Why Product Teams Should Own Data – Ep. 480

Do you fix data chaos by building a “Center of Excellence”… or by killing it? In Episode 480, Mike and Tommy react to a provocative article arguing that centralized data teams often become bottlenecks—and that product teams should own data outcomes directly.

Main Topic

• Kill Your Data Team: Why Product Teams Should Own Data — The article argues that “data centers of excellence” frequently turn into expensive silos: long request queues, insights arriving too late, and constant infrastructure work that never seems to land business value. The alternative: embed data professionals directly into product teams so they’re closer to context and can deliver outcomes faster.

Main Discussion: Centralized vs. Embedded Data Ownership

Why Central Teams Become Bottlenecks

Mike and Tommy unpack the failure mode most orgs recognize:

• A central team becomes a ticketing system
• Context gets lost between requestor and builder
• Delivery timing misses decision windows
• The team spends so long building platforms that it stops shipping answers

The Case for Product-Owned Data

Embedding data ownership inside product teams can improve:

• Speed — decisions don’t wait behind an intake queue
• Relevance — the team building the metric understands the product reality
• Accountability — outcomes aren’t “someone else’s dashboard”

The Tradeoff: Governance and Consistency

The hard part: decentralization can create new problems.

• Every team invents its own definitions
• Metrics drift (the same KPI means different things)
• Data quality becomes uneven

Their takeaway: if product teams own data, you still need shared semantic standards (dimensions, definitions, and governance patterns) so embedded teams don’t recreate chaos at higher velocity.

Looking Forward

This episode ties directly into the Fabric IQ / ontology conversation: AI and self-service scale only when semantic definitions are consistent. Whether you centralize or embed, the work doesn’t disappear—you’re choosing where ownership lives and how you coordinate meaning across teams.

Episode Transcript

Full verbatim transcript — click any timestamp to jump to that moment:

0:03 Down. Good morning and welcome back to the

0:34 Explicit Measures podcast with Tommy and Mike. Hello Tommy. Good morning. Good morning Mike. How you doing? I’m doing well. Thank you very much. Just clipping along here. for those who are listening along, this is a pre-recorded episode. So we’ve been traveling a little bit. I have places we got to be and I think this is around the Thanksgiving time frame. So I don’t actually know exactly what day this is going to be on. [laughter] Neither I do. I just know the episode number. So exactly the same. So it may be near a holiday, it may be not. So anyways, [laughter] this is what happens when you

1:08 Record five episodes in one week and it gets things start blurring together slightly. All right, that being said, we we’re not going to do any news just because it’s a recorded episode. We’ll jump right into our main topic today and our main topic today is why product teams should own data. This is the topic here today. So this will be interesting to see where we go with this one. , give me the the runup of this one. How do we get this topic and like I think that even getting the topic is interesting. Let’s go there. This was a mailbag submission, just the article by the real Mike. And whoever

1:42 The real Mike is, I’m I know that we know they they’ve listened. They’ve joined us on the chat. So, whoever the original Mike is, thank you very much. There was no context of the article. , but it’s it’s a great one here because you and I have talked for five years online and 10 years before that about centralized BI teams. Y and the literal thesis of the article is that centralized data teams are the problem, not a problem, but they are the

2:14 Problem, not the solution. we’re going to unpack that a little bit today and and talk about what does central BI do? is it really effective and what other approaches does this author recommend? just other note here as well this is an article that Tommy found. It was an article from Medium. So you have to have a subscription to see the whole article. The the link to the article the full article if you have a subscription to Medium is in the description below. So you can go read the article for yourself, see about your opinions, see how you feel about it as

2:46 Well. And then yeah, that’s just all the details there. a a fun mailbag. It’s a, , in the dark shot, just just a URL link to an article that we should discuss. very good topic here. All right, let’s let’s jump on in here and I guess let’s maybe talk about let’s just maybe we should define central BI. What does that look like? Because in this article, I guess there’s like a five minute, , summary or wrap-up here. The main key points of the article are centralized teams become the

3:20 Bottleneck. They deliver stuff but they deliver it too late and with misaligned objectives. when you embed data professionals directly in your product teams you get better results. real world examples and he gives a couple here like HubSpot Booking.com snap travel show that a decentralized or hybrid mode offers a superior result than a single centralized team. I think that’s a key point to what he’s describing here. And I this is something I think we’re going to have to unpack a little bit more. the

3:54 The the transition the transitioning effectively involves hybrid approaches, domain ownership, strong communications, needing communities of practice and defined central data functions. So there is some central data functions that happen. But you need to identify what those should be and what those should not be and let the building the the departments build what they want to want to build. yeah, this is another one that’s something interesting here. Tools like DBT database tools should add value but they should support your

4:28 Organizational model not dictate how your organizational model works. Okay, those are like the five like summary points of the article. Let me let’s start there. Toby, what is what are we talking about here when we talk about central teams? How do you define this in your mind first? So, we’ve always said that the sign of a healthy data culture is an actual established team in an organization. And usually when in our world with PowerBI, I noticed that in the entire article really never mentioned a managed self-service type of solution is mostly

