Untangling Workspace, Branching, and Artifact Chaos – Ep. 478

Workspace design in Fabric gets messy fast. In this episode, Mike and Tommy break down the challenges of managing workspaces when you add branching, git integration, deployment pipelines, and a growing collection of artifacts. They share what’s working, what’s not, and how to think about workspace architecture before it becomes unmanageable.

Main Discussion: Workspace Design Patterns

The Problem: Workspace Sprawl

As organizations adopt Fabric, workspaces multiply:

• Dev/Test/Prod per project
• Feature branches creating temporary workspaces
• Mixed artifacts (lakehouses, semantic models, reports) in single workspaces
• No clear naming conventions or lifecycle management

Workspace Architecture Patterns

Mike and Tommy discuss several approaches:

By environment (Dev → Test → Prod) — The classic deployment pipeline pattern. Works well for smaller teams but gets complex when multiple projects share infrastructure.

By domain — Group artifacts by business domain (Sales, Finance, Operations). Cleaner ownership but requires discipline around cross-domain dependencies.

By artifact type — Separate lakehouses from semantic models from reports. Enables different security and lifecycle management per type, but increases the number of workspaces.

Branching and Git Integration

Git integration adds another dimension:

• Each branch can map to a workspace
• Feature branches create temporary workspaces that need cleanup
• Merge conflicts in workspace-level settings are painful
• The workspace-to-branch relationship isn’t always 1:1

Anti-Patterns to Avoid

• Everything in one workspace — Becomes ungovernable as it grows
• Too many workspaces — Creates management overhead and confusion
• No naming convention — Makes discovery impossible at scale
• Ignoring lifecycle — Temporary workspaces that never get cleaned up

Practical Recommendations

• Start with a clear naming convention before creating anything
• Document the workspace-to-project-to-branch mapping
• Use deployment pipelines for environment promotion
• Plan for cleanup: who deletes feature-branch workspaces?
• Separate compute-heavy items (lakehouses) from consumption items (reports) when scale demands it

Looking Forward

Workspace management is one of those “unsexy but critical” topics that determines whether a Fabric deployment scales or collapses under its own weight. As git integration matures and more teams adopt Fabric, having a deliberate workspace strategy becomes non-negotiable.

Episode Transcript

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

0:03 Thank you. Out. Morning and welcome back to the explicit

0:35 Measures podcast with Tommy and Mike. Hello everyone and good morning to you. Good morning Mike. It’s another recording. We keep going. No matter where we are or how much we’re traveling, we’re always going to be here. Well, Tommy, I think we have a bit of a a a reputation to uphold, I guess, is what I would say. We we’ve been doing this since 2021, four years now, and we’re almost about ready to hit 5 years when we hit 2026. Is that right? I think so. We’ve been doing this for a long time.

1:07 When was the first episode? Do we know when the first episode was? May of 2021. Yeah, May of 2021. Wow, that’s a long time ago. Anyways, so we can’t miss a week. We got to have at least some episode two episodes a week since 2021. That’s impressive. , I’m excited to see how this is going to go. Who knows, Tommy? Maybe it’ll be 2031 and we’ll still be doing episodes. [laughter] I I , I’ve had enough of us by then. [laughter] I Dude, I I hope we’re still doing this,

1:40 But it’s been fun, man. So, nothing else will be retired and old people talking about I remember when data was like you had to talk to the computer with a keyboard, like all kinds of crazy things. [laughter] Crazy. So, awesome. , random note here. Let’s let’s talk about our main topic and I’m going to come give you a little quick just beat from the street here. I have something I would that came to mind Tommy as we were talking about the future of things as well. So our topic today is untangling workspaces, branching and artifact chaos. All the artifacts

2:13 You build inside fabric. So just getting your whole fabric experience organized like let’s talk about that. This is a mailbag today. that will be our main topic. All right, dude. I’m I’m excited to get into We’ve been getting a ton of mail bags lately, too. I don’t know if you’ve noticed that, but it’s been great. Well, the callouts have been good here because they all know we’re going to be leaving for holidays. They’re like, “Well, maybe I get my question in for the holidays and [laughter] maybe that’s what people are thinking.” So, , definitely go to PowerBI tips. If you have a question, if you want to put one there, we will definitely address it. We, it comes into our backlog and we’ll pick out the best topics from

2:45 There as well. So, anything’s fair game for PowerBI and Fabric. Feel free to ask your questions. Real quick, Tommy, I want to give you a very brief beat from the street here. , I remember you and I talking like a long time ago on the podcast. I wanted to just say it out loud just to say we talked about it on the podcast. I had originally noted that the the world of AI and how we build apps is going to be able to let you build any personalized app in a rapid succession. Right? You can go get an example of an idea. You can build

3:18 Something that’s a personal app, a personalized app that does something I needed to do with relatively ease. Today, I saw the first article of someone else talking about the same concept and saying apps are now being able to be developed at a fast enough pace and clip that not software developers can go describe the app that they need and most of the main components of that application are being able to be built by AI. So, , Tommy, you had this idea like, “Hey, I want to

3:50 Build a a note-taking app or I want to build a grocery list thing, but I don’t want to pay for the set app to go do it, right?” So, if the app that you’re trying to purchase can’t do this very elaborate barrier to entry item, like, , use maybe AI or do something fancy, you can literally vibe code the app in like an hour or two. So now it becomes a trade-off of do I want to pay $5 for this grocery app that I maybe would use for the next year or two or do I want to spend an hour of my time right now to build exactly what I need and

4:24 Then iterate on that design and build it to be what I want. Like we’re I think we are stepping into a new era, Tommy. Like this is this is a brand new world where you don’t need to go buy apps anymore. you can build many tools and apps as you need them. What are your thoughts? What are your reaction to that? Honestly though, I I haven’t thought about it from the iOS side or from the mobile side, but what I’ve actually been doing is creating apps that is just like hosted on my local network that just run based on various needs. I have like I have

4:58 Give me an example. Give me give me a clean example of this one. I’ll give you a very basic general example. So always dealing with local like actually in local AI models and I want to be able to organize manage categorize them because like if you download from hugging face it’s all convoluted in terms of the files and the types and it the names are not very descriptive like oh is that for this type of generation or is this instructing so what it does is it actually you need like a library of like hey is this image generation or is this natural

5:31 Language or like is this exactly then like how much I like it, , where it’s located. So, basically Oh, interesting. It creates like a SQL like database around the folder and then I can organize it, categorize it, and then basically say, hey, this is yeah, like for captioning, this is for computer, , coding thing. So, I The back end is SQL light on your computer. The front end is what what did you build the front end in? Oh, whatever it built in. I think it was like just JavaScript a basic app kind

6:04 Of thing and then all I do is and like it creates a batch script. So all I have to do is like add a shortcut to my program and say start the server and then I can access it anywhere on my local network. It’s great. So so it’s it’s running. So you’ve actually built a server and it’s it acts like a web page. So basically it’s like a server URL you like local host whatever port whatever web app. Exactly. And then you can just go there and just run it anywhere on your local home network. Exactly. So, so this is my point, Tommy. I think this is going to become the commodity. Like, , when when do I get to have a window into my iOS, iPhone, Android app

6:40 And just say, , instead of going there to to get a bunch of games or things on your phone, why don’t you describe the game that you want? Hey, I want a game that does this. Hey, I want a game I’m I saw this game that does this this thing. Describe into the AI what to do. you could build all these little mini either games or apps or tools or solutions. And I I just want to say I saw it coming. I’m starting to see this become a reality. Like there’s all these different app dev companies that are like now touting build your entire app on top of this program on this platform.

