What is a Data Model? – Ep. 231

In Ep. 231, Mike, Tommy, and Seth unpack what a “data model” actually is and why it matters—from conceptual diagrams to logical fields and finally the physical star schema you build in Power BI. You’ll learn how clearer modeling conversations reduce DAX complexity, improve trust, and set teams up for scalable, reusable reporting.

News & Announcements

• Data Modeling for Mere Mortals (Part 1): What is Data Modeling? — A practical walkthrough of conceptual → logical → physical modeling that gives you a shared vocabulary for planning a semantic model with your stakeholders.

• PowerBI.tips Training — If you want to level up your modeling fundamentals (and not just memorize DAX), check out upcoming courses and workshops from the PowerBI.tips team.

• Tips+ Theme Gallery — A quick win: apply a consistent theme so the visual layer looks professional while you spend your energy on the model and measures.

• Explicit Measures Podcast — Subscribe and browse the full catalog so you can revisit episodes when a specific modeling or Fabric question comes up on a project.

• Mike Carlo on LinkedIn — Follow for Fabric experiments, modeling patterns, and new episode announcements.

• Tommy Puglia on LinkedIn — Follow for BI leadership, governance, and real-world adoption conversations.

Main Discussion

Topic: What a “data model” really means (conceptual → logical → physical)

People say “data model” constantly—but it can mean different things depending on whether you’re talking to the business, a data engineer, or a Power BI developer. In this episode, the team uses a solid modeling article as a guide to separate the idea of a model (shared understanding and trust) from the artifact you build (tables, relationships, and measures).

Key takeaways:

• Start with a conceptual model: identify the core business objects (customers, products, events, tickets, etc.) and how they relate—it’s the fastest way to align language and surface ambiguity.
• Move to a logical model: define the attributes and keys that represent those objects (and validate assumptions with stakeholders).
• Only then build the physical model: tables, relationships, and grain—the star schema (or variation) your semantic model depends on.
• A data model is a trust mechanism: it creates shared definitions so reports don’t collapse into “whose number is right?” debates.
• If you’re writing increasingly complex DAX to compensate for missing relationships or unclear grain, it’s often a signal the model needs work more than the measures do.
• Use the conceptual layer to spot friction early (many-to-many, slowly changing attributes, unclear ownership) before you sink time into ETL and report building.
• Don’t confuse storage with modeling: even with lakehouses and Fabric-style architectures, you still need a deliberate semantic model that matches how the business asks questions.
• Treat modeling as a data contract: agree on definitions, ownership, and change expectations so the model stays stable as the organization evolves.

Looking Forward

Try sketching a one-page conceptual model for your next dataset and use it to drive a 30-minute alignment meeting—you’ll save hours of downstream DAX and redesign.

Episode Transcript

0:00 foreign [Music]

0:32 good morning everyone welcome back to the explicit measures podcast with Tommy Seth and Mike good morning gentlemen and a happy Tuesday to you it feels better today but it is Tuesday it feels like a Friday but it is a Tuesday oh boy great oh boy the workload is intense I think Mike I asked you yesterday I’m like what day is it today I’m like because we we’ve done a few recordings like it feels like Thursday yeah I’m having a hard time keeping

1:03 yeah I’m having a hard time keeping track of what day it is yes you’ve basically been looking at me going Tommy you need sleep Tommy Tommy needed some sleep Tommy’s I can tell how much sleep Tony has been getting by the the status of the five o’clock shadow if the if the shadow gets if the shadow gets darker Tommy has been working really hard and not not sleeping as much is whether or not I’m at least willing to get the hair slightly wet before the podcast basically exactly right exactly right right so a couple quick openers or

1:33 so a couple quick openers or things we’ve found across the internet so Tommy and I have been doing an exploration in fabric let’s call it let’s we’re doing learning fabric series and we’ve been having a lot of fun actually I’ve been really enjoying working with fabric I like working with the preview I’ve been trying out features trying to figure out how what workflows would work for me and just explaining how to do very small pieces of pieces of data engineering I guess around fabric for now so in lieu of yesterday’s video we made a video around hey we’re

2:03 video we made a video around hey we’re gonna go pick a file from the lake house and we’re gonna transform it and we’re gonna drop it back in the lake house and when we did the video or as we were doing the video there’s a lot of questions that came out we hear like why does this when you create a data flow you actually get three new objects called data flow staging something right created by the user who’s creating the data flow and we’re like why are these three objects here what does this mean what is what is occurring when I’m using the mashup engine or the the power query engine to do certain things on top of the lake house

2:34 things on top of the lake house well within literally we asked the questions during the episode and we must have missed the blog but the same day somewhere throughout the day Miguel Escobar which amazing developer on the Microsoft product team dropped a great article and literally answered all of our questions so I will throw out the this article here from Microsoft because it really answers a lot of our questions we were having yesterday around just data data engineering and what’s going on with

3:04 engineering and what’s going on with dataflow’s Gen 2. it’s just very different than what we’ve thought about previously yeah and the best part too is Mike and I have been just talking about the the direction of the demos like well should we actually do a different example for data flows rather than pulling a file out and as I as I tend to overthink initially with things things like well what’s the best use case for dataflow and a lot of this was going with our previous understanding of how data flows worked right yeah so and if I think Michael’s

3:35 right yeah so and if I think Michael’s like well let me answer where dataflows fit in and it’s it’s really actually crazy how data flows actually work on the back end here and there’s a lot of even though it’s the same UI yep all the the naming is the same some real huge important Concepts that are different but which is hilarious even in gen 1 this was Microsoft’s like best practices for Enterprise teams using data flows yes so one of the big ones and yeah so

4:06 yes so one of the big ones and yeah so what you’re speaking to is when I create a data flow the reason why there’s always those three additional artifacts that are also created is because how it’s going to process it previously with power via dataflows like they would talk about you load everything into like a bronze data flow and then if I want to do Transformations I would connect those in the same workspace and that would be the linked a computed entity basically everything gets loaded in from the source system and then the Transformations would happen

