Move Faster in Power BI – Ep.532

Moving faster in Power BI is rarely about one secret trick. In this episode, Mike and Tommy unpack the workflow decisions, report-design habits, and modeling patterns that actually save time while helping teams avoid the rework that slows projects down later.

News & Announcements

This week’s “beat from the street” focused on a different kind of productivity gain: getting more done with less typing. Tommy described a workday driven by voice, agents, and well-structured personal workflows, while Mike connected that shift to the broader way AI is changing how practitioners capture ideas, produce deliverables, and move through day-to-day work.

• From Excel to Delta Tables: Simplifying Excel Data Ingestion with Shortcut Transformations (Preview) — Fabric’s new Excel-to-Delta experience brings low-code ingestion from Excel workbooks into Lakehouse Delta tables through Shortcut Transformations. The preview adds automatic sheet discovery, flexible multi-sheet processing, combined-table or table-per-sheet outputs, schema inference and drift handling, continuous sync, and monitoring support, making it much easier to operationalize the Excel files that still drive so many business processes.

Main Discussion

Mike and Tommy frame the episode around a practical question: what are the everyday Power BI habits that save the most time once you move beyond beginner workflows? Their answer is that speed comes from better structure, repeatable design decisions, and smarter ways to prototype, debug, and organize your models.

• Use a real sandbox phase before you polish. Start by sketching report pages live with stakeholders so you can validate what matters before you spend time refining every visual detail.
• Standardize design with themes and templates. Reusing layouts, background patterns, and consistent styling keeps teams from redesigning the same report shell for every project.
• Organize semantic models so navigation is easy. Hide what users do not need, apply strong naming conventions, and use display folders intentionally so measures are easier to discover and maintain.
• Prototype on a larger canvas when you are exploring ideas. Giving yourself more room during design makes it easier to compare visuals, inspect values, and work through layout choices quickly.
• Develop locally with a reduced dataset when it makes sense. Smaller local data volumes can dramatically speed iteration, refresh, and publishing while still letting you test the report-building workflow.
• Use model views and domain-based modeling to reduce confusion. Focused relationship views and models built around clear reporting domains are easier to debug and easier for users to understand.
• Keep debugging small and controlled. Reproducing a DAX or model issue with a narrower sample often makes context problems much easier to isolate than testing everything against the full production setup.
• Lean into advanced authoring tools as you mature. Features like DAX Query View, TMDL view, deeper Power Query fluency, and template-driven design can compound into major time savings for experienced teams.

Looking Forward

Pick one workflow improvement from this episode and apply it on your next Power BI project: run a true sandbox session, clean up your model organization, or standardize your starting templates. And if your team still depends heavily on Excel files for operational data, the new Excel-to-Delta preview in Fabric is worth watching closely.

Episode Transcript

0:01 Lighting up the sky. Dance [music] to the day to laugh in the mix. Fabric and A. I get your feels. Explicit measures. Drop the beat [music] now. H feel the crowd. Explicit measures. Good morning and welcome back to the explicit measures podcast with Tommy and Mike. Hello everyone and welcome back to a wonderful Thursday. jumping into some more PowerBI conversations today.

0:32 more PowerBI conversations today. Dude, it’s another great day. How you doing, my friend? I’m doing well. Things are very busy right now. I can’t complain., but I’m having a lot of fun. There’s a lot of new things coming out left and right. There’s tons of news in the micro in the the PowerBI, I guess, ecosystem, I guess, from the community. They’re just fun things to unpack right now. So, I’m doing a lot of exploring of what people are building., people are building MCP servers left and right. Microsoft is making skills for things. Like there’s just a lot of things happening right now that I’m trying to test all these different things.

1:03 these different things. It’s only going to get busier too in two weeks. Like that’s the crazy thing. Usually there’s a calm before the storm before [clears throat] a major conference in Microsoft Build, which I really the now I guess the second largest for PowerBI fabric. I I would say Ignite’s probably larger now than Build. I think I think Ignite last time it was there was 10, 000 people at at Jeez Jeez in Chicago when Tommy I think you and I went and then we went to the fabric conference in Atlanta I went to the fabric car in Atlanta you get to make it to that one that was 8, 000 people

1:35 to that one that was 8, 000 people so that’s that is pretty large build I don’t think is slated to be as large it’s going to have a huge online presence but in person there’s not a lot of people coming to the conference and it’s much more developer focused usually right way more developer and so [clears throat] I I get to go speak on some very specific new agentic experiences. Sweet around creating reporting with new agentic like experiences. So I’m going to highlight some of the agents and how they can help you build things. Again, Again, yeah, yeah, I’m going to focus on the creator agent.

2:06 I’m going to focus on the creator agent. The agents are not giving you the answers. The answers are the creators. They’re building the things that you want to go use, which I think is a great way of doing it. Anyways, today’s main topic is move faster in PowerBI. And today we were looking at our comment. So this is actually a note that came from Reddit, which I’m not sure why, but it’s the title of the topic is five things I learned in PowerBI that saved me a lot of time. And I remember reading this a while ago and thought this was a really interesting topic. Yeah, Yeah, for whatever reason, it’s already got

2:36 for whatever reason, it’s already got 167 upvotes and someone already deleted the post. Like it’s like, sorry, this post has been removed by Reddit’s filters. I don’t know what it was doing, but I thought it was just fine. I’ve seen worse on Reddit. [laughter] I’ve I’ve seen worse on Reddit, so I’m not sure what that meant. Or someone from the Microsoft channel removed it or whatever, but it doesn’t matter regardless. there is still all the comments that are still attached to the thread, which we’ll go through some

3:07 the thread, which we’ll go through some of those. And I do remember reading it. The whole gist was, “Have you been building in PowerBI for a while?, here are five things that have saved me a ton of time in working with reports. And they weren’t hacks. They weren’t like, “Oh, update your. exe file and update the binary.” No, it was it was just normal,, here’s techniques that you use. And so, this is a good opportunity for us to one just review that., I believe I made a couple comments on this one as well on this thread. So,

3:37 on this one as well on this thread. So,, definitely would want to jump in and hear what your five things are, Tommy., working specifically in PowerBI. What five things do you do that makes you faster today than when you started in PowerBI? That’s so that’ll be our main topic. Moving on from there, Tommy, what kind Moving on from there, Tommy, what news do you have for us today? of news do you have for us today? So, I have a beat from the street, something I wanted to share with you, and there is a fabric update. I’m going to let you I’ll let you pick. Which one do you want to do first? I think I’m going to like this beat from the street. So, let’s let’s hear the

4:07 the street. So, let’s let’s hear the beat FROM THE STREET. YES. YES. I THINK WE need a different name sound for sound for We probably need a different sound effect. What What would Let me ask that question, Tommy. What would a beat from the street sound like? I’m thinking some like city vibe like if you’re New York City horn like. Yeah. Like a taxi thing or a subway sound. I’m walking here. [laughter] All right. I’ll record it. I can record it. You You could sound just like a New

4:38 You You could sound just like a New Yorker, Tommy. That would be great. [laughter] That’s great. Quick, give me give me the sound clip. I’ll be quiet here. Give me the clean sound clip here. What What would it sound like, Tommy? Okay, I’m going to match up. Hey. Oh, an activate. Oh, I’m walking here. There you go. There [laughter] it is. That’s crazy. All right, good. Maybe we can pull Maybe we can pull that one in with some with some noise in the background there. I got Sounds good., we we have a face made for radio, Tommy. This is good. We Exactly. [laughter] Someone told me I