7:14 Lovable I think is one of the companies or the the websites that are doing this. Claw Code is a lot of is doing some of this stuff for you as well. like they’re very advanced in be able to build massive apps and help code bases. So I think AI right now is really good at writing code and building apps because it is such a structured way of doing things. If when I look at an app Tommy I think about a couple things authentication like how do you sign into it storage of information and then the UI that sits on top of it. Like

7:46 If you can if you can get AI to build those three things for you and you can get like a a pattern of prompting that lets you build those things and build anything you want there. Anyways, yeah. Have you anything like that? I’m starting to I’ve been building lots of little apps that I’ve been throwing away cuz I have a specific purpose. I’m trying to educate myself in a way that can let me build what I really want to build, which is more elaborate fabric workloads using AI. , I recently just did a video, I think in the last month or so with Gird, got a

8:19 Really good number of hits on YouTube, but Gird from the Microsoft product team built workloads v2 and inside that workloads v2, he’s using claude code to generate a lot of the code to build the workload. So to get it set up, to have instructions, to have a readme file. So he’s got a project that you can almost chat with to help you build the app that you need to build, which is pretty amazing. So interesting. I’m I’m trying to unpack like, okay, what did he do? How can I further leverage this? How can I build my own apps? , and I’m now picking up

8:51 Small projects to your point, Tommy, around the house around everything I see now in my family is like, there’s a there could be an app for that. There could be an app. Like, who left the dishes in the sink? I could make an app for that. , , hey, I want to go sell sourdough online. That that could be an app for that. I could build an app for this. every little I could I could see this idea of there’s a lot of manual tasks that I’m coordinating that I’m thinking now in terms of if it only takes me a little bit of time to build the app why not build it and so that’s where my mind is going right now more recently my

9:27 Development has been focused on I have so many media analytics platforms Tommy we’re all over the place we’re on LinkedIn we’re on YouTube we’re on X we have like Rumble like all these different platforms that broadcast to. I’m trying to get my head around how many what’s our usage and metrics and what what things are good content or not good content. And shocker, each of these platforms lock you in with their analytics and they don’t really give you anything. You don’t really get much information out of their analytics. So, it’s very difficult to get a good read on what impact your social media stuff is having an impact

10:02 With unless you’re paying for advertising to these platforms. So that being said, I’m I’m doing a lot more development around, okay, what can I use AI to help me do to go scrape the web, get content, log into portals, and then get data out of there that I need to go analyze later on and put it to fabric. So that’s where I’m at right now. One thing I’ve observed in general with app development, Tommy, everyone uses Docker. Docker is like a thing. And while I like Docker, I’m like I’m I’m torn because I want Fabric to be my backend. I want because

10:35 I have a SQL database. Like seriously, anytime I build an application now, I’m thinking I don’t really want to go buy an Azure based backend data system thing. I want to use what’s already in fabric as my backend. Cosmos DB, KQL databases, or or SQL databases, right? These three databases are operational systems. They’re they’re used to build apps on top of them. And the reason I want to use them is because once I use them as my app backend, any data that’s being used or generated there easily is added

11:09 To my lakehouse. It’s easy for me to do analytics on top of it. And at the end of the day, it’s no use. , you can have tools that help you build a process, but at the end of the day, you need all that information to be sourced and exist inside some reporting solution. So I’m designing like the report facing thing backwards into like my applications. Where do I put this data so it’s easy to get access to? And yeah, there’s some things here. You can do mirroring to a Postgress database. You can mirror it to a Cosmos DB, but I really just want the data to be stored inside fabric. It makes my opinion it makes my life easier. What are your

11:42 Thoughts? No, I I I completely agree and I think that’s one of the things I’ve been trying to do is, , some of the data that I actually want is from fabric. I’m like, “Okay, do I have to open a PowerBI file and do the old school like just an export? What do I want to do here?” Yes, exactly. And like even at one point I’m like, “Should I even embed in this web app?” yes, like a power one of my PowerBI reports. So there’s a ton of good options there. So even if I go to like some platforms like let’s go to Twitter, look at your

12:13 Analytics from Twitter. There’s a download button. You can go into Twitter. You can set a time range and you can download a bunch of data. Great. like I want to automate that. Where’s where’s the API to automate this stuff? Like that to me, this is so clunky because I can’t get the information out that I need and it doesn’t put it where I want. I’m I’m really close. Like the automation is so close to being there, but we’re not quite there yet. Like it’s not exactly there. , one last thought here, Tommy, before we move on. Have you heard of the browser called Comet? Oh, I I did try it out. Yeah,

12:48 That’s per perplexities version of a web browser that has AI enabled on the browser. I’ve been substantially impressed by the Comet browser, an AI based browser. It’s basically I think built on Chromium. It’s a Chromebased browser, but the AI can take over your window and start doing a bunch of things in the window that can like open a page, search for things. And I was just having it do some research basically on some companies that I had on LinkedIn. And I just basically gave Perplexity the the comment browser said, “Go find this

13:21 Company on LinkedIn. Research what they do. tell me what products they make and give me like a summary of how I could best sell a data platform to them, right? Just some general terms.” And it was able to go to the website, find navigation, it was taking screenshots of things. It was reading the screenshots back. It would say, “Oh, I found the search window.” And then it was typing things in. Oh, look look at found the company. And it was clicking on stuff. This is next level stuff, Tommy. And while it was doing the research and going through the browser and doing things, I could shift my attention to something else, which was

13:54 Impressive to me. So this is where I was fully taking and this is the first aha moment I had around taking a task that I needed to do, describing the task to an AI agent and letting it just run and do it and do its research and then I could focus my attention on something else. Th this I think is what AI is going to be doing. I I I completely agree and so I did try Comet. I was a little frustrated because maybe there’s some setting or some extension that was affecting it

14:25 Because it would be like I would ask it a question. It’s like hey we’re doing we’re doing all those changes now and you’re like no you’re not you’re not doing anything right now and it was it what what is it called a hallucination? So mine my thing was definitely hallucinating. So your comment was hallucinating. Sounds like again Tommy these words we use like the things we say out of our mouths like if I went to any person who wasn’t a data person and said I’ve got this thing called fabric and I’m using comet and it’s hallucinating you would think I’m