4:37 happen but with data Flows In Gen 2 that can all happen within the same data flow so already a really big proper ETL pipeline yeah true so all in this yeah building multiple ETL pipelines that have to build upon one another yes and exactly and I had a creepy suspicion that yes there’s data flows and yes there’s power query in the back end here but I had a creepy suspicion that even dataflow’s gen 1 is using some shared SQL

5:09 using some shared SQL serverless machine that’s helping you make the data flow run it would be my opinion and I think this data flows Gen 2 just basically shows a bit more behind the curtain what’s actually happening and this this thing called the enhanced compute engine I don’t think there’s anything fancy I think it’s just SQL Server actually I think it’s just a serverless SQL Server that’s getting spun up that’s helping you do these Transformations with a compute engine in in real time as you’re doing these loads of the data and then it’s going and then it’s

5:39 then it’s going and then it’s spooling down and doing something else I think what you’re seeing here is a bit more of behind the curtains of what was already happening but you actually get objects now that yeah again from our understanding you shouldn’t touch them don’t delete them we did that it broke figs so don’t delete that don’t delete the objects don’t remove them when you create a data flow it will create some additional objects and yeah we learned it well at least from a Content perspective being being out early also includes the often because yes initial Impressions or the

6:09 because yes initial Impressions or the way things work and we’ve outlined this in previous podcasts yes hey hey we’re talking about this thing and the next time we talk about it there’s probably going to be something new well it’s an iterative thing yes agreed here’s your mind shift though with the data flow is that to me even when book The the mic dropped anytime you’ve ever built in power query what’s been one of the best practices the Cardinal things only load in what you need that’s not really to like do the check mark to say enable load or not because if you

6:39 to say enable load or not because if you load it in it’s going to go into your model which will only do what you need and the data flow is Gen 2 that option which should not be called that anymore is actually what enables the computed entity to basically be enabled since the data flow gen 2s are really meant to go to a destination source and again that’s very different from how we add data flows before before whatever I would load in would be available so to speak like this like a SQL table but now if whenever I disable or

7:10 but now if whenever I disable or enable that load that’s basically adhering to how our power bi is going to sense process is it going to use the computed entity it’s going is it going to Stage it in that data low because again it doesn’t matter if everything’s enabled because it’s only about what you’re actually getting pushed on so huge shift in thinking yeah check out the article I think it’s very relevant if you want to know more details around what data flows Gen 2 is doing the articles in the chat window and we’ll

7:40 articles in the chat window and we’ll make sure we put it in the description of the video as well another article that’s coming out Tommy you’ve got one here from what can AI do with a toolbox I would we probably need to make a new nickname for Tommy he’s probably going to have to be the prompt master or something like that you prompt master or something like that something along those lines because know something along those lines because he’s always telling Seth and I is like check out this new prompt that I made around this really amazing thing there working with data asking questions of the business users developing Dax equations or whatever Tommy’s always

8:11 Dax equations or whatever Tommy’s always got prompt things and I think this article you found Tommy yeah so the we’ve talked about this before about like Microsoft’s been pushing you’re like oh yeah Power Wheel is coming with co-pilot and but with fabric and everything going on it’s like who’s going to be on that we’ve talked about the users right recently open AI release what they’re calling the code interpreter so like oh it’s for developers but basically that’s not it at all it’s just basically now Everything Chad TPT is doing it’s

8:41 Everything Chad TPT is doing it’s writing it’s Python and then giving you an intro that way so basically rather the the whole end of the article here is the major Focus here where the author here I think his name is Ethan basically just goes through and it’s just ask you very basic questions but because chat DPT is basically going to write the python itself and then give an answer an answer he basically goes through how and it’s a very brief example he uploaded two files and he basically asked chat

9:13 two files and he basically asked chat TBT to do some merging Transformations and it understood itself like hey these two column names were different in each one so let me merge it this way and he’ll show you all the python code on how he’s doing the Transformations so incredible is like when you think about how it’s actually it’s even updating itself in real time going no this is not going to work let me try it this way then he even goes into not just the ETL of what now AI can do or at least chat EPT but then just allow analysis where rather

9:44 then just allow analysis where rather than giving this very detailed prompt to say I want to see these columns with this he’s like show me some insights

9:51 this he’s like show me some insights like show me what’s what’s good here and if you go through the article it’s insane and insane just with his questions asked and then what basically AI was basically coming out with and we think about what we do again this this to me is the closest we’re getting or we’ve got so far to how that’s going to affect us I think the interesting one about this is it’s it’s not just analysis right it’s it’s running the data through

10:21 it’s it’s running the data through predictive models it’s doing data science activities it’s writing the code it’s and what he was describing is it was correcting itself in column naming in in the day in because the python code it was generating was it was Auto fixing its errors and looking along so I think he comes to the same conclusion that we’ve talked about but but it kind that we’ve talked about but but it broadens of broadens I I think the spectrum of of how in

10:52 I I think the spectrum of of how in where and how AI is going to impact jobs right because one of the things he he points out here is it did something in minutes that took would take him two weeks to do or like took forever to learn in his PhD program around the statistical model and if if you we think about like oh coding like SQL and whatnot but like the power here is like is it the fringes right is it these really long degrees in math and

11:24 these really long degrees in math and science and like that create really complex models and diagrams that have to like pull in data making this hypothesis and then you get an output where AI is going to be actually most useful and dangerous right because you can’t do this like our other points you need to have people that are in there understanding all of those things to say yeah this is actually a sound yes model this is actually a valid output right

11:54 this is actually a valid output right but the speed at which you’re able to understand whether or not that’s a valid model and have all the reasoning behind it and just puts you light years ahead of where you were and it’s the speed to Market or the speed to Insight that AI is just gonna I think still blow our minds with right like we know it’s here we know it’s coming but like when it gets implemented I think he was just like oh my oh my goodness right like you can run through so much more so much