5:09 Exactly. [laughter] Someone told me I can’t even do radio with my face. So, I think that was your dad. [laughter] Oh, God bless that man. [laughter] Awesome. All right. This is great. All right. So, beat from the street. Mike, yesterday I worked harder than I have in maybe in a few weeks to be conservative. A few weeks. Yeah. Yeah. And I’m probably typed on my keyboard

5:40 And I’m probably typed on my keyboard three times the whole day. The whole day I testing out, but really it was because of the agents that I’ve already put into place. all the voice features that are available in Notion, in Claude, and Outlook, the the fact that you can dictate. And I was telling my wife, I’m like, none of this would be possible without the infrastructure that I built because one, I was working on a statement of work and it’s a pretty extensive project. The other one was

6:12 extensive project. The other one was working with the MCP server. We’re trying to diagnose some things. The thing, Mike, is people just think, oh, just talk to AI and it works. It works. Okay. But generally, you look at all the documentation, all the things that are being written and it sound has my style. It has the way I like to organize things more importantly knows what my types of deliverables are. So I have certain skills that I’ve built specialized really just for my own company for building again MSAS

6:42 company for building again MSAS so the server it’s very custom in terms of what it’s reading from notion based on the projects that I’m working on not just build a bunch of measures and the whole day was literally I was able to go outside and just it was literally like I was talking to my team the whole day I told my wife I’m like I talked more yesterday than I think sometimes I’ve done in training and the amount that I got done the productivity because I think again I

7:12 productivity because I think again I still think there’s a misconception if you do AI then you just get to sit on your chair and watch YouTube while things are running that was not the case yesterday yesterday and the fact though too that it’s the productivity side it wasn’t like I was getting more done in the sense like we’ve talked about oh the volume of work but it the quality of the work that I was really astounded by because I was able to refine. I think I really think agents were made for people with ADD because they say the superpower

7:44 with ADD because they say the superpower of people with ADD is the creativity, the the improvise and figure things out. And Mike, I’m astounded. I’m absolutely astounded on what my workday looks like today. I know we’ve talked about the typing before, but, you can get we’re getting to a point now where you don’t have to type, but you’re getting the same work done. I would I agree in a lot of the areas you’re talking talking about here, Tommy. I am finding myself quickly

8:16 Tommy. I am finding myself quickly moving away from having to type on keyboards. I’m finding myself moving away from writing things down and having to type out things. specifically when I’m with my phone and moving around. My wife has been doing this a lot longer than I have or more so, but she’ll she’ll click the little,, microphone button and she’ll just dictate to the phone what the messages those things as well. What I find sometimes, and this is the reason why I’m hesitating here slightly, is is Yeah, I’m intrigued. I don’t think it’s actually a hesitation. It’s more of like I’m just trying to unpack the thought here. Yeah.

8:46 Yeah. Not every transcription service you use on your computer is created equal. And what by this is there’s the idea of sending words and voice to your computer, right? right? But you stutter and you pause and I know, okay, there’s like it sometimes has weird punctuation and you pause for long enough to think about the next sentence. And I I’m trying to be more thoughtful without having to rethink or resay something. I think there’s a lot of forgiveness.

9:16 forgiveness. I know what you’re saying, right? There’s a lot of forgiveness, Tommy, when I talk to you and I’m like, “Oh, this.” Then I switch moment, switch thought halfway through and then I come back and so there’s a I’m conveying information to you but sometimes the words I say is not quite as clean and smooth as if I had written them on paper. Writing them on paper or writing down on the keyboard is like this is a very clean sentence. It has like a a purpose. I’m writing into the statement of work. There’s very specific things but when I talk to them I’m not always doing the exact words I should be using on the sentence.

9:47 on the sentence. So, this is where things like Okay, so I’m bringing this point up because there were programs. I think this is why you’re bringing this up, Tommy, is because there’s things like whisper and there’s other there’s other things that listen to what you say and then it runs the output of what you said through an AI and cleans up the sentence, right? There’s no ums, there’s no ands, there’s no sos. It it just takes what you’re trying to say and distills it down into a better sentence. Let me just pause right there. Is that what you’re doing?, no, actually, and I’m I’m really happy you said that because one thing

10:18 happy you said that because one thing about the way tokens are ingested is ums, even if you were to type those, those are disregarded. Those are not important words to any agent. And that’s been that way since chat GPT came out. If you actually look at how it tokenizes words, words, it knows like,, I’m not going to reason about. Now, I really want to focus on what you said here, even outside of using a transcription service to make it more AI friendly. Mhm. Mhm. When I first started transcribing and trying to talk to the agents, I was very

10:48 trying to talk to the agents, I was very cautious and conscious of how I was speaking. I wanted to be clear like a normal instructions. I would sometimes,

10:55 normal instructions. I would sometimes, to your point, revise it. The last The last three weeks, I’ve tested out what have I just talked naturally. For example, Mike, to your point, I’ll be looking at something going item 8 through nine. Actually, what? items seven through 10. That’s the ones. Those can’t be here. So, you’re changing it mid-sentence, like correct natural. natural. And I’ve tested doing that intentionally a few weeks ago to see how much does it cather, how much longer does it take.

11:26 cather, how much longer does it take. Yeah, Yeah, Mike, it’s been phenomenal. So, I have tested at least with notions transcription service and Claude’s transcription service. Sure. Sure. I can speak in natural language. I can speak to your point like I’m speaking to you where it’s going where you change your mind in mid-sentence. I think that’s the best way to put it where you’re like actually no let’s not do that., and I’ll even pause like I have it open and I’m thinking going let me see here and I’m taking a look. I’m trying to I’m pondering go what if we

11:59 trying to I’m pondering go what if we what if we did this actually let’s let’s do that it may be using more tokens but again the way LLM work initially to take to tokenize characters and words it knows what are relevant words it’s looking for the word next to it so those things are very much disregarded and Mike so I would challenge you try just speaking at least to those two harnesses naturally naturally [snorts] [snorts] Well, I’m going to maybe check that out a little bit more, but

12:30 a little bit more, but yeah, yeah, one of the ones I use the most right now is using Windows H on my computer. So, Windows H is one that I use a lot., I I want to get a little bit better at using voice inside VS Code. That’s one that I know I’m a little bit weak on. I keep typing things and especially now that I’m doing YouTube videos and and showing people how to write it, it’d be it’s a lot faster. I’ve seen people, particularly Matt PCO, who’s doing a lot of this like AI programming. He goes, “Okay, now I’m going to tell my agent to do XYZ thing, and then he just talks and

13:00 do XYZ thing, and then he just talks and then that writes his transcription inside the thing.” So, that’s one area that I need to get better at, especially when I’m demoing to people about how the AI and agents work because I think it’s smoother. smoother. At the end of the day though, I will say this, I like this mode of talking to your computer as opposed to typing to your computer. This is one major interface, Tommy, that’s disappearing for me. for me. Like, if I think about how I interact with computers, the keyboard’s a pretty big part of a computer. You you use it to give anything I would want to