14:57 Crazy the terms these words don’t even make I don’t have any context what these words mean together it’s like a word salad of things anyways that being said let’s get into our main topic today I think this is interesting because this idea of lots of things in workspaces is branching and CI/CD strategy. I think this plays very well into our question for today. So with that, let’s hard transition over to our question. Tommy, to you into it. Alrighty. So I’m the director of data products and analytics and I’ve

15:29 Been listening to the podcast for a few years. My department is responsible for building maintaining an enterprise data lake, PowerBI reporting and a selfservice community at a 10,000 person company. We are all in fabric but can’t get a straight story from Microsoft and our various consultants on what are the best patterns for organizing workface and artifact. Love this already. For example, managing dev test prods bronze silver gold pipeline leak house notebook semantic model report app and

16:02 Business domains. Further complicating that the CI/CD feature still seem partially complete. So how does one manage feature branches and pull requests into dev from a developer without proliferating workspaces? Also love that also why are we trying to standardize around the direct link pattern but that has or we are trying to standardize around the direct link pattern but that has implications on colllocating lakeouses and semantic models. Woof

16:35 A lot of a lot of details here. So there’s a couple statements in here. I’d like to correct a couple things just from an understanding standpoint. I think this is a really good question. Solid, right? this is this is what I would say the fabric adoption roadmap. for those of you who are not aware of that, Microsoft has done a really good document around the fabric adoption roadmap which used to be the PowerBI adoption road map. But fab they they’ve changed some of the language here. And one of the things that I think that makes PowerBI and fabric really exciting for us to use is it’s not a one you don’t have to conform your business

17:08 To what fabric does, right? It’s not a one-sizefits-all type story. You can design the system how you need to and use the system that works best for your business structure. When you read the adoption roadmap for fabric, there’s a lot of effort on diagnosing how your company does work. Who are the people? What are their skills? What is your threshold and tolerance for data and data security? Those are some fundamental questions I think you need to ask upfront and decide

17:42 What those look like first. Once you decide on those initial experiences, then you can back your way into, okay, how does security, CI/CD, deployment, central BI, and community all fit together? Because it all does fit together. There’s patterns here that Microsoft is describing. , again, Microsoft has been working with the world’s largest companies. They all use fabric, I think, Tommy or or they all use PowerBI for at least at least PowerBI. I think I last time I heard was 98 of the top 100 Fortune 100 companies

18:15 Are all using PowerBI at some level. So PowerBI is pervasive. It’s everywhere. and we’re starting to hear more and more adoption. I think one of the recent notes at the Fabriccom Vienna note was the fabric platform is Microsoft’s fastest data platform growth ever. all their data platforms, no matter what the Synapse, , Azure data products, the fabric experience is the fastest growing one because they’re making it easy for stuff to work together. So yeah,

18:47 Let me just pause there. Your your initial reactions to the to the question. this is literally in one all the frustrations I think we are still having with fabric and I know you mentioned the fabric roadmap but it is it is it’s good but is it complete or anywhere near complete not at all especially with the the needs that we have and the the conversations in a in a wider audience just dealing with the workspace and artifact organization is there a simple straightforward way you really

19:21 Have to break that out. Right now all the things on the road map the categories and the scenarios are general and that works but only if you’re using a certain workflow or a few of the workflows which are usually not the case. with at least with PowerBI the workflow there weren’t too many divers versions in terms of how you develop a semantic model right where usually I’m going to import or I may do something with analysis services where the data is coming from but that

19:54 Is the same funnel right obviously different approaches and there is still a comprehensive roadmap for PowerBI on the best way I remember the PDFs like the 220 PDFs when PowerBI first came out on building that up and it was very intense and that was just semantic models. Now we’re introducing all the other features with and and again to one place and I think this is going to take time until I well let me rephrase that. It’s going to take time and feature

20:26 Updates for this to become something that we have a good grasp on. I’m gonna I’m gonna jump into a couple comments you made there, Tommy, around the the Fabric platform. We’re two years into this. I get it. Right. I remember when PowerBI came out. When it first came out, it was good, but I had really no reference point to compare it to, right? I didn’t have any expectations initially coming out. It was like this brand new thing. Didn’t really know what was going on.

20:58 And Microsoft was very adamantly listening to feedback in that early stages of like build what we need. Year one of fabric has been a lot of Microsoft listening and refining and getting things working. I think I think there’s a lot of issues with getting things working. Mhm. Recently though, this last year, I feel like there’s been a lot of technical debt that was required to get out of the way in early year two, so that now we can get to a place where we’re starting to see like real feature

21:30 Production that is helping me build stuff. That’s cool. Like it’s it’s helping me go faster. And what by that is like let’s talk about dataf flows gen 2. Tommy, we gave dataf flows gen 2 a really hard time. Too expensive, too slow, can’t handle big volumes of data. This year we saw a huge improvement with like parallel processing in data flows. We now have accelerators on it. We’re using a new net backend for data flows. Like Microsoft took a really big like that’s a monumental investment to

22:03 Rebuild the engine of data flows gen 2 to make them more performant, faster, and more cost effective. There’s even a whole new pricing model. If the if the model run if the data flow runs for the first 15 minutes it’s a certain rate and then at 15 minutes the the rate cuts in half like it’s it’s a substantial decrease in charge billable rates on cus after the first 15 minutes. There’s a whole bunch of stuff here that I’m like they’ve done a major rebuild. And so when I look at this there’s a lot of technical debt that had to get eaten

22:36 Up and out of the way for us to really start seeing improvements. Now, what I feel like is we’re doing a lot more. Again, we’re back in this in the mode of listening to the customer. What do they need? What are we trying to build? But I see the the whole fabric platform becoming much more mature at a much faster clip. If you ask me when we talked about fabric, Tommy, when it was initially going GA, we’re like, it’s okay. There’s a lot of friction. Give it a year. give it a year to and then now I feel way more confident leading with fabric as your only data solution like

23:10 This is the data solution we’re going to use and you you can be effective with it. Let me pause there. What what are your thoughts on that? I agree to an extent. So I I’m actually curious for you. G give me some examples of those things that have changed or updated where that’s been that catalyst for you feel like your workflow the the that process is actually come to being more productionalized because I’m curious on what you’re seeing as those big feature those advancements in the tool. Sure. Couple of things that I think that

23:42 Are very useful here are one is an improvements on data flows. That’s been a big improvement for me. I like using them. The UI is useful. When I think about all of the users that I work with that come from PowerBI and are moving into fabric, their first inclination is let’s build a dataf flow gen 2. That thing has to be cheaper and more efficient than the dataf flow gen one. The other big thing improvement here has been just ease of using the lakehouse. That’s been a major improvement for me as well. So you easy ease of getting data in, easy to get it out. This is a very nuance feature, Tommy. I know no

24:13 One else likes this except me. This is just me a thing. You can now download files directly from the lakehouse. You can go to the lakehouse in the web and download a file from there. , duh. That should have been a feature that existed a long time ago. That’s so I can create files that are produced from notebooks and now I can download them. Yay. This is what I wanted. Like I want to be the the lake the one lake is a file storage system. I need to be able to get stuff up and down. Should be easy, right? So you could use the one drive one link connector for your desktop and synchronize files as well.