12:24 can run through so much more so much faster than before and that’s that’s crazy I think I think the guard last Point Guard the guard rails are what’s going to be important here right yes very exciting all of this is going to change or has the potential to change a lot of what we do but just because you can type in a prompt and say hey write me an end-to-end ETL and da and run this through statistical model and I need this and this and this and you you can you cannot take that and

12:56 and you you can you cannot take that and run with it and I think that’s what some people are going to do they’re just gonna be like yep I did this thing in in upon review it’s gonna be like no you didn’t because if you did then you would have caught this this this this yes all this is an error this whole output is not good and that thing I think I think the big takeaway for me a lot of this AI stuff one is you better learn how to start prompting things right figure out what prompting looks like I am continually using Bing chat more

13:26 continually using Bing chat more frequently than I’m using Google anymore because not only does it give me a very ability to ask a better question it gives me a faster answer by looking at the pages where the answer is coming from and summarizing details the other day I was trying to look up a a python statement that I needed well I went to Google I asked a question and it gave me the articles I needed but I had I had to go to the page and then sift through the article and get to the point in the article that told me the answer that I needed so it was like a couple extra clicks and some

13:57 was like a couple extra clicks and some scrolling well I went to Bing chat and it just asked the question and boom there’s the answer it literally gave me here’s the code you need to write and it came from this source and it also references it on this source and this source so it’s it’s like some of this how do I know it to me it’s then that’s one one one is prompting the second thing is discernment right when I see these answers are coming out of the AI generated thing what is the AI doing to to confirm this answer to me to confirm the solution hey this is the

14:28 confirm the solution hey this is the line of code that you would need to write and I found it and three examples on these three websites okay that seems more credible than just I went to one website on one thing using a Google search to get there so I think to me it’s going to start evolving into this and the discernment of the answers the discernment of the output is going to be very interesting as far as this relates to power bi I’m very interested to seeing where AI can be applied in the python Library area so perfect yeah so when you get to notebooks inside fabric how will this help me build again I I want to learn

15:00 help me build again I I want to learn more python I want to be better writing at python statements I just need someone to guide me I need I really need like the co-pilot to come alongside and say okay here’s what you wrote If you wrote this statement this way it would say if you compute time or why not why not in Dax give me a you why not why not in Dax give me a the Italian o pilot that that says know the Italian o pilot that that says hey your context transitioning is happening here did this like here’s hey you wrote this it should be able to analyze your statement and say hey here’s three more Dax statements that would accomplish the same thing you

15:31 that would accomplish the same thing you may want to try these and they may have faster Performance Based on this this and this that’s the stuff that we really want because I still need to write code I just need to supplement the code writing with something that’s a bit more knowledgeable or taking that that Library of the internet has the answer it’s somewhere out there I just can’t get my hands on all the time yeah when you say like like in Italian are you talking like Marco or Tommy because oh yeah

16:01 here’s the Dax thing if it was Tommy like every 15 minutes it’d be like are you sure you don’t want to break maybe you need some spaghetti hey I think you should add a cannoli to this statement like halfway through like what are we talking about I need some pasta but we we are honestly Mike we’re not far off now when you think about the examples that were shown here that’s now already available and being able to connect to the your fabric workspace just like hey I have this I do want to do these Transformations yes one really quick thing from the

16:31 one really quick thing from the article I don’t know if you caught it but it’s like called a ghosting moment with gbt and where we’re getting to to just we’ll sum it up this way this new feature he basically just asked the basic question said hey with all the tools you have create a new meme that your experience working with humans on a day-to-day basis and he’s like okay and so by itself that’s all he asks and by itself gbt came up he said I work with humans and it’s a wide range

17:02 work with humans and it’s a wide range of questions there’s four images that generated all in Python the first one set is the computer says I’m an AI developed by open AI ask me anything next one is a person little stick person what’s the weather today the next image is the AI says I’m sorry I don’t have internet access to answer real-time questions and the human says so how do about open AI but that he even says to itself the humor comes from the irony that y can provide a wide range of questions I’m able unable to answer real-time

17:33 able unable to answer real-time questions do my internet restrictions it’s wild it’s wild it’s picking up weird stuff yeah yeah it’s it’s interesting because it’s picking up the nuances of the of the language and the and the the nuances of joking things yeah but guys please check out the end of that article it is we are getting close to just how important it’s going to be on our day-to-day to plug into it I think is it and and this is what you keep referencing it’s the prompting it’s it’s

18:04 referencing it’s the prompting it’s it’s get familiar yes understand that this is here yep and the the more the quicker you understand the capabilities of what is here now and coming and this article is like the next version right yeah the next thing I think he got an early preview yeah is the these are accelerators for you huge oh my God right like you need you need to get on board you need to like Leverage them and think about how you might leverage them in your your day-to-day agreed all right great conversation fun stuff there I’m excited to see where

18:35 stuff there I’m excited to see where AI takes itself in the next two years I can’t imagine what it’s gonna look like in three years from now or a year from now even it’s going to be so different so it’ll be very interesting to keep tabs on that one moving forward with some deeper thoughts here this is a bit more of a conceptual I guess discussion today the article for today is from data Mozart Mozart website it’s Nicola has a really good job does a great job teaching people how to do the the DP 300 DLP 300 I think it’s the test for

19:06 300 DLP 300 I think it’s the test for Microsoft power bi essentially it’s the data engineering test so it does some training around that but has been doing a number of years of Consulting and this blog post is really good and I was curious and he’s like Microsoft certified trainer MV and nvp yeah correct sorry I messed all the things so incredibly smart gentlemen really like his stuff very what the thought on this process is just very well thought out and I thought this was a great exercise and as I was building some new data models I was