5:00 Data science type solutions that they were working on. [snorts] But regardless, we’ve always talked about if we’re going to build reports, we’re gonna understand the data. It needs to go through a central team where we know who owns the data. We can process the content. We can process the request. We may have some more expert knowledge depending on which person you are. You may be the liaison with marketing or the le liaison with sales. But your support comes from your team. that’s other data professionals

5:34 Who distributes content to the organization. , ensuring that people are keeping the same style, theme, data quality. , rather than being left out on an ocean, so to speak, if you are the only person on sales who is the one building reports, which is usually where you also run into a lot of problems. I’m not sure I understood your definition of what central BI was based on that. You just said a lot of things. I’m not sure. I feel like I got like a bit of a word salad here. Like can you give me an example of like what do you feel like what is pure central BI look

6:07 Like to you? All right, I will try to say this a different way this time but no central BI is a single department or more or less of groups of individuals who are their experiences in the data stack. You’re you’re making me question myself, Mike. No, I’m just I’m just trying to get a clear like you said a lot of things. I agree with a lot of things but you’re saying distributing and reporting and like they are responsible for the the data flowing through the organization both from the data quality to also the actual content coming out.

6:39 Okay, that’s what I was looking for. Okay, that’s a lot clearer to me cuz I feel like when when I hear people say let me paint two pictures of two different organizations and one where I think we’re having like a heavy hand for central BI and another one where we’re trying to federate or have a hybrid approach. Okay. So in one area I look at this going central BI is the team where all the data goes. Any business unit that needs data they go to the central the central team and say I need data. When you have a report request whether it’s a pageentated report or a PowerBI

7:12 Report the business unit sends a request fills out a form sends a request to central IT or central BI says I need this. Then central BI is then required to understand the requirements go back find the data curate something and then give back a here’s the here’s the URL here’s the page here’s the report that you asked for and then you go back and forth and negotiate that right so I think of central BI in the way that’s talked about in this article which is it’s a single department where all analytics come from like that that

7:46 Is the extent of it and my opinion here is this only works so well because when you go down this very hard line of all central analytics come out of this one particular team any tools that you give to the business now means well the business is just going to pick up that data out of whatever system put it in Excel and do whatever they got to do right so you get like these shadow IT to use a term here organizations or people in the business that central IT takes too long to get something done shadow IT

8:18 Or individual in the in the business unit, take that data from central and then do their own little analysis of it. So that’s how I see the one company, right? Let’s flip that over to another company that’s saying, look, we know we need to have a distributed EI team. And so in this example, I think in organizations where Microsoft even touts this is a hybrid approach is really the best way to go about this one is central teams start with here’s the large amount of data coming from SAP the big

8:52 Data systems whatever those are we realize that there’s going to be other data you’re going to need to blend with our main central data so we’re going to have handoffs or exchanges depending on your team’s skill levels central team is a smaller group of people defining policy and governance around how data should be handled, educating people across the organization. The central BI team is responsible for running a community of practice that enables analytics to be taught to the business. Hey, here’s what a semantic model is. Hey, we’re going to do an experiment.

9:25 We’re going to give out tables of data using lakehouse shortcuts. Here’s how you get a lakehouse. Here’s how you build things in it. and then there’s a clear the the central BI team describes here’s the data that we are responsible for. We’re going to make sure that this table is up to date and correct and here’s some levels of how we check quality. Here’s quality checks to this data table and we will hand the table and here’s how we’re checking for quality back to the business and the business can take that as truth and build what they want from there. So I I think this article is

9:57 Arguing a lot more for hiring or educating a lot more analytical minds in the business units where they work and and that’s maybe my story here. So that’s how I paint this two picture. Would you agree Tommy? Is that what this article is painting or do you disagree? the definition itself to be honest I think that’s taking to the extreme because you mentioned that you said the the key word the buzzword the killer word you said that you’re giving it to it and that’s the misnomer here where especially if

10:32 You look at a lot of organizations today who have a centralized approach to PowerBI it really is not the IT department who owns them a [clears throat] lot of time 5050 some yes, some no. Let’s say 50/50. So fine, that means there’s 50% or other or like usually finance or technology. But the point being it’s not what we would assume just to be what it used to be. If you were to tell someone you worked in IT, it’s just because there’s no category for data.

11:05 All that being said, I think that’s the more extreme version of a centralized team where that is more isolated from the business. You can be on a centralized team and not be isolated. Yes. Not saying you are, but I think that I think the the goals of that team are different, right? So on one hand, the team looks like a ticketing system, right? The central team looks like whatever you need, fill out a ticket and their goal is to sit there and just chew

11:37 Through tickets. One of their metrics is a ticket processor, right? How many [snorts] tickets did we get through this month? How many report requests did we have? Did we get out? Do we are we producing enough at that central team? So, that’s how one view of this I look at. It’s a lot harder to put tickets around a community of practice. It’s a lot harder to put tickets around education towards analytical tooling, right? Those things are more difficult to quantify. And I think when we talk about central teams, we sometimes fall

12:08 Into the trap of if we if we focus on what we want to measure and what we’re measuring is we’re adding a service to these business units by getting through x number of tickets in a certain amount of time. That’s an easy measurement tool. But does it really help enable businesses to take responsibility and ownership of their own data and and building the skills that we want them to have so they can do things? Now my point around it is I’ve had some experiences and I think you have too Tommy as well when you work in this. The reason I say