24:46 That’s another way you could have gotten files up and down from the the one link but now you can do it directly in the browser as well. So it’s like improvements like that. The CI/CD patterns was very rough Tommy getting anything to deploy from dev to test to prod in those early days it was painful and I had a lot of issues with connecting to git and or azure devops where things would deploy but not deploy not synchronize correctly. there was a lot of bugs I think around that system. To me now most of those are worked out. I don’t have any issues connecting to things. the other big improvement

25:19 This is more of a PowerBI side that I will directly point to is the PBIR format. PBIR has been a game changer for me. Like I build products on top of PBI. This is amazing. So the the level of automation and level of effort we can build out at the PBI level is absolutely incredible. So I really like the PBIR format. That’s another big win and helping you create teams of people to build products together. And so I think these are some of the questions that

25:51 This speaker is asking about. The question here is what are the patterns from organization is one of the tone I hear in this question. Another question I hear is DICD doesn’t seem very robust yet. So what does that look like? Right. And then I also hear standardizing on direct lake patterns. And then they’re talking about colllocating lake houses and semantic models which I’m not sure I necessarily agree with that mentality. a semantic model using direct link must be in fabric but

26:24 You don’t have to colllocate them in the same workspace. that doesn’t that’s not necessarily true. So I I think maybe there’s a little bit of missing on information on some of the direct link items. how I would view direct link is pre-processing the data before it gets to PowerBI. So there’s also been a lot of improvements around Spark as well. So those are those are my examples. Tommy, did I No, I think that’s good. And I think the the PBR you’re dead on. You’re dead on in terms of the advancement that’s made u the feature set and what is available.

26:56 Yeah, that totally. For me though, that doesn’t help answer the whole story at least with the the question here because that’s only we’re we’re still dealing just with PowerBI, right? In that situation, we’re not dealing with the other types of artifacts here. So, there’s still a ton there that we remains to be seen in terms of the best practice. So, I know we’re getting a lot better with the lakeouses, but I think when you are dealing with multiple

27:28 People with multiple roles, let’s just deal with the and I don’t even get into permissions. Let’s just start with the basic setup of a lakehouse and then people accessing that and doing additional actions on them, right? So, if I had obviously the lakehouse do semantic model, great. But probably people are going to want to be doing some data engineering things on it to enhance it. Maybe there’s some data science things and the best practice to unclutter where all a sudden these

28:00 Notebooks are created. Now there’s multiple you have the semantic models maybe someone’s doing an experiment maybe it’s getting pushed to a SQL database because we’re doing the applications right I think there has to be I I wish when you created an artifact that it would give it a or you could set up like a template name right instead of notebook one right I I hate that so much of there has to be some naming conventions right off the bat that you have to implement and that’s not something that’s in a sense a global

28:32 Acceptance or a global process. like how we have DAX format best practices. We don’t have that with fabric. The organization of it, not just if it’s in a folder or not, but how many workspaces are we creating? The idea with the CI/CD and that workspace explosion is 100%. To be honest, Mike, I stay away with some of my projects with dealing with a dev prod with workspaces because I hate seeing so many freaking workspaces in my environment. It’s so frustrating because

29:04 They’re not organized because I may have a get repo with branches, which makes sense, but in PowerBI, you see I it just becomes a nightmare. So, yes, you can do that. Yes. I don’t think that’s I don’t think that’s the best experience. I I think so I’m going to lean back here. So the tooling has gotten a lot better, right? So let’s let’s talk specifically. So patterns for So there’s two things I want to touch on. Patterns for your organization and then CI/CD patterns. I think those are two main topics we’re talking about in this

29:35 Question. The one pattern here is what does this look like for your organization to have central BI and people consuming data, right? I think that sets up a lot of the pattern around how you identify what teams need to do what work, right? I’m a firm believer of use the UI that that PowerBI gives you the data story around. So if you if you want to be an organization that like is deeply integrated as you can without a lot of extra effort like writing a lot of code. If your team is

30:07 Let let me say this way I’m going to step back here. If your team is very comfortable in VS Code and running Git, right, that’s a traditional software development experience. Most business teams don’t see it that way. Most business teams are not thinking in terms of GitHub check-ins. They’re not thinking in branches. They’re not thinking in these other patterns, right? That is a software development world or even maybe a data engineering world that we’re bringing more towards that business user. So, one thing you have to do as an organization is educate your

30:38 Team. What is Git? How do we use it? What’s the integration with it? If you’re a bigger organization, we’re at 10,000 people for this or and again, I don’t know what the ratio is of like how many developers versus how many consumers of information here, right? How much of this needs to be self-service? How much of it doesn’t? Those are decisions you need to make at the organization level. But at at 10,000 users potentially of PowerBI, you are going to need a central BI team. Period. I think that’s a given. Sure. Yeah. In that team, you need to identify what does central BI and data engineering need to own. Is there a team

31:12 Of individuals that are coming from it that is going to build data engineering products or data tables inside lakehouses or is that going to be owned by the central BI team? Then once you have this identified, then you can specify how do we distribute and deploy things. And this is where things get a little bit again it’s getting better. It’s it’s not quite perfect yet, but it’s definitely getting better. inside the deployment pipelines you can use variable libraries. This is one of the the technologies that Microsoft provided that I think is extremely useful here. Variable libraries allow

31:45 You to relocate variables between servers and data flows and all these other things that you can’t quite connect. So as you move items between different environments, you can use the variable library to switch out those variables as you go, which is really ideal. Sure. Yeah. I I I don’t I don’t have any qualms there, but I I think you’re giving a lot of credit. I think you’re being very very generous, Mike, on the workspace organization and that ability because I don’t think you said like we’re getting there. I don’t think

32:18 We’re close yet. I think Do do you use domains and tagging? I do. I do. But that’s my solution for this. Like the the amount of workspaces I get anymore, it’s not a problem to me. If you’re not using domains to build out what you’re doing for like dev things or like domains and filtering is everywhere in the product, it’s super useful and I don’t have any issue with using domains and tagging to help me align what work I’m supposed to be working on to patterns inside the development system. I’d also argue Tommy, you’re you’re the

32:51 One setting things up. So you are the one that can see the proliferation of all the workspaces. When you start talking central BI, I think there’s a pattern here that developers don’t need access to test and prod. They should not have access to test and prod. So if I’m a developer building products inside fabric as a data as a data developer, I should only have access to dev. That’s the only thing I need to have access to. So when you see a proliferation of workspaces, you should never be deploying from desktop to test, from desktop to

33:24 Production. There should always be some a deployment pipeline or some automation that’s deploying out what you want directly in those environments outside of this. So going back to the CDICD patterns, once you identify proper structure in your company around central BI and one lake catalogs and what you’re going to build there, you need to be using domain tagging. You need to be figuring out who your developers are and you need to be building a developer level workspace and thinking about when does the developer create their own effort and merge their changes back into the dev branch for