19:37 building some new data models I was really going through and thinking yeah this this is a this is a valuable exercise to go through with

19:43 exercise to go through with anytime you’re building data models to really understand the different types of data models you should be building and it’s the same data model but it’s different levels of depth on what the data is doing together and I think this dovetails into a lot of conversations when working with teams around do you unders like how how does the data engineer or data analyst how can you understand what the business is talking about in their data what are the different things that interact together and how does that data interact as a

20:14 and how does that data interact as a model so I’ll drop the article here in the chat window any initial thoughts on the article maybe we should summarize briefly about the article I think I think perfect timing for me yeah I was I was discussing data modeling like the importance of a new data model different Paradigm different way like we have we have a base now we need to build this new data model in Long conversations yesterday even good and as I was thinking about it tonight or like last

20:44 thinking about it tonight or like last night like how do I describe this body of work yes the business right because it’s when you say hey it’s hard for people to understand like we we’re already using this data we already use this data in different ways yes but where we want to take it is the next level next steps and I’m like how do I describe that body of work so like coming into this morning I’m like man I love love it’s the way this is it outlined this yes process by which you can you break this down

21:14 can you break this down the differences between the types of models that you talk about and and ultimately there’s some key phrasing he using uses in here but I love love the article spot on and it’s a great great one for being able to articulate the components and what you mean when we when we and I’m I’m a huge I’m the individual that you over uses not overuses but you say data model A Lot yes I do yes this I think is a

21:45 Lot yes I do yes this I think is a valuable exploration not just to help the audience here articulate but anyone who’s listening is just to go here check it out all of a sudden you’re like oh that’s what all these things are and it locks locks those into place and will ease your conversations with the business for sure and one thing I’ll let’s you and one thing I’ll let’s talk about the article so know talk about the article so this Con we say in the podcast we say probably the buzzword all the time data model and then there’s basically two things that are going along with this one and I think I’ve had it made a distinction in my mind a data model is

22:16 distinction in my mind a data model is more of a concept and then the data set comes from powerbi. com so the data set is the representation of what a data model is doing and so this article talks about three like the three stages of a data model it talks about the conceptual model what are the core pieces of data what are the physical things right for example in the article it talks about going to a concert everyone’s been to a concert or an event at some point in time there’s there’s

22:46 at some point in time there’s there’s the attendee of the event right the person who’s attending an event May buy some food therefore you now become a customer of the event the event is running in a park a stadium there’s there’s a venue that applies to the event and then in order to get into the event you need to have a ticket so these are all like physical things but they all interact together to represent the experience of going to a concert so you start off with this high level conceptual model that’s your first

23:17 that’s your first point of contact in the article what are the what are the main objects and how do they interact together and then from the conceptual model you work on okay let’s talk about what is The Logical data model you start saying okay what data points do I have related to or what are the columns that represent customer information do we have do we have a primary key do we have their name their address things about the customer what is on the ticket did they buy a certain type of ticket what does that look like so there’s all these details that support these different

23:47 details that support these different objects and how they interact together and then finally you get all the way down to the bottom where you start saying I’m going to build the physical model where I’m literally creating like the power bi data model with columns with tables and we have various objects that are actually linking things together right the stadium needs to have an ID column that links to an event right the event occurs at the stadium ID so what does that look like and I think to me this really helps explain

24:17 to me this really helps explain how in my mind I’ve been doing this all along along but now seeing it represented in a very clearly articulated way I thought this was just incredible so that was my impression initially yeah like when you hit the article over or you put it in our in our topic board I’m like Mike I think we’re a little past what a data model is at least that’s what the the topic was but yeah no when you actually this goes perfect too honestly with conversations we’ve had around fabric on like the lake house and how much table or data we’re adding into

24:47 how much table or data we’re adding into one lake house but I think even more I feel like we just had this conversation the other day there was yesterday but the idea though to now we’re like we it’s more the this conceptual data model it’s bringing in the business and this also an investment for the business of building the data model and I love that that sense concept there where think about when we what data models have meant to us for the last seven years like

25:18 for the last seven years like well it’s a power band model we build that then we try to then frame it around with the business seeds then we’re asking the questions that’s true right yes so yeah and I think if you go back to like the conceptual model I think is a lot of again as a data developer right I did engineer data analyst whatever you want to call it right as a data developer I think a lot of the times we miss the boat on the conceptual model yeah you can literally step back and say let’s look at our entire business like I run a a lemon you

25:51 I run a a lemon you I run a a lemon what is that a lemonade stand right know what is that a lemonade stand right I run a lemonade stand let’s just step back and say what are the things that are important to us what do we care about as a business any business can sit back and say let’s talk conceptually around around how we do business what are the artifacts what are the different objects that we care about and how do all those relate once you start with that conceptual model I think a lot of other things become clearer and then when you start talking about these other details details there’s always

26:21 there’s always challenges when you start talking between okay well how does a customer relate to our products what does that look like okay we have another example here we have seven different channels by which someone can buy something from our company right we could they could buy our product through Amazon they could buy our product at Walmart they could buy it Direct ship from our website like there’s there may be multi-channels of sales that are going on at the same so how how do you articulate that complication well the conceptual model allows you to

26:51 well the conceptual model allows you to pinpoint areas of friction or where you need to spend extra time designing a data model because there is more complex interaction between the different data points well and it’s that it’s in the importance between that and again the having the business more involved we’ve talked about the idea of this data contract and I think this is a big part of that where hey I can only do so much I can build a data model for you and I can make it work run like a Lamborghini but if we want it accurate

27:22 Lamborghini but if we want it accurate if we wanted to be able to be flexible you have to tell me business what are the like how does everything work and you need to tell me this from a business point of view what are those little those trigger points what are the the thresholds here on how we’re getting the information in what’s important again not the technical side but what are those important elements so again it becomes we’re speaking the same language language from the outset I think I think the two things that strike me right off the top