12:40 It is because the systems that usually generate all the data to begin with are usually housed in it right those systems can’t go down your operational system. If you’re a larger organization, you’ve got manufacturing plants, things across the world that you’re producing things through, that team needs to support those systems to be up all the time and not fall down. So there is a paid group of people to make sure that these systems do not fall over in that system because they automatically have access to the databases and the tables

13:12 And the things that need to be producing the data. People assume or put the data role. I think this is maybe a bit of an older concept, but they put the data role like the data BI team. Okay. Well, you’re just part of you’re just part of it. The funding I’m I’m I’m physically talking about where does the funding come from, right? The data team is funded out of it budget just because that’s closely aligned to what they should be doing. Like, yeah, you’re going to have systems. Yeah, we we also need to get those data out to the business units. So, therefore, the spend

13:45 Comes from it’s budget. That’s why I think it aligns there. A lot of it is follow the money. Yeah. And it’s it’s funny you bring up the ticket system, too, because I I I would go a step further and say it doesn’t really serve the BI team either, just to only work on tickets from a from a life point of view. Yeah, there are requests. Again, the the role of a centralized BI team is to support the quality of the data, the distribution, and the trust of the content. It’s funny though too that you bring up the when not necessarily the opposite of a

14:18 Centralized team but you’re saying when I go out to the business I am building the support for the technology through the center of excellence through helping people understand the submitting model. Now I will argue that could absolutely happen with in a centralized area what I would consider being more hybrid with the with the organization. I don’t understand your comment. So, so you that you have the centralized team who’s doing all the data things and then you have people who are the champions or those stewards to help

14:50 Promote and preach the basically preach it who that they can still live on the in the centralized team. You didn’t say the opposite federated like where does it where does their paycheck come from? Right. So central there’s like a small I think this in this in this case there’s a smaller central team like there’s a a person who’s in charge of like the center of excellence. There’s a person in charge that that that team is in charge of making sure that they understand where data is coming from and they understand how data is being distributed out to the business units, but there’s other users that participate in the center of excellence, but they’re

15:24 Being their salaries are paid for by the business unit. So, it’s people who participate and can bring their voice to that central team. So, this is more of the federated approach. the federated approach is there’s multiple teams that need to work together to break down data silos, articulate their needs, explain where they’re having challenges or issues or problems like those teams are being brought into that central experience. So yes, it is a central team. It is being paid out of that central budget a CTO it’s coming out of the maybe it’s coming out of IT

15:56 Budget for the central team, right? But I think that shrinks this the number of people that you need there. And what happens is they they do a lot less of like I’m going to fill out a bunch of I’m going to chew through a bunch of tickets and there is probably some of this but there’s also the ability to say look that central team is really try to enable users to build stuff like central team says we’re going to make six semantic models. These are what we’ve decided that are the best way to share out data. Okay, here’s a handful of pre-built reports on top of these

16:28 Models. Here you go, business. And then in that business, as they become comfortable with it, you can they scale up, right? Then say, “Okay, now here’s access to the model. Build what you want.” So it’s interesting you’re going that approach because there’s a few ways when it when we talk about the where someone actually lives so to speak because I’ve also seen the extreme Mike where you had the technical user who’s the one supporting all the data in organiz in a particular department but

17:01 If you’re too far away from again where the data is coming from the data quality and if you like let’s say just supporting a department and you in that approach the ins and outs of not just helping people understand the data but what those definitions mean what a for if it’s sales what a good account is and a bad account because you live and breathe the business definitions in lingo so that’s one other the extreme side now that comes with

17:33 Its own cautionary tales Because if you don’t have support, if you’re getting paid for by the business, then any support that you need from BI, you’re just again, you’re not going to be treated that way with when it comes to access. So, you’re isolated. And a lot of people, this has been a very consistent thing we’ve heard from people is they feel overwhelmed because they’re supporting an entire department. There’s not enough resources. That being said, when we look at different ways to

18:05 Support the business, it’s more than just also to promoting the technology where it’s also the initiatives done. And I I like that you brought up the tickets idea. Tickets are good. Tickets are important. Like I am not going to ever stray away from the fact that you need a centralized place for people to put requests in. But if you actually really want to be impactful with a department that you have to understand what they look at on a day-to-day basis and it’s not just building or giving them access semantic model and

18:37 Upskilling them it may be initiative based what any of everything that BI should be doing anyways should be aligned with the objectives in organization whatever and it should be dynamic for every year that was the for me my eyeopening moment my first company where I was a data analyst. I was the individual supporting sales and marketing. We we had two other people who were the data analyst, but we did it all. We did it all for them, all the

19:09 Data, making sure it’s great. And it was completely overwhelming. And we also had no every request we did too because we also had no way to say no because there was every the person over me was sales who did not understand the technology. So, this goes back to a lot of these principles and I’m I’m going to jump between basically the article here and then also jump into the the PowerBI or actually the fabric adoption road map now because I think there’s a lot of really core concepts that that align to