33:56 Integration testing and then how do you get from dev to test to prod. I think it’s with the deployment pipelines and I think it actually works pretty pretty well here. I don’t have a lot of issues with it. So you you would say completely recommend in the deployment pipelines as where you would go with this. This is a 10,000 user organization. I don’t think there’s enough people in this organization to actually hire a DevOps team to auto deploy everything that you’re going to need everywhere. That thing I think I think we’re not big enough yet to do that. , I think at this size company, you’re going to have

34:28 A smaller central BI team, maybe three, four, five people max, right? And at that scale, I think it makes a lot more sense to have the deployment pipelines just handle a lot of that work for you. And again, I’m I’m thinking here in this pattern, lakeouses for data, semantic models for modeling, and then reports on top of that. And then then everything flits fits very nicely through this pipeline. And with you’re using those objects, you can easily design around variable libraries and semantic models to get everything linking correctly

35:01 Together through the different environments. [sighs and gasps] And here’s the thing though when you’re also dealing what what would you say then just about from the naming convention or the that’s one thing about the automation of pushing things for me when I’m thinking about this the and we’ll set aside which is impossible the CI/CD features here the organization and I I do love in the mailbag how he he does outline the multiple kind I don’t want to say scenarios but a

35:35 Higher level than that. You have dev test prod and self-service. You have bronze silver gold. You have the pipeline lakehouse and the models. And I don’t think the business domains there because here’s the thing to me I honestly feel business domains still still feel like to me that they are intended for it’s like an it’s an organization or business thing not a development thing. And this is what I this is what I’m trying to say. consider when you look at GitHub branches. If you don’t know what a GitHub branch is, it can look intimidating, but

36:09 Branches make a lot of sense, especially the way you actually navigate it. , domains are just more or less a category. It might it’s a glorified tag, and I hate saying that, but it’s not really meant from a development point of view. You can say it is but when you look at other patterns of organization and for different versioning domains are not helping there. It it can help in terms of your departments but again we were trying to already solve

36:40 That before when we had like workspaces and Microsoft’s or best practices was build your workspaces around your departments. Okay? So domains could be a higher level than that. Do do you see what I’m saying here? There’s there’s the business view of categorizing and organizing information and then there’s the development view. Yes. But one is not required to have the other. What I would would argue here business domains are I’m the HR team. HR team human resources stuff. I am the

37:14 Sales team. I’m the research team. Right? those teams may need similar data from the same tables or they may need entirely different data. So your strategy around and there’s a lot of terms here that are being mixed up that I don’t really like in the question here. They’re talking about like okay how do we manage dev test prod okay that is a a procedure that you use in devops development operations to help you provide consistency with what you’re deploying right you do dev test prod and you need to have clear definitions of

37:47 What those mean that is not the same as bronze silver and gold those are not the same stuff inside dev you may have bronze silver and gold enrichment of data and that may be carried through all the different environments. So just wording all these things together doesn’t really give you the answer. So dev test prod is designed for a central BI team to regularly deploy content at a at a good standard clip. Bronze, silver and gold are ways to organize and collect data from any data domain that

38:21 You’re talking about. Whatever the system comes from, right? It could be SAP, Oracle, right? You bring in the raw data, you transform it, and you make it better. that pattern bronze, silver, and gold will exist in dev, test, and prod if you build the process correctly. So you’re you’re you’re it’s like a subset. It’s another line of thinking. And then we talk about data domains or business domains, right? Business domains. I don’t need the same data that you do, Tommy, all the time or I don’t look at the data the same way that you do. So to me this is a you can’t you have to look at your organization and

38:54 Have that central BI team decide on some patterns you’re going to follow and then you use what data does this business need you use do we do bronze silver gold or other layers of enrichment of data again bronze silver gold is just a concept it’s taking data in at a raw level and producing it to be consumed by somebody else that’s all that it’s doing it’s It’s it’s a it’s the data warehouse. There is bronze, silver, gold is simple in my in my terms. I don’t think people over complicate it. They

39:27 Try and make it more complicated. It doesn’t mean that. It just means it’s a data warehouse and you’re refining and creating efficient tables to be consumed. So when I look at this, I go, okay, you need to align your business on what pattern should you be following. The central BI team should define how many environments they have dev test prod and then what does a self-service workspace look like. So to me I think you need to understand what data is being currently developed what are the needs of these different departments and do they need the same data central BI

39:59 Should be organizing all that that this is all of that is being organized and managed independently of fabric. You don’t even need fabric for this conversation. You you just You don’t tell me why you need fabric to organize what data is used by department. Fabric is not required to do that. You should know that before you go to fabric and then design fabric around what your business needs. If you’re using fabric, then fabric’s required. I I I see what you’re saying

40:31 To a degree, but once No, no, no. The design of your data should come from your business and your central BI team. You’re just choosing fabric and you’re going to conform the pattern of how best to deliver that data into your fabric environment. I could do dev prod. I could do not cicd deployment pipelines. I could use as your devops deploy everything. I could build my own tool. Like there’s a whole bunch of other tools and solutions you can put on top of this. You shouldn’t look at fabric and say what does it do and have it conform to your business process. You

41:04 Should instead say how do I maximize business value first and explain and describe my company in what data and how they need it. Let’s start with that process because you may find you don’t need a test environment. No one’s going to test anything. Great. Then that doesn’t become part of your process and it just informs how you build the pipelines of deployment. You see what I’m trying to go after here? I I do, but I’m not sure if I completely agree with your sentiment

41:36 There. or or at all. But that’s that’s besides the point. I think I want to move on to something else here, too. We’re we’re talking about You can’t disagree and just not and and not give a reason. So, I understand the rabbit hole. I don’t want to go down the rabbit hole because That’s all right. But don’t disagree and not give me an answer. Like, go ahead. What What am I missing, Tommy? like what what what has to happen here like once you the technology is going to dictate so much of your organization and your process. I see what you’re saying

42:09 To a degree but Mike if let’s say if you’re using Git, you are utilizing the branch strategies. You’re utilizing version control the way Git intended it to be. Okay, pause. Hold on. Pause. There are so many different patterns in how you do Git. There’s Git. There’s GitHub. There’s GitLabs, there’s trunkbased development, all of these there’s just because you say the word git doesn’t mean you’re developing in a single way. There’s a whole bunch of different patterns and and elevation like just all git is is what Greg talked about in our

42:44 Recent podcast which is git is just storage of files. That’s it. No way. Absolutely. Yes, it is. And just you use Dropbox then. No, it’s not. It’s not. It’s it’s concurrent use. You can develop concurrent products at the same time. It’s a storage system for files that tracks all your changes. Git is not just a it’s not a it’s not a process. Git isn’t a process. Git is a way of you using a technology to inform a process. Git is a technology with a process around it because no Oh Mike Mike if if

43:18 Git was just storage for your files if that’s all all was then just use Dropbox and make it easy to have something easily synced on your computer instead of having the headaches of committing and versioning and you’re wrong. You can decouple the process away from Git. Git is you develop the process first. That’s why there’s different deployment patterns with like different processes. You can develop a different process and then use git in different ways. Okay. So let’s go down that route. So