27:54 strike me right off the top that I 100 appreciate is he he lays out he lays out the main reasons why you you go about this process and and I think one of the I’m not gonna say us as bi developers are hacks but like there are there are we we cover the spectrum of engineering data engineering ETL models reporting like they’re we cover a lot of ground yes but there are there are roles there are larger companies people who do data modeling data Architects right like you could argue

28:24 Architects right like you could argue are two different roles or the same as well but ultimately he he starts from the right premise which is his point is that I agree with 100 is data models are about creating trust a shared understanding between the business and data professionals the goal is to increase business value with data start there start yeah right like this is everyone’s gonna agree on that this is why we’re gonna go down this this route and I’ve found myself like in in situations where somewhere we’re talking

28:55 situations where somewhere we’re talking a lot about data modeling because the points here are just because you build a pipeline of data and do some transforms and you now have a table of information that you might be using does not mean you have a data model you’ve ingested some data and you’re getting some value out of it I love the that’s very true analogy he uses here with the house and this is I’m I’m straight up stealing this conversation point so so thank you Nicola because because this is this is exactly

29:25 exactly what you need to engage in right away with the business in the conceptual model and a lot of the ideas that I’ve been articulate like obtusely

29:34 been articulate like obtusely describing I’m gonna take away today and go straight to building a conceptual model that I can have conversations with the business because especially with data and all of us anybody listening knows this there is so much nuance and when you try to elevate or build a data model to move you in different directions to do things above and beyond just wrote data coming in there are a lot of challenges that we need guidance on we need under we need

30:04 need guidance on we need under we need to understand what does the business want to do with things and a lot of times business confuses that with them telling us what to do that’s not the case and that’s why I love this house example because building an ETO like to ingest some data and make it useful to the organization is like your foundation if if you’re the the build like if you’re the owner and I’m the builder of the house you absolutely care where rooms go you care what goes in the room if you care what like how things look

30:35 if you care what like how things look and how they’re structured and throughout the article as he lays out these different types of models I think he keys in on that important aspect of if you don’t pay attention and do this modeling that is extremely important from a conceptual logical and Physical Realm and go implement with like the investment as an organization can you whip things together without that you betcha but if

31:05 together without that you betcha but if the owner comes and he all of a sudden sees you put the bathroom on the wrong side of the house how long do you think it’s going to take you after the fact to tear it down rebuild it over on the other side of the house as opposed to doing it the first time the right way and that’s the argument here right that I love in detail that will walk through in these different types because it’s very specific of a lot why you do these modeling but it shapes the conversation yes and puts it in the platform of like yeah you can you can you can skip this

31:35 yeah you can you can you can skip this but this goes back to our point we were talking about something related but different where I was just like yeah that’s a black oh maybe this was the where business starts to choose to have really important information only in the reporting layer that isn’t part of your Source system right you can build this thing sure but if they come back and they’re like well now we want to tweak it right well in in this analogy I I finished your bathroom 100

32:06 analogy I I finished your bathroom 100 and now you want black tile instead of white tile and every wall and every surface is tile well I can do that but I can either spray paint it or I gotta start over right right and and this is where I I just I’m so keyed in love the analogy he uses and I think it’s the framework for that Bedrock conversation because these aren’t simple things this isn’t something you whip out the door but if you have that like proper process conversation there are absolutely ways

32:37 conversation there are absolutely ways you can incrementally iterate towards Solutions but it’s it’s all built on the right foundation and the right architecture yeah when when we say trust I think I’m just going back to it’s just aligning expectations from the onset of not just like what we’re gonna what the end result’s going to be but what what is this supposed to do what are we actually and again what is the actual concept here right like what is all the logic what’s important to me from the business

33:07 what’s important to me from the business all that being said though this does not require more time or should require more time from the business to be even more involved because you think about everything up until now it’s build me a report and I’ll give you some brief information you should know where our sales data is and then there’s all the nuances there that we may miss from the onset and then we’ll again you’re like to your point Seth we’re building that in power bi some convoluted way and there may be a calculated column somewhere because we

33:37 calculated column somewhere because we have to adhere to that but if we’re going to do things the right way every time right then this does require a better conversation with the business and more time spent from them with more of their input too one thing I’ll add here from my experience on this one has been the more when you spend time to build out the conceptual model all the way down to a physical model when you start with conceptual things first you start identifying earlier in the

34:07 you start identifying earlier in the process a larger data model that can serve the needs of more reports because you start thinking about all the other data elements that you may need to bring together now you may not have the data to support all the different objects but at least what data links together and and what that looks like and as you as you go from the conceptual piece into what the data model is doing there is some really relevant you have to decide where the relationships live inside your data

34:37 relationships live inside your data model and I think this is a very good method by which you can say we do need to understand the interaction between customers and sales how do these two Separate Tables relate together and then you can really have these conversations around okay when we’re talking about information about a customer what part of that data is a factual piece of information or a dimensional piece of data because that again in the chat someone was talking about a lot of times people are trying to solve very complex

35:08 people are trying to solve very complex data challenges with more complex Dax which in reality boils down to typically your data model is just wrong like the way you shaped the data is not correct to support the analysis that you want so if you step back and say okay let’s let’s re-look at what the data model is doing and pull back a bit I think that really helps you design and build better data models so I I really think that this is a if I’m teaching star schemas if I’m teaching individuals

35:39 star schemas if I’m teaching individuals how to data model to provide data modeling capabilities to your team I think I’m starting with this I think I’m really starting with like hey and go ahead yeah yeah this the conceptual data model approach simplifies what we talk about all the time which is lead with a visual right if we’re having a conversation about complex topics within the business right like having a conceptual model which is just high level we’re talking customers how does