19:42 What strategy your organization is putting down and based on that strategy it really informs okay this does make sense. So in the article it talks a little bit here about why centralized data teams become the problem and not the solution. And I think this is a really good point here the the telephone game problem or playing telephone right I have to fill out a ticket I have to articulate my requirements to somebody else they have to understand what I’m asking for and then from that understanding then build out something and come back to the result. And sometimes that’s a couple times removed, right? If I’m lower in my

20:15 Organization, I may have to go to my boss, say, “Here’s what we need.” Boss has to then fill out a ticket. Ticket has to get sent over to a team. That just starts slowing down the process there. It becomes more arduous to get the information done. I remember working on a project where central IT team had to own where the data was being stored for customers and customer was sending us 200,000 records of data a day or sorry a week. It was one week every 200,000 records and we didn’t have the tools or we were weren’t given the tools to build our own databases, our own

20:47 Systems to load things. There was no tooling that the business was able to have like we couldn’t go into Azure and run an Azure data factory and put the file down, load it into a table and edit a database and move on. That was all required from the central team. Well, the data format kept changing and so the requirements kept moving and so what a project that should have taken like about a month for central team to figure out and fix actually took a year and a half. And so the business wound up standing up their own solution inside

21:20 Excel with Power Query to ingest millions of records of data because central team couldn’t figure it out quick enough and eventually they got it because the format settled down. The the files that kept coming in started getting consistent. It wasn’t changing as much. Again, this stuff happens. It’s part of the process. But , was it acceptable to have a year and a half of time to Probably not. Is there better tools out there that could let it could have let the business unit use it unit do its thing? Yes, there is. So, this is the telephone game that occurs, right?

21:52 It I have to every time something changes, it’s going back more communication. It it it’s talking across teams and Rob Collie in his book Power Pivot was it Power Pivot, one of the Power Pivot ones, one of his early ones, Rob Collie talks about it’s much faster to communicate to yourself than it is to have to communicate with a second or third person in there. If you took all the code, like if you work on a software project and you just wrote out all the code that you needed to write to build a project, if you knew exactly what you were going to build and said just type it out, takes you, , a

22:25 Couple days to type out the the book or type out the code. And by contrast, it doesn’t take days to build an app. It takes months or years to build an app because you have to communicate the need to someone else. They have to interpret it. They got to build it. and then you had to come back and refine the idea. As soon as you involve more people into the process, things potentially get slower and it it takes more time to get that communicated. And I think that’s what he’s describing here in the telephone game, right? That one that one makes a lot of sense. And I think a lot of times

22:58 The other thing he points out here too which I think is a a trap as well is when we start talking about the modern data stack there there’s some there’s some and the tooling should be a part of your you have to pick something you can’t have no tool and get this stuff done [snorts] but this is where I go over to the adoption road map in the adoption road map it talks about like your maturity levels and it talks about what is the data culture what culture do we want to instill That that’s probably the first thing we should identify is what that looks like.

23:30 Then we identify what is our business alignment. What is going to make us be successful as a business? And then on the maturity levels they talk about there’s another image here that I think that works really well. There’s types of adoption right at the organization. We’re going to adopt something at the organizational level. We can do user level adoption and we can do solution level adoptioning. But I see this as also being the idea is you have to have buyin at the organization level. Yeah. We have to get buy in at the user level about what our data culture and analytics solutions need to be and only then once you have buyin at those two

24:03 Levels then you should be looking at like okay what technology solves this problem and I think for a lot of businesses and this is why I think we’re seeing an explosive growth of these call them entire data platforms right it’s data engineering it’s data science and it’s BI and analytics all in one platform all of them are doing it snowflake’s doing it data bricks is doing Microsoft’s got one. When you look at the Gartner charts and the the Gartner, what was the one that just recently came out, Tommy? the Forester Wave. Forester. Yeah.

24:34 Forester Wave just came out and said, “Here’s companies that are rebuilding and re reimagining like an entire fabric data platform that doesn’t just do one thing. It handles real time. It handles bulk loading or batching of processes. So now that you have business alignment on what do we think about analytics then you can go choose the tool. Sorry I said a lot of things to you Tommy. Yeah. So apparently we’re having a lot of salad today. So no I honestly this is my word salad. That was my word solid

25:06 For you. Back to you. we’re we’re just where we’re both going where I think we both probably recommend to where if you look at a single centralized team is not going to be the right approach because you are going to lead into you’re going to get closer to it but we also don’t want that isolation of just having someone who’s the analytics which I’ve seen time and time again a department just hires an their analytics expert who has no technology support at all. they’re just they know that they they’re good

25:39 At that, but they’re not going to get supported. We’re really I think that always seems to be the most flexible approach. I’m not going to say it’s perfect, but it is the hub and spoke and I really do believe more and more that and I’m seeing a lot of organizations begin to adopt that. There’s actually a link to HubSpot talking about this as well when they scaled it down their team even though they’re talking about DBT. Well, yeah. Yeah. I wanted to I wanted to point out that too. I was curious. He he points out three companies that killed their

26:12 Central data teams or removed them and what they built instead of those central data teams and what were your thing your your thought on that Tommy? And it is so it’s eerily similar at least from the diagrams and what they’re talking about to yes when PowerB’s recommendation around the hub and spoke. Well it should be it should be similar because that’s a good approach that’s great approach. So there’s a there’s a the technology of PowerBI is the enabler for you to do hub and spoke. Right. So Tommy,