43:51 Let’s say you want to use git for totally different reasons. Are you going to be efficient and is it going to be organized for you? You can do it. But here’s the thing. every technology if I’m creating databases whether or if I’m storing files there is a best practice a recommended approach to using that tool I don’t use GitHub and I don’t use git to store all of my I don’t know let’s say all my old movies I torrented years ago thing I don’t thing there’s better ways

44:25 And better products to do that and in the same way with fa with fabric We are dealing with not just a technology but you’re dealing with again the what that content is. If I want to view my movies am I going to use GitHub? No. I could you’re to your point I absolutely could but that’s not going to be the most efficient way for both the user experience and then just trying to organize that too. But would you agree or disagree with that where yeah I could use Git for other purposes. There

44:58 Are but there’s a recommended way of using that tool. I don’t like your analogy of [laughter] movie. I don’t like your analogy of movie files because I can’t tell you the last time I needed to store a version of a movie file. I don’t version them. So, but you just said git was nothing but storage. It is. It’s storage with version control. Just like I said, you’re not listening to what I’m saying. It’s storage. It stores code that you would deploy with versioning. It allows me to have multiple files all in different

45:31 Forms and versions of them. Again, we just talked about this with Greg recently, which was I thought I really align with what his definition was. It’s version control. It’s, , I’m building something different. I’m changing code. I need to park that change and go back and fix something from production. That’s exactly what gets for doing. It’s storing the files and telling you which version every single file and change is being tracked. So that’s what git is doing. The process on how you get things out the door that depends on what type of process you’re

46:03 Doing. Bunkbased development, , GitLabs, GitHub development, like there’s different patterns, processes that you put in place for how your team works, right? So those processes are what you can use to get things out the door. What we’re going back here to this client question is talking about how do we unpack the central CHCD pattern [clears throat] again I’ll go back to yes this is technology that exists you need to understand how it works but you still have to go back up to how do you develop does your team

46:35 Like to develop trunk based development how many environments do you need what what is the mentality of getting things out the door so again I think the question here around I don’t like this idea Microsoft fabric can’t get a straight story from Microsoft or their various consultants or what are the best practices of organizing workspaces and artifacts. That’s wrong. That line of thinking is wrong. It’s not there is no best practice. It is start with your business needs. Align. This is what the adoption road map does. Read the adoption road map. It’s clear. Define

47:07 How you want to do work. Figure out what process works for your organization. Come back to the technology. Look at the technology. And then from there you’re able to formulate your process. So are there best practices? Yes, there are some. There are trade-offs from going one way or the other. If you use DevOps versus GitHub, there are certain authentication trade-offs you have in each each solution. I get it. That’s that’s a trade-off you have to operate as a business. Anytime you decide to make a trade-off, you either are deciding to take the easy route and use

47:40 Something that’s automated from Microsoft or you’re taking a more ownership responsibility around having to build more things on your own. You don’t need to use deployment pipelines. You can roll your own deployment strategy. You can track everything in Git if you want, but I would argue the skills of your team will not align to self-service using Git. I would argue your central BI team needs to be educated and trained on what this new process is. We’re going to use PBIR formats. That’s what we’re going to do to build that is for a central team. if you want to roll that out to business

48:14 Units and have them help them, you can provide support around that. But I do think you have to start with the business strategy of like what you need to do to ensure quality and and value out of the information and keep it secure. So I think you start there. And from that it forms a lot of your opinions around what is the best practice for your business. Would you agree or disagree? Regardless of the tool or in a sense your process, there are still at least some we’ll call them fundamental recommendations, right?

48:47 Regardless of what you’re doing, for example, naming conventions does not exist like that exists outside of fabric. and in terms of the from the organizing your content obviously that’s a little more u dependent on the technology but there I I think that there are some universal universal best practices here that are not necessarily for like artifact specific I like yeah

49:18 I I like your I like your thinking here Tommy this is going to be something we’ve for the ones in this podcast for this episode we’re agreeing on because we’ve we’ve been adamantly disagreeing the whole time which is good. It’s good for it’s good for TV. but okay so let me go let’s define I think what you’re you’re hinting here is what is best practice right so best practice is aligning the tradeoffs with your process of what you’re going to build I think is what best practice is right so when we say best practice yes Tommy we don’t

49:50 Need to use deployment pipelines we can get rid of all that but the trade-off to that is your team now needs to know how to use APIs or semantic link labs to deploy items or write powershell to deploy the things you need to deploy into fabric. Right? So the tradeoff for that best practice is does your team know PowerShell? Do you want to manage a lot of extra code? Do you have a team to support deployments when something goes wrong? If the answer is no in some of those areas, there’s only two options you have. Educate the team and figure it out

50:23 Or don’t use it. That those are that’s the trade-off. So when I look at best practices, I have to take a lot of input in and say what is your skill? And this is why I keep going back to like the business, what does your business know? What are best practices? How much effort do you need to spend on these different activities? Do you need to have central reporting? Yes or no? Okay. If yes, here’s some decision things I need to be considering around that central management. If I’m going to use GitHub versus Azure DevOps to do deployments, where’s your team skill? Do you even

50:56 Have licensing to one or the other? How big is your team? Do you have GitHub already attached to your company as an enterprise license? Do you have policies on that that are going to inhibit you from using Git to integrate your systems? Right? There’s a whole bunch of things already there that are going to potentially prohibit you from going down that path. So to say there’s a single best practice, I think is a bit naive, right? You have to define what your team knowledge is and what is the thirst of how much skills do we need in order to get all this to work. Does that make

51:29 Sense, Tommy? Because I think there’s trade-offs to everything. I I completely agree. And and I think we have to Good. Ship it. We’re done with this episode. Yeah, we’re good. We’re good. We’re good. No, [laughter] I I think regardless, we’re just going to deal with fabric where there are there’s not going to be a one obviously without being redundant, a one-sizefits-all, but I think there I think that with all the recommendations and all the approaches, I think you said it on the the nail on the head. No matter which way you choose, there are trade-offs to it. There is not a un , like it’s

52:02 Not like, hey, we have managed self-service with lakeouses and also data science. Okay, template here. That’s how it was perfect. Like, well, you could there are multiple ways to approach that. There are some better than others, correct? And also depends on your team skill. Like, if your team comes from like superdev world and you’re always in VS Code all the time, I have no problem with you taking more of these more ownership on deployment things. Interesting point you go here where you’re saying what the team’s comfortable with. I struggle with this one because it’s that chicken before the

52:35 Egg type of approach where if the team doesn’t know something and like that’s available in the product do we cater the organization to that or do we upskill so this is I like [laughter] this is also an interesting so two points that I would consider in this area right the fact that we’re bringing a fabric environment to a team regardless is going to require people to learn things. So like at the end of the day, we’re going to have to learn stuff period. Right? I would argue here the two areas I think