36:09 level we’re talking customers how does this relate to something right that can elicit so many different conversations within a business or a particular business unit you’re saying hey what we’re doing here is we’re trying to understand and figure out like how do we get from point A to point B and this is what this looks like from our high level and that alone I think elicits a lot of conversation that would would help you drive for building a better model now like one of the the nuances I want to make here is like we could be talking

36:40 make here is like we could be talking power bi model like a tabular model or you could be talking about a a data model in the back end that’s going to serve much more awareness that live in the warehouse yeah this is where having that larger conversation it like builds this ecosystem point of objects that that could be required for your business to move forward yes but does that mean that you need to build the whole thing before you get any value and no

37:10 and no but understanding all of those pieces absolutely is a foundation for moving forward because otherwise it’d be like well we got the basement figured out so the main level in the the roof that’ll all figure itself out as we go I want to build this side of the house and not care about what the other side it looks like so you’re gonna build three levels and not connect it to anything like we could do that but well and and I think too there’s a big part there where as you go through that

37:40 part there where as you go through that exercise you can identify too where the where the flags are where the potential barriers are and again that’s also the expectations and with the business where it’s like I really know you love you love the beach you want to build a beach house but there’s alligators based on what you’re telling me there’s alligators out there so let’s talk about this and I don’t like oh did alligators can climb fences by the way I did not but oh boy one more reason not to live in a foreign foreign but identifying like I know this is what

38:11 but identifying like I know this is what you’re trying to do but based on what you’re telling me right like there’s there’s some things we got to figure out if you if we’re going to get the end result be too often we realize when we’re trying to do the visuals oh it’s not going to work that way because all the way Upstream there’s this additional logic or how the again the way the business thinks yes and I think I think there’s also flexibility here so Donald you make a really good point in the chat here Donald’s talking about you don’t Kim so Kimball kind you don’t Kim so Kimball like the a stake in the data modeling

38:44 of like the a stake in the data modeling world right and the Kimball modeling system is how you would model data together kibble says you should build and model around a business process and not a particular report and I think a lot of times we get tripped up around thinking I need this report with this data on it and you and we focus a bit too much on the report level and then we don’t ever come back up to what is the conceptual model how does all the data interact together and when we don’t focus on the business process and I think

39:14 and I think this also alludes to when we talk about this all the time on the on the podcast which is which is we need to take action on the data just presenting you with information is helpful but I think the enrichment of

39:26 helpful but I think the enrichment of that data or being able to look at something walk away and take action on that or know what you need to do to fix the numbers basically right because the the monitoring the reporting this is all it’s all a behavior modificator right it’s it’s there to support your actions to do something different that changes the direction of that line chart that bar chart whatever you’re you you’re looking to get to you’re looking to get to a goal and the numbers are trying to help you track to that end goal so identifying more clearly what is the

39:58 identifying more clearly what is the business process that you’re doing and aligning that to your main goals and objectives and I think as I as I get into this sometimes we don’t we so I think sometimes businesses lose or or over complicate the business process there’s too much in the business process which over complicates it and therefore makes it a lot harder to use the operational system to collect the data to to write the reports and

40:28 and yeah anyway I’ll have another point but I’ll leave that for there now it is yeah business process on top of data the I thought I thought you were going I thought you were going to like and correct me if I’m wrong like in the business process conversation like yeah levels at first I thought we were talking about like levels of scalability right like if you think about an area of business that needs a report are you gonna give them that specific report or are you looking at that business area and what they entail and you build a model and then you can build many

40:59 model and then you can build many reports off that right so that’s where I was going yes okay yeah so because if you model the data in a prop like as the process is running for the business process right that’s the goal I business process right that’s the goal the right the goal is here here’s a mean the right the goal is here here’s a data model that can serve your insightful needs build what you want to build the report level I’ll I’ll teach you how to build your own reports but here’s here’s how your model here’s how your data interacts together right these things are related these tables are have these relationships here’s the calculations you can use when you define those things now you teach

41:30 you define those things now you teach that business unit to rely on that data model and that becomes their source of Truth for whatever they need to do so now you’re you’re spending less time building reports and you’re doing more time data modeling well and also would cause to go farther Upstream too with any Transformations we’re going to do because if we know certain Concepts or trigger points we can we can then kind trigger points we can we can then integrate that part of the Upstream of integrate that part of the Upstream detail process like well they never include this or that we always have to account for X Y and Z or whatever the

42:02 account for X Y and Z or whatever the whatever the framework or the scenario is which there’s again always nuances and we can tackle that farther upstream and part of our own process I I do want to ask you guys something though because as I’m going through this and I’m thinking about how does this actually work in a real world setting so to speak right because I immediately go to okay are we doing this for every Power bi data model when does this conversation occur and then basically like how much investment do you need from the business

42:34 investment do you need from the business with you what’s acceptable in terms of the amount of time and what are we trying like I guess I’m thinking what is the actual output that we’re expecting from the business and then are they willing to do that I was just writing that note right we talk we talk we’ve talked a lot about like what is our report build process and asking questions to the business is one thing but I think this I think this structure of even on a simple level

43:05 structure of even on a simple level building a conceptual logical physical data model like it it seems like a perfect fit for talking data behind a report right because if I’m if you’re the business user and sometimes this could be a super simple conversation but more often than not it’s not right how do they understand the business and if you’re breaking things down from a like even from the reverse right like you even from the reverse right like where the data is you’re going to know where the data is you’re going to break it you’re going to build this facts and dimensions this is you’ve already got your physical model like

43:35 already got your physical model like figured out like layering that up to the high level conceptual hey business user let me let me describe for you like how I pull your data together right this is what we’re going to do on a high level here here like having that conversation with the visual part to it I think creates a really firm firm foundation for you to move forward and make sure that you’re all on the same page with the data right whereas before I think this was something we we’ve talked about where you would just suss out with