13:30 to give anything I would want to communicate to the computer was through the keyboard. This whole new world of saying things and having to type things out for you.,, this this is substantially changing how I maybe would work with a computer. And I’m even thinking, why can’t I do more on my iPad? Why can’t I do more on my remote devices? devices? Because then I could I can talk to them. I have a bigger screen. Really, what I want, Tommy, and I don’t I don’t know how to do this yet. This is one that I wish it was a little bit easier to set up, but there are some

14:00 easier to set up, but there are some iPhone applications where you can do SSH right into your computer. So, is a good one. Yeah. I’ I’d really like to have like,, the co-pilot terminal, right? I’d really like to have an app on my iPad that is this app is just a terminal window into GitHub Copilot, right? Or or whatever the whatever the cloud code co anything CLI. I’d really like to have that and then I could have potentially multiple tabs for

14:31 could have potentially multiple tabs for all the different CLIs and I’d like it to all communicate back to my main desktop machine. So maybe this is a software that exists already, but I can I can really I can really feel like this is another movement for me. And then I could go to the CLI, talk to it directly through my iPad. iPad. I do feel like Tommy, especially with the impending doom that is coming up with all of our tokens going up in price, specifically around GitHub copilot copilot and notion doomed, they’re all doing it. They’re all they’re all ripping up the price for the

15:02 they’re all ripping up the price for the the free launch of cheap tokens is over. They’re like, we’re ready to make money now. now. We think we have enough volume of people getting value from code. And so now what we’re finding is people companies are now I Microsoft I think just announced or I saw an article came out that said they’ve stopped all usage of Opus or Claude because it’s too expensive and they want to use other models because the prices of these models are approaching the price of an employee. So like well that doesn’t make sense. Why would I why would I spend so much money in these

15:32 would I spend so much money in these heavier models if they can’t run well? So I there’s a the war is coming Tommy and I I al honestly think having more localized models to do exactly what you’re describing Tommy like why don’t I have a localized model on my computer that just does the audio transcription for me that that pulls it down simplifies it does a little bit of pre-prompt engineering for me and then fires it off to the the fancy model and goes from there so there’s there’s a whole new world I think coming

16:02 whole new world I think coming orchestration of agents is coming and dude I’m telling you like the orchestration side notion just had their main like world like major conference and this whole orchestration side Mike I was outside because the kids it’s this is the last day of school so I brought out the the slide where they all do the whole water slide and I’m on the deck and I was working on my desktop so I got my Surface was able to connect to the new MCP d and I didn’t even bring my

16:33 the new MCP d and I didn’t even bring my heatboard with me to your point. Mhm. Mhm. So, it was just the Surface tablet without the keyboard. Yep. And I’m just clicking on the audio button on each thing. And Mike, seriously, I am at the Yesterday was at the point where it if you were recording me a fly on the wall, you would think I’m talking to a team of people. There was no The way the work was being done was so conversational and just so productive and again, so it wasn’t just getting stuff out the door, right? And I the big

17:04 stuff out the door, right? And I the big thing I want to emphasize is the fact that that the quality of the work and the things that need to get done. And it’s pretty amazing. It’s pretty amazing. And we are we’re still we’re still 1999. I think you you mentioned that or someone mentioned that like we’re in the world in 1999. We’re like this internet things pretty cool. Yeah. And I we’re going to I think Tom we’re going to continually move away from keyboards and mice and we’re going to move more towards touch and voice. I think that’s where we’re going to be moving towards., and so there’s gonna be more of

17:34 , and so there’s gonna be more of this. this. I think I invested on a 140 keyboard. Yeah. Right. The ducky. You ever heard of ducky? I love mechanical. I I don’t actually like to make keyboards are just so loud. I have a Microsoft keyboard. I like having keyboards that mimic my laptop honestly because I want to well I like those flatter keys just because if I’m here working on my computer at home and then I go travel I don’t want the keys to feel awkward in any way as I move between different locations and so I like to have the keys all kind I like to have the keys all consistent.

18:04 of consistent. I will say this though is the fact that I’m moving a lot more towards development with just my development cycles have changed. I’m seriously looking at getting like a Mac and just having that become like my laptop will be a Mac and I’ll get rid of all the Windows computers. Yeah, my Windows computers are all getting relegated to running various agents and just sitting on all the time at home and then I’m going to like Yeah, openclaw Hermes, all these agents are running here at home. I let them

18:34 are running here at home. I let them just stay here and then I go use other things. So, I still haven’t quite figured out the right system. I did as a side note I did go over to Azure Foundry and Azure Foundry I just ripped out a new version of DeepSeek. So for my open claw agent I’m trying Deepseek version 4 which apparently is pretty impressive. So I’m going to give that a whirl and see how that performs as like an agent that you can talk to. Speaking of models, there’s a bit of a t or a bit of a side conversation but do you use chat GPT at all? Any open AI

19:06 you use chat GPT at all? Any open AI models anymore? Yes. Yeah, substantially. I have you really? really? Yeah, I use 54 and 53 codecs. Okay. Okay. Pretty regularly and have there’s some things where I every so often I’ll dip into Opus to get something done or something heavy to that’s being used, but a lot of times I’ll use a lot of the five five Wow. Wow. the 54s 53 is right now. I’m also exploring a lot of other models right now. The open source model community is actually getting very performant. They’re much cheaper to run

19:36 performant. They’re much cheaper to run in my opinion. U I’m seeing anywhere between four to 6x cheaper in token usage. usage. I’m I’m cool with it. Like I’m okay having less spend on these things. So I have noticed, Tommy, when I stopped using GPT54, my spend on using GPT54 went down substantially in my Foundry spend. So Wow. Wow. Well, and Foundry doesn’t really hide the token usage away from you like it does in GitHub, which by the way, we’re one week away from GitHub increasing

20:06 one week away from GitHub increasing their prices. So, next week you’re going to find figure out how much token usage you actually have on GitHub. Crazy. Well, hey, we got a pretty interesting article. Let’s go let’s go back to legacy. Let’s go to Excel. You want to go to Excel? We got this whole AI world that we’re on. But what? It will it was is and will always be. That is our Excel application. And there’s a great article on the fabric blog called from Excel to data tables simplifying data ingestion with shortcut

20:38 simplifying data ingestion with shortcut transformations. So what do we got here? So we all know again you Excel is the universal language of business data billions of critical files again it’s the most used application in the world outside of Outlook and what we actually have available now is a zero code ingestion from any Excel file with fabric using shortcuts so automatic sheet discovery you can process multiple sheets at a time you can have

21:08 multiple sheets at a time you can have two different types of outputs either combined table. So all the sheets merged into a single delta table or table per sheet. Each sheet each sheet becomes its own delta table. The automatical schema inference. There’s cross sheet schema validation. There’s continuous sync because it’s in one lake. the sheet name helps with the val table names. It manages error handling. manage intelligent skipping of any banners or

21:39 intelligent skipping of any banners or metadata. So,, we’re talking about some crazy stuff here. Well, here’s everything, Tommy, and a kitchen sink. Like, you mean Yeah, this is about to go right to All the presents you wanted for