26:46 Right? , yeah, we’ve talked about this a number of times on the podcast, which is we believe that analytics is now becoming a commodity because Microsoft is trying to simplify something technical and give it out in an easy way for people to click buttons and get to those develop, build, create things quickly using the tools that they’re providing. So by removing some of the technical aspect and hiding that behind a user interface that’s clickable. I don’t have to write as much code. It’s clickable interfaces. I can

27:19 See previews of data as I’m transforming it. Like those are core principles to helping democratize or share or make it easier more accessible for people to use it. And when you do that it enables the hub and spoke more. Yeah. And on listen, we’re also talking to about the actual end users becoming more creating like I think what we’re talking about at least from just the just the root of the hub and spoke is to me when you were trying to think about how does that actually work? I always go

27:51 Back to well who’s supporting you and really you may be working focused on a team department and yeah it’s funny you said it too. I’ve even seen some companies adopt more like a tiger team where the they have a team of people who just jump to whatever the greatest need is based on the business goals and what they’re trying to accomplish, but they are supported by the BI or the centralized data team. That is where they get support from. They they also own up and they’re accountable. Yeah. At the end of the day to the centralized

28:25 BI team. It doesn’t work if I’m just meeting with the BI team every week just to see what their latest updates are. The accountability here is where this becomes very efficient. Now I may be sent out like an ambassador to a department and talk to them 90% of the time but my support where who’s doing my and is and to be very practical who I’m doing my one-on ones with and who’s actually going to review my year annual review should come from that centralized data team for this to be

28:59 Successful. I I I like this idea of the tiger team and I worked in a company once where a very successful objective was identified by a CTO leader and he said something he goes that I thought [clears throat] was very fairly impactful. He goes, “If you give me two or three really smart, very sharp individuals, we can solve any problem. And what we need to do is we need to go find the hardest problem, the one that is the most difficult for us to solve,

29:31 And we need to put all those leaders at the top, put them all on that really difficult problem.” And he goes, “Anything else that is not that problem, we don’t solve.” Like those people are dedicated to solving this extremely hard thing. you focus, you have to shed all other, , asks from things and really let them, , the smaller tasks, you just got to let them be dealt with, but you need to hyperfocus on this one big problem. And I think to some degree that’s that focused attention to solve a very challenging thing. That’s a

30:05 Lot that we don’t I don’t think we have the some of the businesses don’t have the stomach to come in and do that thing and really revolutionize their business in peace. And so what with this company they were in the the shipping and transportation business and they were able to identify the right products at the right location. And so they had buil we there was a system that was being built to take in all the orders align them to where they need to be distributed and then they were actually able to find which locations which storage locations had too much inventory

30:39 And not enough inventory and how to rebalance their inventory. Just by them rebalancing their inventory and putting the right parts in the right location, predicting where they think the sales were going to come from, they were able to save a million dollars a month in just shipping costs cuz they were just shipping things from too far away cuz they had a they had a guarantee like they they’re playing with Amazon. They’re playing against Amazon, a two-day shipping thing like if you need to order it, you order it, it gets there quickly. So they’re spending all this extra money shipping things from really far away to get to these locations. Not effective. So just by figuring out this

31:12 Really hard difficult problem that wasn’t easy to solve, it made a huge impact to the business. And I think that tiger team effect can really have an impact on certain things. Now I I don’t know if it works in every single situation. There’s probably edge cases here, but I I like that idea. I personally have seen it work and it make a really big movement inside a business. And on in order for that to be successful, I think you you made some very good pre prerexs where you do have to have a pretty healthy and mature data culture at your organization

31:44 Because they have to be willing one first off for those people just to only work on those things and that there’s going to be those devoted resources to that and also you a healthy data culture. people are willing to show, , the worst processes and the worst data that they have and go through things because those are usually what a a tiger team would go on. But I’ll give you a diet version of this that can also be approachable if you don’t have an executive sponsor who’s going to make you basically jump from one problem to the next where when you think

32:18 About not just I am the analyst for a department where again I for me Mike it’s going to be hard to solve anything that way because you can’t you have no ability to say no to any request by the team again this goes back to the accountab ability, but if my accountability comes from the centralized BI team and my goal is to the highest level of impact with their data and what I can actually do is rather than hey just send a ticket for a new report and that’s what I’m going to

32:50 Do. That doesn’t do anything. Mhm. instead, one of the things that we actually adopted was we would do road shows once a quarter with like subsets of the each of the department and we went through just an exercise where we weren’t talking about the data or the reports. We’re talking about well what’s your initiative this year D and going through these things and trying to basically get some themes together on understanding one what are the goals of the company for that time period. to what would be one thing that they wish

33:23 They could have more than anything? And we identified so many oh man it was I wish I learned that trait a lot earlier but just asking that question that way and it’s not just no I’m just the guy who builds reports where no we just make your data work for you and there are people who who are have to do those mundane requests because you got a ticket system is important but my accountability came from my boss who was not in sales they She was the boss basically this CD