53:09 Is do we already know how to do it and then are we willing to support it moving forward, right? So if I have a team that’s using something inside fabric, for example, let’s talk about deployment pipelines. fairly easy to learn, has some tricks of the trade you need to use to get it to work correctly, but I would argue like the learning curve on the deployment pipelines, it does some things really well and it will help you move items between dev to into production, right? You have to understand what is the level of effort

53:41 For you to train your team to use that tool, that part of the tool versus we’re just going to roll our own and we’re going to deploy everything ourselves. If you have a dedicated deployment team and the funding is there for you to own that entire process or spend the additional time, the only difference between that and using deployment pipelines is a longer learning curve, more testing, a person to support it. What happens when Tommy goes on break? Is there a second person, a backup person to manage and own that? How much knowledge do I need to give them to train them on how to use

54:14 The process? Right? So there’s an extra burden of education that goes along with owning more of the process. And so if your company decides that’s a willing an acceptable threshold for you to take on, yeah, fine. Do it. Does that make sense, Tommy? So I I agree with you. I think at the end of the day, I’m saying you’re going to have to learn something regardless. It’s going to have to be done. But to your point, Tommy, it’s a I I not sure it’s really chicken and egg. I think it’s more around what strategic investments are we willing to make. What’s the commitment to make this

54:47 Happen and do we have ample backups? And again, Tommy, for a long time, you and I both did this. We’re the onestop shop. We ran the whole show. It was a one-person show. Like whatever I learned and whatever I thought was best practice was best practice and we did it in our organizations. We built it for them, right? That that’s the starting point. So, I think the reason you bring in consultants is to help you weigh some of these decisions, right? , a consulting firm should come in and say, “Here’s three options for CI/CD. You can use the the built-in one, you

55:19 Can use a hybrid approach, like a little bit built-in, a little bit not builtin, or you can use a roll your own. Here’s the three options. We know how to set up all three of them. Here’s the trade-offs in each one of these areas. And so if you’re doing the deployment pipelines, here’s best practices for using that process. Here’s best practices for using a roll your own solution and CI/CD, you need consultants to show up and show you, here’s this, here’s the patterns that we’re seeing. And then you as the organization can say, okay, we’re going to choose the deployment pipeline. Let’s provide an MVP and let’s evaluate does

55:53 That really make sense for us in our organization? And if there are weaknesses there that you’re like, this is not good or the trade-offs are too great, then you can step back and say, okay, now let’s go try this more elaborate version of like, let’s roll all of our own stuff. Does that make sense? So, , you need consulting firms to show up and have this. This is why you call Tommy and I. This is why you talk to us, cuz we’ve seen we’ve been around long enough. We’ve been here since the beginning. We’ve seen the patterns start. We’ve seen the technology appear. And we have been able to identify trade-offs by choosing one

56:25 Direction or the other. Dude, I love it. Are we I feel a little better about this, but I still think that there needs to be more created around that like we’ll call it the fundamental guidelines. , the fundamentals I’d like that I wouldn’t say guidelines, but like guardrails. What guardrails do we need to adhere to? I like that. That’s a good term. That’s that’s a term used in the Microsoft adoption roadmap a lot saying you need to be able to give people a lane to drive in but you put guard rails

56:58 Up to say don’t go out of these boundaries. That’s that’s not what we’re going to do as an organization. So I think a lot of central BI team should be thinking about how many lanes and what is the messaging on the highway. That’s what we should be thinking about in central BI building process and educational patterns that we want to instill in our organization. And then from there you get really big value on the the the funnel of learning. Right. Right Tommy. Like I think I think that’s something that’s really important here. Especially when you start talking self-service. If you’re doing self-service anywhere in

57:30 Your organization, you need to figure out what’s the funnel of learning to get people from little to no knowledge down to we’re now inside the guardrails. We understand what to do and not to do in self-service. Oh yeah. And and I think the self self-s service is going to be interesting with fabric too because are we just talking reporting? Are we talking everything else in between? But I I think that for someone in this situation whether they’re it’s a large organization that they’re dealing with here. Oh, actually let me actually ask you one more question and I and then we’ll do

58:02 Closing thoughts. Sure. With an organization this large, I want to hear and we’ll see how much we align. How what’s your take on having a universal approach in terms of like hey if you’re going to do get and yada yada yada here’s the practice you need to you need to follow if you’re doing manage self-service o over here this is the practice you need to follow or is it much more organic like there’s the those fundamentals like very bare fundamentals but in terms of the process one

58:35 Sub sector of the company’s going to use compared to another even if they’re both doing it. How much should that be followed? This is interesting, Tommy. I I think I think there’s a I’ve been working on a lot of software development stuff recently, and I’ll maybe pull an analogy here that that makes sense. A lot of times when we start talking to customers or clients around software development projects, you first need to understand where where their mind is, what is their organizational structure, what what are they doing, what is the requirement, what are we trying to hit, what’s the

59:07 Milestone. So to me a lot of times I do a lot of work on the front end side of projects is what are your expectations? What are we trying to solve for? I want a lot of those clearly stated objectives at the front of things. Once we start when we clearly identify those main objectives what are we solving for? For example, our organization has a lot of Excel documents and we export a lot of our data out of business objects. We want to transition away from straight exports. And instead we want to start serving models through semantic models,

59:40 Pre-built reports, and allowing users to explore the data inside a fabric platform. That’s a statement that we can say like there’s a there’s a place where we were and we’re trying to shift to a new direction. Right? Clearly stating the objective of where we’re trying to get to. And you can then tie that back to again going back to like OKRs. We talked a lot about this with Seth back in the day which was what are the OKRs of the business, right? What what is a measure of success of doing a data platform project or any of these things? Well, the success here is speed to delivery of insights. more reports

60:14 Being used by central teams faster execution of changes to systems, right? Those could be your main objectives and all those can tie back to the reason we’re moving away from this old world is there’s too much stuff locked away on people’s computers and we’re trying to move to more of a centralized world and seeing what people actually interact with. So that shift is happening. Stating those facts up front helps I think inform a lot of the what do we do next? What’s what’s the next step here? So, I don’t know if I answered your question, Tommy, a little

60:46 Bit, but when I I spent a lot more time on the definition of what we’re trying to do because that informs a lot of how we do it. Yeah, I think this is where we’re always going to slightly differ where I’m always going to have a slight feeling towards not just what are we trying to do, but it’s looking at the the technology landscape and identifying like the resources involved. It’s identifying do we do we actually have that skill because a lot of times when you adopt or a new technology if they’re

61:19 All in a fabric I will make a good bet in saying that there’s a good poss there’s a good probability that some people don’t have the necessary experience or skill for using fabric or they’re not aware of what can be done right just because it was like hey we want to do this it’s one of those let’s understand the the ideal approach to using fabric what what if you want to do SQL databases in fabric okay what’s the ideal person look like what’s their skill set what’s their

61:52 Knowledge etc and then identifying okay is that even available it’s like actually my my best friend’s dad he was an engineer but when I started getting a data analyst he told me something that I’ve approached now to businesses as well he’s like anytime you start on one of your projects asked these three questions and I’m going to analogy this to people. He’s like, “What data do I need? What data do I have? And what do I need to do to get what I need?” And I’ve really actually used that approach every time. But I