21:51 All the presents you wanted for Christmas is here, Tommy. Like, get all this cool this. They should have waited for Bill for this. this. There’s a lot of commitments being made in this post. Mhm. If they deliver on these commitments as they have written them, which I’m hoping they do, this is a strong feature. Like the amount the amount of volume of data that’s being handled and and managed inside Excel files is phenomenal. And to be honest, Tommy, Tommy, Excel is the best input system for tables, period. What’s that? I love working in tables inside Excel

22:22 I love working in tables inside Excel and that’s awesome to me., getting that information quickly out of Excel and directly into Delta tables into the lake makes a ton of sense. One thing it doesn’t really quite tell me what it’s doing here is I don’t really understand how often or when it triggers a refresh on the information. And that’s one of the things I imagine you do a pipeline, but it’s shortcut. I don’t know. It’s a shortcut. So, you I don’t know. It’s a shortcut. So,, if we have a team of people

22:52 know, if we have a team of people working editing in the Excel file. How often is that refreshing? Does it is it like just so a lot of times what I think I find sometimes here is sometimes Microsoft is just listening for like it’s actually watching the file like okay if the file is not changing it checks it every 5 minutes to update things. I I don’t know what the speed here there was not really clear indication on like the speed of this. there’s also this idea of like not not doing a pull but doing a push. So

23:23 not doing a pull but doing a push. So the the thought here being if the Excel file changes push an alert to fabric and say hey it’s changed now go update it right cuz SharePoint you’re basically saving the file in SharePoint it’s making changes SharePoint has this ability of saying like hey look something changed there’s a notification I can listen to that notification bus and then go make the change. So yeah that’s I I just want to understand a little bit more there but other than that wow this is amazing. What what sticks out to you Tommy? I, Mike,

23:53 sticks out to you Tommy? I, Mike, like I said, I think the biggest thing is the fact that it’s that automatic side. So, we’re talking about here sheet all the different sheets and I’ve been doing a bunch of I did a ton of Excel training in the winter. Yep. Yep. And it’s how many people are utilizing very unstructured Excel, but again, the company relies on it. So, the fact that you go just get a one link and we’ll take care of the rest. there’s there’s actually like maybe one or two statement of works right now in the pipeline I may have to revise if it works as advertised because we can get

24:26 works as advertised because we can get started right away. And I think also too there’s this other idea of this proof of concept that we can get much too much faster. faster. Yes. Yes. Which I I I actually just was considering is I know we have all the stuff in Excel but it’s coming from a database. Well, guess what? Let’s see what it looks like in fabric. How do? to send me the Excel files, I’ll have it by lunch. And I think it’s there’s something to that. I really like this feature for that reason. And so I’m I’m very excited for

24:57 reason. And so I’m I’m very excited for this one. This is great. The other one I will just note here that I think would be super powerful is this one called automatic schema inference and schema drift. drift. So it will column names, data types are inferred from each sheet. The schema is validated across the different sheets, mismatches are detected, logged, and handled gracefully. So if this tells me when some user decides to add a new column, it just ignores it and moves on. Great. Right?

25:27 Great. Right? Go back in and change the schema and so you can absorb that new column. This is one of the largest challenges I think I have in Excel in general. People change stuff, they modify it, and it breaks my my pipelines and things. So having this available to you is awesome. I’m cur like I think I think the biggest thing is thing is how long is it going to be in preview? And that does mean everyone would have to download probably like seven years or something like that. Yeah, I think about Do we have English to Spanish yet?

25:58 to Spanish yet? translation. Okay, that’s still preview. preview. No, we’re still preview. All right. All right. So we’ll take just keep it in preview forever. Never have to finish the feature. Anyways, very cool article. Check it out. It’s in in the description here below. It’s also in the chat window if you want to check this out. I think this is a great new feature., yeah, people are going to love this feature. This is going to be awesome awesome a thousand%. All right, main topic. Let’s do it. Moving faster in PowerBI. So, there’s also a a Reddit link here. And so, the downside of this one is you don’t get to

26:29 downside of this one is you don’t get to see the actual five answers that the original author had wrote for this. So Tommy and I are going to give the author I guess I don’t know it’s historical listen 358. Thank you very much for your comment. I’m going to try to infer some of your five things that you learned that makes your that saved you a ton of time. and I’ll I’ll probably pick I some of them I think I picked up myself. Yeah. Yeah. But let me I’ll throw the link here in the in the chat window as well in case you want to go look at the actual link. All right, Tommy, let’s kick

26:59 link. All right, Tommy, let’s kick things off. What five things have you learned that have made you faster in PowerBI? [laughter] Yeah. Yeah. Over these what 11 years now? There was this great thing called quick ribbons that Oh, wait. Can’t use that. That was a great time saver. No. So, I think when we look at this, Mike, I I’m I was thinking about this question in terms of just the actual development and then the process of building. And here’s what. There is one thing when I’m actually in the computer in desktop building, but there’s also the whole

27:30 building, but there’s also the whole journey of getting a report up you working with the business, right? And I I so I want I may have a combination of the two because there is things that I found to be huge time savers when it comes to actually helping out the process from we have we have needs to here’s your final report. Sure. Sure. And I think both of those are pretty critical here. And what I’m also going to try to do, Mike, is stay away or avoid things we’ve probably mentioned in the past., so I’m going to try to be

28:00 the past., so I’m going to try to be a little more creative here. So, jumping in, I always think about the building of PowerBI into stages. I don’t know if you do this and I’m not sure of your exact process, but what’s been the largest timesaver for me by far and away is the methodology on how I build and again with a stakeholder or with a client. Most people try to build everything off,, right off the

28:31 everything off,, right off the bat. I have the data, I have the, you bat. I have the data, I have the,, the scope and we’re going to try know, the scope and we’re going to try to build everything to perfection. Mhm. I think Paul Turley actually had an article years ago about breaking it into two different phases. And since then, Mike, I have sweared by this. And what I do is the beginning of any new model or new report is what’s called the sandbox phase. phase. We don’t focus on colors. We don’t focus on titles. We focus on features,

29:02 on titles. We focus on features, functionality, and story. So, for example, I meet with a client twice a week and they know that the port report’s not going to be perfect. I’m not we’re not doing data validation. I specifically tell the stakeholders, don’t worry if the numbers are accurate. That’s not what we’re here for today. What we’re here for is is this relevant to you? Is this meeting what you were expecting? The point of the sandbox phase to me is it’s real time. I have desktop open, not the service. And what

29:33 desktop open, not the service. And what they do is like I go hey we have a this drill through page what’s relevant here. They go I need to see the team and I you They go I need to see the team and I we really need to see x y know we really need to see x y and z like okay well we have those measures and we put it on the page in real time. These are 25 minute calls that that completely,, really accelerate the build of a report because now we’re not focusing on well, I don’t like that color for that or are you sure that’s the number

30:03 that or are you sure that’s the number for,, X per hour. Mike, I can get through complex models because I know what’s relevant to the user because they’re finally seeing the things rather than,, trying to make it up, right? Because it’s hard when you are in this, you have the wide angle people say, I think I need sales,, by a team. You show them, then they go, that’s actually not important. So getting that interface and that feedback is an acceleration. Then we go

30:33 feedback is an acceleration. Then we go into the second phase, which is the development mode. We’re working on the quality. We’re working on the polish of it. And this has been my greatest accelerator that I would recommend to anyone. So, I know I feel like we’ve talked about this before, but what’s your take on that? Is there a methodology that you use as well when it comes to working on building a PowerBI model and the report with a client? client? Yeah, I do a lot more of this now. A lot of this is talking about the