33:58 Officer of data and that’s where I had to report to that’s where I had to be accountable for now I had to show value for my time and this was the best way to do it and the nice thing was I was I could tell you every acronym abbreviation that that department used. I knew what they were, , what made a good leadup and I it was more than just the DAX statement. But having that knowledge and being almost like a mini consultant in a in a

34:30 Department can be very impactful and at least and again can you survive that way? Maybe not, but almost acting that way where we’re going to just, , uncover the biggest things. We were even we began to do stuff that wasn’t just PowerBI. we realized that everything they did was in a single Excel file that got passed around. , and we built the Power App. So, it was just a data solution at that point, too. So, all all that being said, there’s a lot of approaches here where we’re just not it, , because the internet

35:04 Went out or whatever, , because that’s when I when people think it support, they think I can’t send an email anymore on my phone. Can you help me thing? And if if that’s all we’re doing, then I don’t want to go over to my boss at the end of the year and say this is what I did. Yeah. I’m I think so. I want to keep doubling down on these things. I think in the article to to your your notes here, Tommy, around what you’re talking about the decentralizing and getting, , effective leaders of analytics in these departments, right?

35:37 How like to your point here, how do you do that? What does that look like? And I think these examples here in HubSpot, in Booking.com, and and looking at Snap Travel and Stash, there are some case studies that show that in the article here. And I really like how he called these out, right? So like, , HubSpot they dismantled their centralized business team. Instead, as the team scaled to 3,000 employees, the whole team, right? This is HubSpot in general. The result was they had developed 120 data analyzers, a team of people that

36:09 Were like more focused on data, , that that representative from central BI, but they were embedded to your point what you said earlier in the conversation, Tommy, was they’re embedded directly alongside business colleagues. They’re not just analysts serving the business and being a ticket pusher. Now they’re in the business. They’re learning the acronyms. They’re responding faster to queries, right? So they’re given elevated permissions against what data they’re able to get access to helping build reports. So it’s that embedded approach of okay again I’m going to assume here

36:43 Those analysts are dotted line to central IT meaning they have access to systems that not everyone has access to but they’re able to build data things and produce data elements for those business teams faster and then they can more easily understand the integrated business communication and what the needs were at the business level but their salaries are paid for from that business unit whatever ever that unit is doing. So I thought that was really interesting. Booking.com had a very similar effect. They saw they they were driven by the need of speed and relevancy. Their

37:19 Central team become had become a bottleneck and their data scientists lacked the context to give meaningful contributions to those different businesses. So Booking.com moved away from central data science and put data scientists in the business units where they were solving their problems with data science technology. I thought that was really interesting. , Snap Travel used an embedded growth team. Analysts were assigned inside the different departments. They had an agile central team. So, analysts came back to that central hub, that federated approach. , and they had better

37:51 Collaboration. And they had full stack pods. So, they merged their data engineers and their analyst into a single like combined unit. So, had someone who’s very technical could build the data engineering and then someone who could like take that data engineering and push it into reports or semantic models. , and then they had this like hybrid domain approach. , which I thought was really interesting. So again, a lot of I think what this feels like to me is a lot of these silos that we’ve been traditionally having need to be broken down. And the more we have tools that allow us to easily collaborate between data silos. And this is one of the

38:25 Things that I think is really nice here is fabric is fairly straightforward. You can still have security by different data silos, but you can also break those down and make it easy for people to get aggregations or move data across different teams using fabric. And one of the areas that I’m really pro about right now or really interested in is this idea of like shortcuts, door cuts across the lakeouses is really interesting to me. So one team can own a data silo or the data that’s in their department. they could aggregate information down to an appropriate level

38:57 And then shortcut that data to somebody else. So the original data team can own that information, but then they’re easily able to give it out to other teams as needed. And it’s a it’s a secured connection to that. One team can own the processing and creation of data. The second team can then own the consumption of that data. So this is really interesting to me here. And I I honestly think though too with fabric this is really going to enable us because before with PowerBI is like hey how I can solve your problem the same way I can solve every problem

39:30 Would you like a report I can make it a dashboard like right so if you were just a PowerBI pro your solutions what you can actually output was limited from a streamlined point of view. Yes. And that was the frustrating thing. We were like I remember it was I I remember complaining about this where I was a data analyst before PowerBI. Sure. And then I went to the biggest business intelligence tool out there and I became the report guy and I wanted to get out of that because but that again that’s the fruit of your labor. That’s really the only thing people could see. You couldn’t really do too much more with

40:03 That from the business point of view. That is completely changed with fabric. And yes, funny this is almost aligns more with when you’re talking about like the DPT is that modern data stack where yes your solutions are going to come in many shape and sizes right so this also allows us to not just to work on what because if people think you build reports they’re going to ask you to build reports for reports and I think there’s there’s also this lay I’m working in an organization right now that has there you can identify users that need just reports right there there is a there is

40:35 A skill level of your people in your organization that are just report level people or even I’ll reports are built and they’re pushed to an app like that’s all we get like that’s the extent of what that user needs then there’s another level of those when you start getting a little bit more technical and say there’s like scales to this okay so instead of giving me reports we have another group of the team that’s like okay they need to build their own reports but they need to understand what a semantic model is and do that so there’s like there’s like another more technical role that is that and then there’s even a further advanced or super technical role where you’re saying

