna be some level of that, you know, that collaboration we talked about earlier. is intensified if every time I have to explain to you the marketing framework, so there are marketing strategy or, or whatever, and you're, you don't really understand that. So is that a problem we can solve by just either building career tracks for people or having. Yeah.

Celina Wong: 22:17

Yeah. I think that, you know, I think Sean, this is some of the stuff that we were chatting about like before this recording, but the concept of having embedded data team members versus a centralized data team. I disagree with the viewpoint of having embedded data team members, but hang on, let me explain myself because I know that it always ends up with, but depends, right? I think that it depends on the size of the company, to be honest with you. Right? Like what I've seen work well is that when your company is. Let's say the, the marker that I've mostly seen is around the 200 to 500 people mark. I think that data teams should be centralized and people can be assigned to a team, right? Like, Hey, AB is assigned to marketing analytics. Myself is assigned to finance, you know, and supporting revenue analytics but not having them totally embedded because I think that they would, yes, they would gain a lot of business context, but I think that they become very redundant in some of the data work and modeling that gets done. Because some of the modeling work behind the scenes and actually starts out centralized and then gets broken out. Right. So if you're sitting at embedded in the team, you miss out on some of the nuances of, of like, well, so and so was doing it for marketing analytics and that feeds what finance needs.

Sean Collins: 23:43

So you're saying that the, the reason not to embed is because you will then have, I guess, multiple models that are kind of answering different versions of the same question is you'll have a no source of truth kind of for a better word, and it'll be like, okay, well, the marketing team says our LT is this and the finance team says it's this and here they're cutting it different ways. And then you, then you spend your time fighting over which, which model is the right model. Okay. Okay.

Celina Wong: 24:11

Exactly. And you said you don't, you're not a data person, but you just sound like, sounded like a data person there.

Ibby Syed: 24:16

That's exactly what the problem is. So you end up running, running into problems like that where, you know, another good example is customer experience data. So a CX person is most likely looking at customer experience data to you know, analyze you know, what's wrong with a specific person's order or what's wrong with a specific person's you know experience with, with the platform versus someone on the product side is probably looking at the same data, but try to understand what improvements they should make to the platform, right? And they're trying to answer similar questions, but if you have those folks as embedded analysts, more likely than not, they're going to end up coming to, they're going to end up creating, you know, very similar, but not altogether the same versions of the same data for their own use cases. And suddenly, like when you start to answer larger level business questions. You know different teams will disagree and so you know more conventionally you've got folks in a centralized data team They create those models in a unified way, so it's like here. Here's my source of truth on Here's my source of truth on subscriptions. Here's my source of truth on, I don't know, email. And you know, we'll go to these external teams and answer their questions using, you know, the approved single source of truth, quote unquote, which, you know, we could go into why that itself is kind of a lie, but

Sean Collins: 25:36

the other problem that you, you identified there is you know, I think we all have this dream of like. Data for everyone, like data democratization and getting it so everyone can self service and I think you open yourself up to a lot of risk when you start trying to make everything accessible to everyone to, and then, then, you know, people don't know to clear filters or to add filters or what, what, do you have a data dictionary that's actually maintained it so people know exactly what events equals water, what attribute equals, like, what is the definition of this? And that's, I mean, you know, I guess, it's, Maybe that's part of the solution, I guess, for what I'm saying of why we need, should have an embedded analyst is like, can we build a self service data platform that gives someone directional numbers and insights and remind and train and educate people that this is not like a 100 percent solution. This is your 70%. You can make a quick decision, but not an expensive decision off of this.

Celina Wong: 26:36

Yeah, I think the keyword you said there is training and consistent training. Because how does someone know that they can make a directional decision without it being a expensive decision? And I think that comes with experience as well as training and communication, right? Like, that's outside of, outside of the technical stuff, it's the communication piece of it. Sean, I have a question for you. Like when, when data, you know, people come to you and say, I've got a self service BI solution for your self service, whatever, what's the first thought, that comes to your mind? Shit, another tool. They're going to try and make me learn that we're going to stop using in three months. I think that's so important for people to hear because I think from a technical side of the world, people probably think we're automating everything away and people like you are going to love it because you're not going to have to wait for a data person. Having that perspective is, is I would call it magic because there's still a trust implementation viability element to all of this, I

Sean Collins: 27:45

think, you know, that might be actually the perfect cherry on top of that. We haven't had a word. We haven't said yet is trust. Because I think the problem with self service quote unquote analysis I think a big problem that comes up a lot is that people get a dashboard and think that by looking at a dashboard you're doing analysis and, and like, or you hire someone super junior and say, give them the title and a list and think that everything they say is going to be like fully baked, well thought out, ready to be like shipped and make business decisions. And like. The thing we keep talking about is like speed to insight, speed to insight, but like just seeing a number doesn't actually tell me what to do, right? Like you could tell me that a click through rate is dropping okay. So does that mean I should, what do I do about that? Does that mean I should change the creative? Does that mean, but is the quality getting better? And so we're having less clicks, but, but higher quality, like I think that too often we get tied to a number and then. I think, think that having a number will make that obvious on what to do next. And the most important piece that can exist even, even more than like empathy is, is, is trust. If I, if I can know, and one of the reasons, you know, I love working with Ibi and the team at Cotera is like, when, when they give me an insight, it isn't just a number, it is like a, Hey, here's the approach we took. Here's how we try to solve this problem. Here's what we think it means for you and what you can do with it. And I can trust that, like, they thought through a bunch of contingencies. They tried different models for it. They listened to what I asked for and came to me with, you know, this is the recommendation based off of the all the things you said. And I think that that's That's probably what's missing from most you know, quote, unquote, from like the quote, unquote, data people to the non data people is it's you are putting a lot of trust if you don't know how to do data analysis, if you don't know how to code or how to build any of these models, it's a black box and you are putting your entire business on the line based on what, you know, the, the data culture team says, or what the Koterra team says, or what Selena says, like, My business, I'm putting my millions in my career and my, my, my, my own life, happiness and satisfaction and safety on the line based on the guidance you're giving.

Celina Wong: 30:00

Yeah, I think that's just spot on of trust and that. You've thought through the questions that I would have asked, right? Like I trust what you're giving me. And on top of that, it's not a number. Like you said, it's what do I do after that? And how does that also, I think a part of the trust and thought process is have you thought like a business leader would think, right? Even though I'm presenting something to you, I, you know, I think the last example I'll give here is we ran an AB test. Right? Yeah. At Tula, where we found that the conversion rate lift was incredible for short term metrics detrimental to LTV, which I know Cotera you know, services quite a bit. And I mean, it'd be cover your ears if this just is going to scare you, but the, when the test said that our short term metric for, you know, conversion rate lift in that time period was great. We ended up making drastic changes and plummeting some of our LTV. And I know heart attack, right? Cause then what happened next was my team jumped in and well, somebody junior was running the analysis and then the rest of my team, including myself, jumped in and started asking questions about what does this do for the longevity of this segment? What are we doing to ourselves? We're not gonna, you know, with some of the changes we made, we're not going to be able to reach out to them if they didn't convert. Like, yes, the people who converted now, we have them, but the people we lost down the road that's way more expensive. We didn't, you know, now you have to double down on CAC. Like, I think no one says better LTV, or, you know, I don't want better LTV. But sometimes you have to make a choice. You have to, you know, do what you gotta do for short term gains. Yeah, those are hard choices. I also think that we're running up on time, and I think that's a great place to end. Selena, thank you so, so much for coming on today. Again, everyone check out data culture. You guys have a website did a call data culture. com data colt. com they called. com. All right. Well, Hey, thank you so much. Thanks for having me.