31:03 of this is talking about the requirements like what you’re trying to see. I think Mhm. Mhm. organizing things out. I guess my area would focus a lot more on what what things have saved me a lot of time. time. I’m going to start with the theme file has saved me a lot of time. Okay. Yeah. And even talking with other people who are building things like for example when you drop a slicer on the page usually by default the slicer has a setting of drop down or something like that or check boxes. Well,

31:33 like that or check boxes. Well, if you change that setter, the slicer setting default to be drop down all the time, it saves you a ton of time. You don’t have to keep clicking things all over the place. And yeah, Enterprise AR in the comments here, you’re right. The the main message of this has been removed, but the comments still exist. So, we’re going to go off of the some of the comments here. Yeah. Yeah. So, I think for me, one of the major speed improvements is theming and wireframing, right? Get your background images corrected corrected. don’t spend a lot of time positioning the

32:04 spend a lot of time positioning the visuals on the page exactly where they need to go. Instead, use a a background image and we call them a scrim. Use a background image that is pretty and has shading and all the things you want to do and organizes the page well. having templates or files that are templates and ready to go, basically pre-connected semantic models, reports that are already ready to go is a huge time savers. And I’m going to point on this as well. I think this is so valuable to people and teams that we’ve built an entire tool around it called power designer. It’s available for you

32:36 power designer. It’s available for you now today in fabric. You can go use it today for free. You can change your theme files. You can make a you can get a sample report a template basically and you can deploy that report on top of

32:46 and you can deploy that report on top of a model. So the amount of time I I don’t want to spend time polishing and making it look good. Let someone design what that looks like and reuse that template over and over again. And and my design my design methodology is not building a single report for a very single purpose. It’s design a generic report that can be used many different ways., hey, this is a left nav. Okay, every page should have the left nav. This this report will have a header. Great. What does the header look like on all pages? Build that in six pages of report and then

33:16 that in six pages of report and then just have different patterns on the bottom section of the the report page. So that’s the stuff that I I think is really useful. Do you use still templates in terms of the PBIT file? So there so it’s interesting. Okay. Okay. So because it’s interesting Mike you can do the theme file which is great. Think about how you would do it in desktop. I just use power designer for everything and so I don’t need any files. files. Okay. Okay. Cuz I do all the work I do the 100% of the work in now power designer. So power

33:47 the work in now power designer. So power designer let you add the background. You can add the visuals wherever you want. everything is there built into the report without any data and then there’s a one button press and I say publish this you pick the semantic model you pick the workspace and the template autobinds to the data set yeah the report autobinds said semantic model so I’m like yeah you take my template from power designer you say publish you pick a semantic model it automatically binds the report to the semantic model and publishes the template

34:18 publishes the template you create is going to be off of Because you’re saying semantic model, which is the data. the data. So you’re telling me I Okay, so let me just make sure I’m clear. No, no, let me make sure I’m clear. Yeah. Yeah. Yeah. Think about the workflow, right? If I’m wireframing something. Mhm. Mhm. Right. If I’m just doing the wireframe portion portion Yes. There’s usually two individuals working on two different workflows. Right. Tommy, you’re talking with the business user and saying, “What KPIs do I need? How do I build stuff? What what semantic model do I need to create?” you’re you’re working with the business and stubbing out what is what is the

34:49 and stubbing out what is what is the main objective of the report right so meanwhile you’re building like okay does this KPI look right does this measure look right do these dimensions work well with this measure is that what we want so you’re stubbing out with the business most of the semantic model and even I would argue a majority of the semantic model can get stubbed out with just good star schema good data engineering and a lot of sums and counts across the model like that’s with you that’s a first pass at a model. There’s like other nuancy things you’re going to build later on, but like a good first pass is here’s my fact table, here’s the

35:21 pass is here’s my fact table, here’s the sums, here’s the aggregations, I’m going to organize it all and put descriptions in. Boom, done. Yeah. Yeah. So, in relatively short time, you’ve got a sample model. At the same time, someone else could be building or doing the design work for what the report will look like without any data attached to it. And so you spend time somewhere else building like, okay, we’re going to have a we’re going to have a six I’m going to do six sample pages and I’m going to have a navbar that looks like this. The background images will look like that. And they and you use a bunch of design work to make the the framing

35:51 work to make the the framing the report of the report. Now you need to marry these two together at the end. And so what I’m saying is Power Designer is the tool to build the template in, but it’s a couple clicks until you actually have the full report attached to an existing data set. And then you’re just dragging and dropping data fields from the semantic model directly into the report. So the template we’re building is stored in Power Designer. Power Designer lets you keep the template without any semantic model attached to it. But when you publish the report, you do want the report bound to some semantic

36:22 report bound to some semantic model. So it’s part of the publishing step of the template. It’s an interesting task flow because I you the nor the the more basic one that you would assume is you would bind your theme to a report, right? Because that’s the canvas, the visual side. Okay. So theming is so essential. It’s one of those though too, Mike, I would almost even it’s not pushing back, but lean in more is probably is not only is that a timesaver, this is that’s a given at this point. That’s like saying, what’s a timesaver? Star model,

36:54 what’s a timesaver? Star model,, star schema. You’re like, no, that you have to do that if you want to survive. So, yes, I agree with that one. All right. So, yeah, go ahead. I was going to say, let’s go back to the article again. So, one one of the comments down below, we’re decipher here a little bit., related to point number two, someone makes a comment. It says, I recommend creating a dummy table for holding measures. it will make navigation it no it’s not mine this is it will make a navigation and locating measures much easier especially when you have 20 plus measures to navigate through inside the model create

37:25 navigate through inside the model create a blank table and then add your measures there there this is a I don’t really this is one I I disagree with this one I don’t like putting measures in the measure table and the reason because when you do the measures in the measures table it removes the context of where was that measure calculated from cuz you may have if you have one fact table fixed that no how do you see that oh that was old Q&A I was thinking old Q&A never mind never mind you can’t see that so I don’t really

37:55 you can’t see that so I don’t really like I really don’t like this ability to have the measure table because the fact table then is just like another dimension like everything else it doesn’t really make sense so I my preference is in the fact table hide all the columns you don’t need and make sure that only measures are in the fact table and then when you hide them all, the fact table will just move up to the top as if it was a measure table. table. But that also requires you to pull out a lot of dimensions out of the the fact table. table. But not really though. And I see I I was

38:26 But not really though. And I see I I was actually going to say very something very or the same thing as my third one, but here. But but let’s lean into this. I don’t think a lot of people still know that the fact that you can have a measure table from your facts. And to be honest, Mike, if you’re doing your model right, you don’t need any columns from the fact table in the visual side, right? Because I’m not going to be dragging and dropping fields from my fact table into the report. Maybe I’ll do it on audit table, so I’ll unhide it, but I’m not going to

38:57 so I’ll unhide it, but I’m not going to frequently do that. So, if you have a good fact table, Mike, you should be able to hide all the columns. I’m going to say in general the answer is yes. I feel like there’s there’s always situations where there’s edge cases of like little things you don’t want to keep in the fact table. There’s there’s some dimensional data that is just makes sense to keep it there. It’s it’s a it’s a I’ll say it’s a guideline, right? It’s a guideline. I to your point Tommy if I had a perfect world and I could do that for every