41:07 Here’s here’s the lakehouse here’s the tables write your own SQL we trust you like what you’re doing so I think there’s there’s definitely levels or scales to this on like how you distribute that data out to the teams depending on their needs what their skill level is and I think what we’re going to in generally in general find because these tools are becoming more accessible to a broader audience you’re going to see the market slowly getting washed with more and more talent that is

41:40 Able to absorb more and more data and the expectation will be is I’m going to have better tooling that will help me analyze manipulate and work with actual larger piles of data that I can go push more to the business. the workforce I think is in general going to move more towards like these doing their own data because that’s the way the the techn is going. Well, absolutely. But I I you might have missed one of the points I said that was at least for me a very it was my biggest

42:13 Pain point where and I think you’re also going to see it grow as well because the what again what we can build is different too. When I was the PowerBI expert, the only thing people asked me to do was build reports because that’s the only thing they saw I could do. Even though there’s a lot of the things I wanted to do, the solutions architect. I wanted to help build empower apps. I was actually talking to my boss about that one day where was it was frustrating, but you’re like, “Oh, you’re the guy use who uses the tool that builds reports.”

42:45 So, they’re going to think in that box. But now with what we’re be able to see with fabric, we’re probably going to start seeing the business ask different questions of their data team than they did before because they’re going to see what solutions are or like what else is out there. They’re going to see other ways that they can in a sense interact with their data. Again, eight years and Mike, eight for eight years, I don’t agree with that. Oh, they don’t they’ll never they’ll I don’t agree with that statement. the business won’t naturally ask for more because they’re not going to go learn fabric to

43:19 Understand what it can do to ask for different things. Right? The this is this is what you’re describing to me is the role of the central business intelligence team. role of the central business intelligence team is saying we have looked at the landscape of fabric and think that these parts of fabric are effective to our business and then they centrally train and teach the business to say here’s where we’re going to do these things. this is what we’re going to build in order or we’re going to let you start doing and if you’re interested if this idea is going to add value to

43:53 You or if you want access to a semantic model to build your own stuff if you want access to write your own SQL against a data warehouse or a lakehouse come to our session we’ll teach you how it works and the rules of the road right so central team still needs to provide guardrails but the business is never going to ask for anything they don’t understand or don’t know how to do and no nor would I expect them to what’s going to change though is that solutions architect on what they can answer to the business because a lot of times my I agree that I do agree that

44:25 But that’s [clears throat] where the questions are going to change once that starts becoming consistent because if I had if I had a question like oh we need a we have emails and we have people is the site oh we need a way to be able to connect this stuff well before I was like I can build you a semantic model that does that you can’t touch it well now people are working in fabric. They have this whole arsenal of tools and they can start asking different questions too to be like do you need like are you sending this out to anyone? Basically like is there any automation? Oh, we can do this all in

44:58 Fabric. Actually there doesn’t even have to be the end goal of a report. Next thing that now becomes part of the knowledge part of the u memory of the business. Like remember that one thing you did with that? I think you called it a lakehouse and you were just sending it to my other my all my other applications. Could you do that for I have another question like that too because they they’re going to start seeing what’s possible. It’s a slow build. This is a slow role but the thing that is the only thing that’s going to be dynamically changing is

45:32 When we get a problem when we get a project we have so many other in a sense ways that we can manifest a solution outside of a report and this is also going to really dynamically change I think the types of questions we have so that might be diverting a little here but in order for I don’t think so in in order for any of this to actually work though again we’re making a lot of assumptions here that there’s a healthy data culture the accountability is right because you’re we’re doing a

46:05 Lot of things here so I want to back up a little there because how do you you asked the question earlier how do you get that working right you don’t just go in tell your boss I’m only going to work on the hardest things I’m going to just talk to everyone and they’re gonna I’m going to hear what’s aligned with the business so right Well, first and foremost, it starts with leadership, right? So, at the end, at the end of the day, leadership has to have this buyin and they have to be mindful enough to know where to most effectively put their people. And I’ve been in some

46:37 Organizations that had a great vision for this. I’ve been in other organizations that did not have a great vision for this. And so, you have to find what matches for you. And I’ve seen a number of people, Tommy, you’re one of them. I think I’m one of them as well. At some point your knowledge in PowerBI and fabric outgrows the capabilities of what the organization will let you do and you need to move on like they’re they just don’t get it. The leadership doesn’t understand the value of analytics and I think the analytics space is so hot right now that if if you feel like you have outgrown your organization like you’re only able to build PowerBI they

47:08 Won’t turn on fabric you’re interested in learning this other space. I’ve seen a number of people get really technically good at PowerBI and start to want to desire to do more in the fabric space. They just have to find the organization that wants them to do the fabric things. So all that aside, I I do want to come back. I think we need to wrap here. So my [clears throat] final thought here around is I liked in the article here, it actually talked about how to transition from a centralized location BI team to an embedded data team. He gives five approaches here that I think are really

47:42 Solid and I really liked this they’re the written out in the article. Right. So the first one is number one start with a hybrid approach. Don’t disband your entire team your central team overnight. Don’t don’t get rid of it all. Keep them on the core infrastructure. They all started with a hybrid model. They started with central and they started embedding individuals into the business from that central team and making them part of that embedded experience directly in those teams. So that was really really good. I really like this one. Assign clear domain