44:06 where you would just suss out with questions right and and I think you do that more so in visuals but I I love this example because it gives structure just enough in many different scenarios I could I could easily simplify this for a side conversation of a report or it could be much more complex or it could lead to me like misunderstanding a key element of what their area of business is and and therefore like I this is something I’m going to introduce as a structural piece in in that report build process three

44:37 process three I’d also like to note here as well as you think about this one I I’ve had this nagging thought in the back of my head for a portion of this we’re talking about does every single data model get this does this happen every single time I think this is I think you’re you’re made a really good point there Seth around like some this could be a lightweight version of this right you could like distill this down to a simpler version but you could also round this up to a bigger investment of time and say if we’re making certified data sets that

45:07 we’re making certified data sets that live within the organization right I think for sure there’s more rigor the more important those data models become right so for talking in larger data models that are going to serve many reports if it’s coming out of the central bi team you’re spending more time and rigor around around doing this type of activity really documenting what this is and I would also argue once you’re done with this we’ll put this back on your center of excellence like there’s going to be in your center of excellence or your community practice page right here’s the data model that we built in this data

45:38 data model that we built in this data model here’s the logical model that we’re going after I think a lot of this to build trust is to communicate back the output of these deliverables in a way that people can read understand comprehend here’s why we’re doing things the way we are because I can’t tell you how many times people ask things of a report that it’s not designed to do and I think this is another way of saying this report has boundaries on what it can and cannot do or this data model has boundaries of what it can do here are the boundaries so Raphael makes a crack in here in the

46:10 so Raphael makes a crack in here in the comments so you have a few full beautiful CI CD and you get an email like I need this in this column now because board meeting whatever and your release cycle is monthly releases like to me this conversation about modeling is a a relevant one on a report by report basis it’s it’s these structural pieces are good conversation points from there we’ll talk about logical data models and like the the elements of that conversation that bring forward challenge points to make sure we’re on the right track like there is and and

46:41 the right track like there is and and then like in the context here are we talking about building Enterprise data models and houses and in just like investments in time yes right I think organizations need a balance of both one I don’t think a team or like this thing is a blocker for business to get what it it needs I think it should have data and the tables available to it to generate their own reporting and meet the needs of every day being a fire drill for a business user right there’s a balance in here where you have

47:13 a balance in here where you have the organizational development and movement in a direction that is going to serve the needs of the Enterprise holistically as well as a self-serve method methodology where people are getting their answers as they go along with an understanding that it doesn’t come with the same type of slas right and as you and we talk a lot about like if that ecosystem is working well there’s this graduation path like how do you build build or move things that

47:44 you build build or move things that have gotten a lot of adoption into the Enterprise realm or how do you substitute structures and things that have been built now that have slas that have all that certification behind them into other reporting and we talk a lot about that a lot so I don’t think I just want to iterate this conversation is not about like hey it’s going to go in its own whole and go build its own thing without any input from the business or whatever business from access to data it’s not that well see I took it the

48:15 that well see I took it the scalability side of it Seth we’re thinking about where does it applied your house that you build is only going to be as good as the neighborhood Who in a sense around it so the framework is so important too where no data model exists on its own siled if you did this whole practice and everything on the outside was completely unstructured it’s not gonna have a lot of success so if the same idea of building a neighborhood right it’s like do we have the election electrical

48:46 do we have the election electrical and the plumbing that’s already can work so if you need to build another house a good example is like marketing data if you’ve ever worked with a marketing team they’re like hey we need a report and you’re like okay what are you doing now it’s like well we have 17 sources that we’ve never had to talk to each other and we basically look at all these things and yes we just want you to combine all this it’s like hold on it’s like and and to that point though there is in someone’s mind there’s some kind is in someone’s mind there’s some weird like in in the marketing data of weird like in in the marketing data piece of things there’s a lot of data coming in that doesn’t necessarily join

49:16 coming in that doesn’t necessarily join together right yeah it’s a good it’s a good segue Into The Logical data model yeah so yeah exactly that the concepts here right even if even after you get approval business is like Yep looks good look yep making relationship from customers to tickets sounds good okay we’re gonna invest some more time we’re going to dig into the detail the details right and that that’s where you start looking at keys and relationships and fine-tuning those like high level relationships in that logical

49:46 high level relationships in that logical model and it’s to to Nicholas point it stress tests the concepts this is exactly where you start running into those problems be like okay yes I understand one of the business objectives is this we don’t have that data there or all the data in this Source system is only partially completed if we’re going to rely on these this relationship you guys need to go fix all this and we’re going to do that before we get down this road because what we’re building here is a blueprint for us to actually go build something and that

50:17 actually go build something and that that logical data model I think addresses a lot of those those or challenges that need to be discussed need to be overcome and then also shifts into what is the business going to do what is the business responsibilities and actions to make all this work we’re going to go build the physical while you’re fixing these things and there’s going to be this symbiotic relationship about us moving towards a Direction that’s going to serve our organization long term long term and then you have the side side stuff we’re obviously like you’re supporting

50:47 we’re obviously like you’re supporting your your near-term things but this isn’t like an all or nothing all your data is going to get stuck behind the stuff which is so important that the business understands the Via concept of us and their role and the modeling side as you start I keep going to examples that we’ve had we’re like why can’t you do we we’ve built this like yeah but you did it one time in Excel and yeah I’m a data analyst I’m one of the bi guys here but you have to understand if we’re going to build this we need something in

51:17 going to build this we need something in a way that it’s going to work and Empower beyond that like then why use power bi because again the idea of 17 unstructured sources in a Way businesses have done things like yeah I’ve just downloaded everything and made it by date it’s like it doesn’t work like that so having them understand too when we become part of a project on like what our goal there is it’s not just the visual it’s not just we’re going to build a fancy report some things that I’ve seen when when doing joins across different data sources there usually seems to be this