39:27 world and I could do that for every single thing. Yeah. It would be nice. Yeah. Yeah. The the complication of this I think is when you do that every single text column needs to become some reference column in some dimensional table. So what I think that does is you get this ability like you get too much too many tables sometimes and you you have to like so Donald what Donald’s pointing out here is this thing called the junk dimension table right you add a

39:57 the junk dimension table right you add a couple extra columns to a table and you have to create a unique key so basically it goes let’s look at the fact table let’s pick the right collection of columns these three columns these two columns those one columns bring them together and provide the right combination of those and make a brand new key that goes between the dimension and the fact table. And again, they call that a junk dimension, right? The dimension table gets a bit higher granularity, but you have to relink that back down to the main fact table, but you still need a one to many relationship to join the

40:29 a one to many relationship to join the two tables together. It’s more data engineering work to do it, but it definitely cleans up the model and makes it a bit easier to work with. So th this is one area that I think the modeling side is hard to do. But if you can think about naming conventions I so let me say this way in my ideal world Tommy I wouldn’t even have tables in the the model pane. I would have dimensions and I would have measures. A lot of other analytical tools just do

41:01 lot of other analytical tools just do that. Here’s the dimension things. These are all text and I know what you’re talking about and that frustrates me. It it it frustrates me that we can’t get like a cleaned up view in desktop of just dimensions and facts cuz cuz I would rather you could think of it your mental model could be each table represents a folder customer sales product like you could think of those as folders in the right hand side. Mhm. Mhm. But sometimes you just want it to be a bit cleaner I feel like. And I think I think sometimes it gets very

41:31 think sometimes it gets very overwhelmed. Oh, Donald, please don’t say the word metric sets. You’re gonna make Tommy go and spin off on a a random tangent here. But yes, Donald, you’re Those don’t work, by the way, in case you you metrics. Metric sets were what we were looking for. It’s like that, right? Here’s a here’s a measure and here’s all the columns you could or could not use with that measure. That makes a lot of sense to me. So, this is this is where I’m I’m struggling here., yeah, perspectives aren’t supported in desktop or in the service for viewing them, but these

42:01 service for viewing them, but these these would be great opportunities to resort the columns and fields to make it easier for users to build with. I I know what you’re saying here, but I maybe it’s just because of the 11 years of using PowerBI, but the tables make lot. That’s a mental model that makes sense to me that’s more organized rather than alphabetical. Speaking of organization, I we’ve talked about display folders multiple times. So, I’m not going to say just display folders as a timesaver, but enhancements to what you can do with display folders, the naming conventions

42:31 display folders, the naming conventions and the organization of display folders. Kurt Buer [clears throat] just put out a killer post about this just recently. It’s like he has like an mind mind reading because I when we were doing this episode when I wrote everything down like ah like but I saw the same thing this morning dude but going off of Kurt going off me and he we thought of it at the same time. Don’t care either way. It’s a good idea. [laughter] It is a good idea. It doesn’t matter who claim u the biggest thing here is for me

43:02 claim u the biggest thing here is for me personally with the way I organize my items is I will write it out first especially when I know I have more than 10 measures I’ll write on a piece of paper what would make the most logical sense and I also have common organization based on measure types for example Mike anything that’s time intelligence has a certain name to it and it’s always a nested folder. So, another thing you can do, again, if you don’t know it, you can nest display folders. Hopefully,

43:32 can nest display folders. Hopefully, most people know that by now, but I try to organize like, hey, these are all for fuel. Okay, so here underneath that is a display folder for all the time intelligence for fuel.

43:41 intelligence for fuel. Yep. Yep. The other thing I do too, and I think Kurt’s article had the image in there. Again, Again, I just I just sent you the post. He made a post on X around this one and he said don’t use this which is data that doesn’t have any structure to it but then instead use the data structure where he has basically a whole bunch of like invoices under invoices there’s measures there’s value measures and then there’s quantity measures then there’s lines and then there’s facts and so he’s got everything organized in a nice clean folder structure and he’s even using like a nu

44:13 structure and he’s even using like a nu like a numbering schema he’s got level zero level level a like and it’s he’s doing a outline of how all the data is structured. This is really it makes a lot of logical sense. Now the only downside I have of this is there’s a lot of clicking to get down to the measures potentially that you need but it’s very organized and honestly for me like I never I don’t do the alphabetical or the numerical organization you but the reason why he’s doing that is because power to your point when it comes to organization PowerBI is alphabetical.

44:44 PowerBI is alphabetical. Yes. Correct. If you had a display folder called fuel, it’s going to show up in the all the columns next to F. So I always do underscore underscore and again the categorical information. I try not to stay from three levels of nesting because it can get to your point pretty tedious to find. find. Yeah, don’t go that deep. but generally because again if I have five six display folders in a table and then everything underneath that is like you everything underneath that is like time intelligence dynamic variance

45:15 know time intelligence dynamic variance whatever that case may be that’s a lot easier for me than going three levels down. But again I always start with the underscore certain things and I always try to have like dynamic or rolling period measures. I’ve gotten pretty good at that and I just found out Mike more and more that everyone wants that. they just don’t know that they want it. like so you can have a slider say hey show me three weeks first the last three weeks. Ever since I started introducing that to people they’re like we love this and we want actually all the pages to

45:45 and we want actually all the pages to have that. It just seems to be a very popular use case. But going off of this it’s underscore time intelligence is TI dynamic is underscore dynamic. So, and that is for me been a huge timesaver in terms of how I actually just navigate. And what I like that also Kurt did here too. If you don’t if you can’t hide all your columns in a fact table to your point, you can put those in a display folder.

46:15 you can put those in a display folder. No one you can’t don’t really people don’t even think about that. A lot of times too, I’ll put a Z like underscore ski Z and then like these are the columns so I know it’s always gonna be at the bottom. Yep. So again, just because PowerB is alphabetical. So that’s one of my time common time savers. So Mike, what do you got for me? I have another one here. So point number one, I think they were talking about making templates and things a little bit. So we already talked about the template piece earlier., but one of them was making a prototype. And Tommy,

46:45 them was making a prototype. And Tommy, you were touching on this a little bit earlier with your like, hey, I’m I’m wireframing this from you. Yeah. But prototyping things is really good. I will this is maybe like a nuanced of this one is when I’m prototyping I’m changing the page size. So the page size in a report is set to like a certain size screen and it was 19 no 1028x 720 initially was the original size. They just recently upgraded it to 1920 x 1080 which means it’s an HD screen. I now I I’ll go in

47:18 it’s an HD screen. I now I I’ll go in I’ll make my page 300 3000 by 3000 a huge square th00and by 3000. Yeah. And and while I’m prototyping and while I’m I’ll build multiple styles of visuals or multiple images or iterate or I’m doing data quality checks. So I’ll make the report [laughter] I know WHERE YOU’RE GOING. OH, WOW. YEAH. So I’ll make the report canvas really almost no one almost no one changes the report canvas size from what powerb gives you by default. Almost

47:48 what powerb gives you by default. Almost nobody does. I whenever I do training nobody changes it. So you’re using this as like a like an actual sandbox like a whiteboard like like like a sand. Yeah. So hey I want to build this I haven’t thought about this. [laughter] Well I’ll zoom out. It’s it’s almost like zooming out farther and you get more. So, one of the things I wish PowerBI desktop would do is like sometimes I really want like infinite canvas mode, right? I have a lot of tables on a page and I just want to have it be,, I don’t actually want a page. I just want to just be able to put

48:18 page. I just want to just be able to put things anywhere I want and just zoom out and zoom out and zoom out and zoom out and keep getting bigger and bigger and bigger. But we don’t really have that feature, that capability. So, what I have to do for prototyping is I’ll make a really big canvas. I’ll put a lot of visuals on the page and I’ll just start, okay, here’s a bar chart. Okay, great. Copy that. Make another one. Okay., what does this look better as a line chart? Okay. Do I want labels on or off? And I’ll I’ll just keep copying things until I have like a section on the page that’s like my PowerBI mural.