48:15 Ownership. All your data that comes in has domains. Who owns that data? And there’s a lot of these transition moments where I have data coming in. It’s from a specific area. The idea is who owns it and then how do you transition that domain ownership to the business unit? What is the level of comfort of where you transition? Is it at the report level? Is it at the app level? Is it at the semantic model level? Or is it at the lakehouse and table level? You can define that and it doesn’t have to be the same for every unit. But I think

48:46 Like there’s this assignment of domain ownership and when it changes hands from central team to business team, it’s just having the clear understanding of like here’s the transition that’s occurring. We will own up to this point. You will own beyond that point. I thought was really good. Create communities of practice. Tommy, we’ve talked about this at length across the podcast. We totally agree. A community of practice, a central of excellence, a center of excellence is essential for having these things work. Redefining the central data

49:19 Function. This is what we were talking about earlier, very early in the conversation, Tommy. We were talking about they should be worried about core infrastructure, data quality, training, enablement, cross team collaboration, how to govern, what are the standards of that? Those things aren’t ticketable items. This is exactly what I was talking about earlier. These are different roles that need to be held by someone in the organization. No individual business unit wants to handle all governance and standards. They don’t want to do that. They want to just use the data and move on, right? So, someone needs to be in charge of that. And I

49:51 Think that’s a better role for that central team. So, I do agree with this one as well. Again, the Microsoft adoption roadmap speaks highly of these areas and goes in way more depth around training and enablement data quality needs. All these are all addressed in the PowerBI and fabric adoption roadmap. And then the last part here is train and hire differently. I do agree with this one too. I think your teams if you feel like your central team is slow and not doing what it wants, I think you need to work with your business units and enable them to hire more technically skilled

50:25 People in analytics and then integrate those individuals directly with the central team. I think this is making a lot of sense. , , you need you need people who are going to be able to be comfortable with with ambiguity. They’re going to have to be strong communicators. Can they communicate well? They have to be really focused on business outcomes, delivering things, and they’ve got to be very technical diverse. I agree with all these things. Yeah. Anything you’d add to these, Tommy? Honestly, it feels like it’s right out of the implementation road map. The biggest one I would highlight though is the accountability. And one thing I would probably add is more or less build

50:58 Your mission statement on what is going to be what is the aim for a hybrid approach meant to achieve. So you can actually say what is and again the the values but more importantly what is our objective what is our our goal and that may change from year to year that probably is going to be a dynamic statement but I need to start with something that everyone can get behind. Agreed. The last section here he does some questions for your organization. I’m going to end on these questions because I think this is a really good

51:29 Like reflection point for those of you listening to the podcast. Reflect on these questions. I think they’re really good. One of the questions you should ask yourself as an organization is how many steps separate your data people from the product decisions that they inform? Count them up. What percentage of data requests actually get implemented into a real product change? If you feel like you can’t find that number, assume that it’s zero. Also really good. Like if you’re not getting data to drive action, that’s

52:02 A problem. How long does it take to get from a question to an actionable insight and this may vary by department? I think I think this is very department ccentric. I have examples of this but we’ll hold it up for another time. If you embed your analysts tomorrow, what are the organizations and he uses this word antibodies that would resist? If you put your analysts directly in the business units, what is the natural resistance that the business will have against that? What would that look like? And so identify those pain points and

52:35 Address those pain points and address those concerns directly by embedding the analyst directly into the team. And then lastly is ask yourself what is stopping you from trying a hybrid approach with just even a single team as a pilot. Try it once, right? Do it do it on one do it with one team. I’d almost even argue when the organization I was in, the team that I was in, the sales and marketing team, we had a pretty decent data culture and we were doing a lot more of this hybrid

53:07 Type approach and it was working really well for us. We were unblocked. We were building to we were adding value and we were able to use the customer data. It was when we had to go back to that central team and have them control every single thing. That’s when we got ourselves in trouble. So, I do think there’s value in this hybrid approach. And this is a great article. I thought this was a really good good mailbag, lots of good discussion. , with that, I’ll say thank you all so much. We appreciate you listening. We know you could spend your time anywhere. If you would, if you wouldn’t mind, please make sure you like and comment on this video. It helps us and the algorithm to let you

53:40 People know this was a good episode. Tommy, over to you. Where else can you find the podcast? You can find us on Apple, Spotify, or wherever you get your podcast. Make sure to subscribe and leave a rating. It helps us out a ton. And guess what? This was a mailbag. And if you want your link, article, question, please try to add some context. That’ll help us out though. , just go to head over to powerbi.tips/mpodcast. Leave your name and a great question. And finally, join us live every Tuesday and Thursday, 7:30 a.m. Central on all PowerBI.tips social media channels. Thank you all so much, and we’ll see you

54:13 Next time. Are you down?

Thank You

Want to catch us live? Join every Tuesday and Thursday at 7:30 AM Central on YouTube and LinkedIn.

Got a question? Head to powerbi.tips/empodcast and submit your topic ideas.

Listen on Spotify, Apple Podcasts, or wherever you get your podcasts.