62:26 Think for people saying they want to do notebooks, they want to do manage self-service is like okay, what do we need? What is the skills, experience, permissions, etc. there. What does the organization or that team actually have right now? So if we want to get there, do we need to give them I think to your point, do we need to give them knowledge that this is an approach, right? Because if no one knows about git, that doesn’t mean that we can’t use git or if no one knows about using the fabric API, that doesn’t mean we’re

62:58 Not can use it. it need to put on the table to say this can really speed up yacht x y and z rather than we just want to do our old process. So that’s the only place I would slightly differ. It’s interesting Tommy. So what you described to me is exactly how I’ve approached what my statement I think was earlier which is let’s go back to the the analogy of that we use a lot on the podcast. If this report if this process doesn’t make money or save money, why are we doing it? Right? So, I

63:32 Think there’s there’s a reporting piece of this that is if this process or thing that you’re doing doesn’t make you money or save you money, you should really evaluate seriously are we are we really requiring to do it. So, also I’ll note here as well and I’m just I’m using Comet the browser the new browser here which is cool. I’ve been googling some things here on the side here which has been super helpful. and it says what my question that came was what Git brand strategies minimizes production risk and I thought that question was really interesting because whenever we’re applying process or technology git

64:06 Process deployment pipelines workspace development all we’re doing is we’re trying to minimize risk to develop something I think is what I’m seeing here. So the reason you put git in place, the reason you do version tracking on individual files and roll things out is because you’re trying to minimize getting something from developers hands into a production environment with the least amount of risk. And this is a technique a a proven solution that is being from the software development world that’s being brought forward into the PowerBI data world. And

64:39 A lot of data developers have no concept of this. If you’re just in powerbi.com, you’re using SharePoint and versioning files by using final final final on the end of them or you’re using SharePoint to version files and deploy them strategically that way. Like there’s there’s very rudimentary functions that we’re doing to make sure that we have data being developed and data being able to produced with reliable outputs. So, , all this to say, Tommy, to go to go back to your question here, right? , I think what you’re describing to me is

65:11 Going back and reaffirming more of my point around we should not be scared about educating people. We also need to understand fabric or have people that understand fabric fully who help us understand the art of the possible. This is where I think consultants go. This is where you get online education. Go to learn.microsoft.com. Right? as central BI leaders or leaders of of people developing in fabric, you need to be educated or bring people around you who know what it does, how to make it work, what are the trade-offs. That’s the reason if you’re hiring consultants and

65:45 They’re not helping you avoid potholes in your process, why are you hiring them? You’re just now hiring hired work, right? So, in my opinion, when I look at a contractor, it’s not someone there just to be there as a warm body to help you get stuff out the door. It’s someone there to add guidance what’s been there before and help you build things or skills in your team. You’re bringing those people in to educate and develop talent on your team that you can reuse over and over again. That’s the real value of consultants. So, I’ll go back to your t your point, Tommy, right?

66:20 Why do this? What about the skills that don’t match in the team? Yes, it’s going to happen. You need a plan for it. But just be mindful that just because people don’t understand or haven’t been haven’t learned these processes in fabric doesn’t mean you should just throw out fabric and not go forward with it. You just need thinking about the training funnel on how to get people educated and up to speed. There’s a ton of really interesting trainings, educational tests, exams you can go get from Microsoft. It will encourage you to go learn things on microsoft fabric.com.

66:53 You can you can learn this stuff. And the neat part is, Tommy, we’re in a world right now where almost anything you need to learn about fabric already online, already in the internet, and someone’s probably made a video on it. Like, it’s all there. I’ll even point to Tommy, let’s learn fabric. You and I have done like 18 episodes, 12 episodes, I don’t remember how many it is. We’ve done 18 episodes around just getting your feet wet and working inside fabric. And Tommy, to be frankly honest, we should probably pull that up again and do some more. Let’s learn fabrics and building some more things cuz there’s a lot of new things in fabric

67:25 That have shown up recently. Let’s do more of those things. So, there’s probably a lot more we could potentially add here as well. And I run an entire class training.tips that helps people in the community get together, connect, and and learn about these different new features around fabric as well. So, that’s another area to get to get involved with as well. All right, Tommy, we’re definitely at time. We probably need to wrap here. Let’s wrap this up. Any final thoughts as we go forward here. No, I I think what our our final 20 minutes or 15 minutes is exactly where you need to have to approach where

67:57 You need to understand and you need to probably outline and deliver the multiple or choices on how we’re going to actually deploy this and come up with those limitations and what those pros and cons are. But yeah, that’s the the experience is going to be needed there. But there is not going to be a set standard way that’s always going to work. You’re always going to regardless of what process you choose how you do choose to organize or implement it’s always going to come with some hiccups.

68:29 What you need to understand is where do you want the organization to be? What is the standards that you want to implement and then letting the technology be part of that? So that’s all I got. The stuff is very new. Microsoft is building the stuff as we speak. We’re we’re we’re learning best practices as we go here. I think going back to the Microsoft adoption roadmap, fabric adoption roadmapback is the way to go. Figure out what your business needs. Align there’s different strategies you have. Evaluate whether or not you can learn that technology and then are you willing to support it. Right? If you

69:02 Pick a part of the technology that you’re not willing to support, I probably would argue it’s probably not the best solution for you. Once you have picked an alignment on which CI/CD pattern you’re going to use, is it bronze, silver, gold? What’s your lakehouse strategy? Document it out. We’ve done a number of projects with clients where we just sit down and talk about how we design your world so that it works for you. And once we do that, people get a much clearer picture around how to build their workspaces and their semantic models and orchestrate their environment so that it works for their business. Some businesses do a lot of

69:35 Dev prod. Some organizations do dev test prod. Others have test dev test QA and prod. All those things come with trade-offs. Each one of those processes have best practices in them. That’s where I would say you’d have to hire consultants. So this is a this is an evolving space in the BI world and I think it’s going to be a lot of new interesting technology that Microsoft has continue building that’s going to make this interesting and fun and will keep Tommy and I busy for the next hopefully 10 years working in the data space. It’s a very it’s been a very good

70:07 Career move for both of us. That being said, thank you very much for listening to the podcast. This is a little bit of a longer episode. Tommy and I got heated and a lot of concepts to unpack here. Great question. Thank you very much for submitting this question. person who submitted the question. We appreciate you and hopefully this gave you some food for thought. Reach out if you have more questions. With that being said, make sure you like and subscribe to the YouTube channel. That’s where a lot of the content comes out. That’s our main focus for how we distribute and put out content. That being said, Tommy, where else can you find the podcast? You can find us on Apple, Spotify, or

70:39 Wherever your podcast. Make sure to subscribe and leave a rating. It helps us out a ton. you have a question, idea, or topic that you want us to talk about in a future episode, head over to PowerBI.tipsodcast. Leave your name and a great question. And [clears throat] 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 next time.

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.