48:48 my PowerBI mural. Yeah, it’s like a PowerBI mural like and yeah, that’s that’s where I’m going a little bit. So that’s how my mind thinks. My mind thinks more in like board level big view of things and then I can go test through things. This is also a really good technique I think to use when you’re looking for data quality issues. Mhm. you can add a mix of slicers and tables and then as you add granularity to those tables you can figure out where the data points are actually failing or having the wrong information in them. So a lot of times I’ll do that additionally. I’ll build out data quality large pages where I do

49:21 out data quality large pages where I do data investigation right there. So for me that’s helpful and helps me speed up the the the there’s a knowledge increase between me building the report and me building the visuals and and the data that’s in the columns. So sometimes what’s in the columns but sometimes you don’t. And so you’re like I don’t know what’s in these columns and I’ll just make I’ll just literally drag columns on the page and just see distinct lists of everything that’s in there. Okay, great. Okay, yep, I see this. Okay, these are the date ranges. It’s so hard to do when you have a

49:51 It’s so hard to do when you have a normal canvas site. Correct. Because you put like two visuals down, you’re like, I’m done. I’m out of space. The text isn’t small enough. So doing that large canvas really zooms you out there. So that’s another technique I think that points to their number one, which is wireframing, prototyping, getting familiar with the data. So zooming out, making a bigger canvas. That’s another one that I really like to do. All right, Mike, I’m going to let you choose because I’m I’m proud of us. Since we’ve started up our time savers, we haven’t mentioned AI at all. And I I think we deserve a You just ruined it, Tommy. You just now

50:21 You just ruined it, Tommy. You just now you got my mind thinking about it. It’s going to now you’ve ruined the whole thing. Tommy, I’m going to let you choose. Do we continue down the route of,, like caveman style, which is preai, because I have one there, or would you like one around FCP again? And I’m going to let you choose because I can do both. I can do one or the other, my friend. I’m gonna I say we just keep There’s a couple more comments here in the thread here that I’m gonna go through, I think. Okay. So, one off the

50:52 I think. Okay. So, one off the prototyping or testing data out. Some people were mentioning upstream. I’m going to steal one from you that you talked about a long time ago., this is so I don’t know how long if you give a hint on the podcast how long until it’s publicly available to use, but I’m going to use yours. Okay. Okay. When you’re trying to not only look at the values in the data, but trying to diagnose a measure, it’s really hard, and you’re going to know where I’m going with this in the next sentence. It’s really hard to do when you’re looking at the full model. Like why is that? Hey, that measure

51:23 Like why is that? Hey, that measure should be 15, 000 but it’s 1430 in this case case and you build a minor version of the model. model. So I will because this again this is your tip is I’ll build the same model schema but each table has like five rows of the data. Okay. Okay. Rather than having all Yeah. Yeah. Yeah. This was one of my tricks. Okay. Keep going. Y so I like I said it’s now in the public domain since it was like eight years ago when you mentioned this but yeah so again rather than trying

51:53 but yeah so again rather than trying to to huge huge developer trick on this one time saver too. Yeah man Yeah man talk about this first maybe it was you I thought it was you I don’t I use it I’ve talked about it but I believe maybe this is a technique that I’ve pulled from Chris Webb potentially he’s a genius so like ahead of everything. Yeah, but so let’s dive into this. So again, if you’re trying to say why is this measure not showing the value of whatever in your normal model

52:23 value of whatever in your normal model and sometimes it works sometimes it doesn’t go ahead filter the data to have only five rows in the facts whatever just distinct the the the distinct values that you’d want and then just try to say I want this to calculate the four then you can really figure things out go oh because this one particular value is affecting everything else and the amount time I’ve saved building DAX, which is probably where the bulk of your, you probably where the bulk of your,, effort goes into for me at least.

52:54 know, effort goes into for me at least., you go into this, Mike, and you’re realizing very quickly what is D, you can diagnose quickly where the context is coming from, why the context is the way it is, and you can test it out real time without having to go, okay, most things look right, but some things aren’t. I know that there’s five values normally. I can test it out. And then you unfilter, you can test these out. Not only is that a timesaver, but I think like to your point, it’s

53:24 think like to your point, it’s it’s hard not to do for me now, if that makes sense. It’s hard for me not to actually do that in my normal workflow. So the the gist of this one really is, and I believe I lifted this one directly from Chris Webb, but the the idea here is use a parameter in your model. So when you’re looking at your model on your local machine, right, limit every row of table to a thousand rows or filter down to a thousand rows. So that way your model doesn’t get super big or even 10, 000 or even 100, 000 doesn’t doesn’t even matter. You you want to work on your model on your local

53:55 want to work on your model on your local machine with a limited set of data when you move it to the service. You may want to unre remove that flag, right? So what I’ve done in the past is I use query folding. So query folding bundles up a query especially when you’re coming from like a SQL data warehouse. You’re able to fold those queries back up to a source system by adding the filter context there. You’re able to cut down the data. So your local desktop file is still,, a couple megabytes in size, but when you publish that file to the service, there’s a lot

54:25 that file to the service, there’s a lot of things that go right here. One is you’re not waiting a ton of time for models to refresh. It’s much faster. So when you if you’re making a change to the data structure, it almost refreshs

54:36 the data structure, it almost refreshs instantly because you’re only grabbing the first 1, 000 or 100, 000 rows of data. Awesome. Second thing it does is when you’re uploading your file to the service or trying to publish it, the published artifact contains data in it and then you can just publish it and that is smaller as well. And you go to the service and you refresh it, then you can walk away, go get your coffee, and it spends all the time loading all the data. Just be to be clear, this is a parameter that you’re setting locally to make it a smaller amount of data. When

55:07 make it a smaller amount of data. When you publish to the service, you go back into the service, change the parameter and say, and say,, load all the data. And then the model can then chew and eat and load all the data. And so now you’re not burdened by this file that’s continually heavy on your local machine cuz every you’ve seen this Tommy the large data models on your local machine any measure change you have to sit and wait if you’re on the canvas view another time saver actually. Yes exactly the table the data view or the faster super quick. Yeah.