51:49 sources there usually seems to be this need to have a an area where there are people that can edit modify build you people that can edit modify build data that is homeless it has no know data that is homeless it has no home and what particularly for the marketing data it is it is data that is so these other systems that you’re grabbing data from the marketing efforts the spend like there’s all these third-party tools that you use to help the marketing team do their job it feels like a lot of the times the the rigor on which we enter data into these third-party systems we think about these tools in a way of I

52:20 we think about these tools in a way of I just can use this tool as opposed to how am I going to tie all this data together again when we come back and bring everything back together into your to your point set I think earlier was you your point set I think earlier was when you start building this know when you start building this logical model you start understanding hey you can’t abbreviate the product code number I’ve had I’ve had cases where they’re like oh we only have five digits to have in this field and you digits to have in this field and we have a seven digit product code know we have a seven digit product code so we abbreviate this and add some other data here that helps us get close to what the match is that’s not a match like we’re adding a whole bunch of business logic because the tool you’re

52:51 business logic because the tool you’re trying to use is insufficient for what you want to build or we’re selling things on a platform that doesn’t have a full descriptor of what we’re trying to look for like there’s not a clear link between what that is and everything else so there’s all these really weird scenarios where people need to be involved to generate data in order to bind and collect the data together yeah and I think the point here made by another one in chat was Data quality is another big topic on this one because what we’re talking about here is the The Logical model starts really

53:23 the The Logical model starts really pointing at how how good is the quality of data how many data fields come back with blanks is that going to be a problem yeah and this is this is like it’s not part of the camera it is part of the conversation but not but like but the modeling is the like the first thing in order for you to be able to do data quality in data governance so if those are important factors in your organization you are investing in Enterprise type activities towards data models we’re going to run out of time here but like physical data model is

53:53 here but like physical data model is essentially where you we’re taking the the previous models which are really logical like disc wrong word tool agnostic right they’re they’re this is how data Works within our ecosystem and the physical data model is taking that logical and and laying out the the platform related steps actual implementation of tables columns Keys constraints indexes whatever we need to do to make the the things work but ultimately I love the fact that like this is a great way in in both tangents

54:24 this is a great way in in both tangents one to create a blueprint for the technical builds but also set the stage not just in the initial conversations but also have it have the place to go back to when changes are going to be requested by business and what that creates is a shared understanding even at even if it stays at the conceptual model a shared understanding between business I. T and is a I think foundation for moving forward in adopting new things

54:55 moving forward in adopting new things into the model or adjusting as you go along right and I I just I love how he laid this out I think it makes a ton of sense and and I’m gonna get a lot of use out of it so I’m excited about it yeah the last part that probably could be a whole episode in itself as I think about this is once you build a house right it’s also the investment like can you pay your mortgage and so then there’s going to be things that the no that can you afford it can you afford it right because there may be the upkeep so

55:26 because there may be the upkeep so just the piggyback on that the marketing example we one of the proudest things I’ve done career-wise in the state was basically creating or working with business on creating a whole marketing platform tracking so basically creating one unified tracking code that went to all their systems and they didn’t understand why we’re doing it at first but it’s like no you put this this and it’ll actually go sent everywhere but it’s the same key and we’ve built this power apps but the idea here was they had to follow that structure because there was nothing yeah so it’s

55:57 because there was nothing yeah so it’s like yeah you’re changing the business process such that the data is actually able to join back together and that and that’s the power here right and so yeah but the idea is you’re only going to be as good as you have to understand do you have the investment from them yes again are they going to be able to pay whatever they’re asking for I think that’s why the executive sponsor is there right that’s that’s why we’re trying to like Stitch this stuff together and and get it all working because without that executive sponsor we can’t push on other teams to like hey we actually need to change our process you’re literally giving us not enough

56:27 you’re literally giving us not enough information to make what you want a reality help us build what we need to build in order to to bring the relationships together in 2008. So yeah so I’m gonna I’m gonna jump in here chatbt I ask quick chat gbt what’s the difference between the conceptual logical and physical data models and I thought it did a good job of summarizing part of this article but also also with what we’ve been talking about here in the conversation like the three models are the conceptual model is used to establish their entities their attributes and their relationships The

56:58 attributes and their relationships The Logical data model defines a structure of the data elements it sets the relationships between them and it’s more detailed than your conceptual model but is also technology agnostic and business oriented and I thought those two points were very relevant it doesn’t it you’re not you’re not really building a party I did a Model it’s still in that higher level version and Seth to your point earlier it made it 100 this these could be tables in your gold layer in your medallion architecture in your lake house that’s what those could be

57:28 lake house that’s what those could be the final tables You’re Building they’re they’re designed ready to be consumed by any tool whatever you’re going to build and then finally here you have the physical data model which is that describes the database specific implementation right of the data model right that that is build it in power bi build it in SQL Server whatever that physical object looks like that is the it incorporates all the different data types the conventions and the limits of that data model so so anyways really good conversation today

57:59 anyways really good conversation today if you like this conversation if you thought this was really cool if you enjoyed talking about data modeling and getting super into the weeds definitely go check out the article from from Nicola around data modeling and it’s also in the description so check that out great great article go read it are only asked if you enjoyed this conversation if you enjoyed this topic please just give us a like or a thumbs up and tell somebody else about it that’s how we get this the word out about this podcast if you enjoyed it let somebody else know I found medium

58:29 let somebody else know I found medium value from it you you you can you can relate to the the modeling that we were talking about and discussing thank you all for your listenership and we really appreciate you listening to the podcast Tommy where else can you find the podcast and you tell those people you can go to Apple and Spotify make sure to subscribe and then once you start enjoying it Go send us a mailbox topic on what you want us to talk about power bi. tips slash the podcast and then just watch live Tuesday and Thursday 7 30 a. m Central even if you don’t enjoy it just give us give us

58:59 you don’t enjoy it just give us give us some feedback what should we do better what can we do yeah we’ll take feedback we’ll pick corrective criticism as well thank you all so much we really appreciate your time and we’ll catch 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.