55:37 the faster super quick. Yeah. Yeah. But again it’s it’s the idea of like they’re not batching calls until you go to these other different views. So your your experience like the amount of time you waste just sitting for your model to like in the desktop to just get the change into the model is astronomical. people are probably wasting hundreds of hours a week just waiting for the model to do something while you’re making changes to it. Like don’t do that. There’s a word of caution though. I can hear Greg Baldini’s voice in my head as you’re speaking. The

56:08 voice in my head as you’re speaking. The voice of reason the voice of reason. The voice of Dax to me. so the caution though with the parameter view if you’re going and then say hey once I’m done I’m I’ll publish in the service change it to all the rows. Yep. Yep. You Mike there are certain functions in DAX that can eat up time in terms of rendering and actually doing the query. for example, you the I my only question with that time saver is that can get you into some bad habits

56:38 that can get you into some bad habits with the types of functions that you’re writing because with a thousand rows it’s quick but if you have a million or five million rows and you’re doing this filter iteration or multiple filters in the same measure that’s going to be really slow. So you’re not going to know that though until it’s published. So, I would actually recommend another step where just test out those measures in the desktop to see yeah,, how long does it take to render? I I like what you’re saying, but it’s you can fall into the trap of, oh, this

57:09 you can fall into the trap of, oh, this works. I didn’t know it’s going to take five minutes to run with all the data, but but Well, it’s a difference between doing the data modeling work and doing the report work, right? So, yes. Yeah. Some of the report work doesn’t need all that data, right? Yeah. So, if I’m just doing the report work, like I don’t really need all the But to your point, Tommy, like you’re right. If you’re doing DAX and data modeling, there’ll be times when you need to open that window up and maybe it’s not 100, 000 records, maybe you need to go all the way up to like a million records and go there. So, yeah, yeah, but at least that way in your desktop file, you can control the amount the

57:39 file, you can control the amount the volume of data you’re playing with, right? And so when you’re doing heavy things like to your point, if I’m testing something with a copious amount of filters and I need to do very complicated DAX things, well one I would argue why is the complicated DAX there to begin with. If you’re doing stupid complex filters, you probably check your model. Check your model because you you might need to go back to the model and start figuring some new things out. Yeah, you might want to be pushing those calculations upstream to your proper system, SQL or Spark or notebooks where

58:09 system, SQL or Spark or notebooks where you can actually use more horsepower to calculate those things., , let me give you my last maybe tip or trick or time savers here., I think people were also alluding to this one. Star schema is very powerful. Is it a timesaver though? I would argue it would be because would argue why else are you what else are you doing? People build all kinds of dumb stuff, Tommy. Like people try and build like invoice headers and invoice details and put them all in separate places and not

58:39 put them all in separate places and not fat not flatten them out or not. There’s just a bunch of things that happen that you’re like, what, I’m not sure if this really makes sense., but I think leaning harder into the star schema idea, but I’m going to go even a little bit further and say domainbased modeling is probably also a a substantial time saver. And the reason why I say this is because when very large models are given to new users in PowerBI, lots of tables, lots of information them. It’s very difficult for users to figure out which columns

59:10 for users to figure out which columns and which tables are supposed to be used together to work together as a team. Yeah. Yeah. The more tables and things you push into a single PowerBI semantic model, the more people get confused, period, unless you have really good documentation, and that very rarely happens. It’s getting better with AI, but it’s still very rare. So, I would argue think really hard about trying to conform as much as you can to the star schema. You can have two or three fact tables, that’s fine. Think about how you’re

59:41 that’s fine. Think about how you’re organizing the information on the page, like the tables, right? So, another another hack that I use to save myself a ton of time is if I’m looking at models, I’ll do the waterfall approach. I got this from Rob Kie in his book something about Power Pivot Power Pivot Pro one. Power Pivot Pro. Yeah. Yeah. And so, he puts all his dimension tables at the top of the model and all the fact tables at the bottom and he watches all the relationships go from the dimensions into the facts. I use this technique all the stinking time and it helps me get through all the different details and items that are there., so that’s a

60:13 items that are there., so that’s a technique that I really like to use. And then we we start focusing in on, okay, what is the intent of this report? What are the least amount of factual tables I can present to this report to make these groups of reports work well? Right? Right? How can we use that?, and so I focus our a lot of our attention on that kind our a lot of our attention on that conversation. And then I’ll build of conversation. And then I’ll build another domain model specifically for a different group set of reporting. So a lot of this is tell me what reports you need. Mhm. Let’s lump them together in the same domain area. From that domain, we then

60:45 domain area. From that domain, we then dictate, okay, what columns or dimensions and factual things do we care to calculate? Doing that saves me tons of time because when the model gets big, I just spend a lot of time debugging like where did this birectional relationship come from? Why is this data not working right from this table to that table? Maybe these two tables shouldn’t be joined together. Maybe I should just simplify my model so that it actually works better. So like there’s all these like I spend so much so many cycles around trying to fix just nasty yucky models and like let’s just throw this out. We should have just made a good model from the beginning. What are

61:16 good model from the beginning. What are we trying to calculate? Let’s start from there and build out the dimensions we need. need. I’m going to I’m going to say my closing one too and it’s going to be really just an addition to that or just parallel with that. I swear by the views in the model view especially when you have a lot of tables 10 plus tables you I I don’t know why people don’t use this more but creating different views of the relationships so in the model view in PowerBI desktop you can have different tabs which you can see oh I just want to see dim

61:47 see oh I just want to see dim product to dim sales okay the these are the relationship for dim product and being able to organize that helps out so much to go oh what does product what does team relate to what are all the complex thing with quote and budget and it’s just a simple view rather than trying to take a look at everything at the whole time Mike before we close I want to give some honorable mentions things I’m actually surprised we didn’t bring up but things that that have been time savers a little more recent and I’ll just lightening these DAX query view a quick way if

62:20 these DAX query view a quick way if you want to see your data like the visual side it’s a really great way to understand what your data is doing. tindle view if you actually want to quickly add things again you gota learn it’s a little spend time to spa spend time to save time but I think those two things are awesome and then really this was mentioned in the comments spend time on power query if you still using power query to learn some of the functional language spend a lot of time it’s going to save you a lot of time so those are my

62:50 of time so those are my honorable mentions honorable mentions for mine would probably be use more of the model view. The model view is actually way more powerful than I think working on the report canvas. I see a lot of people just working and building the models and measures on the report canvas. Model view is way more powerful now. I really like using that a lot. another super big hack for me is when you need to edit multiple measures or multiple things at the same time. Tindle view is incredibly powerful. So TMDL find replace you can do a lot of really cool pattern

63:20 do a lot of really cool pattern matching. matching. That’s a good one. That’s really good. Yeah. Yeah. Yep. Another really big time savers for me is when I’m in Timle, it has a really rich text editor and control shift L will let you select multiple items of the same text format. So control shift L inside the editor in Tindle allows you to select a piece of text and select every occurrence of that text. Multiple cursors is a huge timesaver. It’s so wild. It’s a little hard to get your head around it, but once you figure out how to make

63:50 it, but once you figure out how to make it work, it can really make you very powerful inside in the formula bar is control shift D. I think think that’s that’s select the next occurrence of the item. Oh yeah. Oh yeah. So control shift D is select the next item that is the same formatted highlighted text. Control shift L selects all version of them. And then you can then use this to manipulate and then you can actually add multiple cursors. So multiple cursors in code editing. Super super crazy., absolutely love it. So, that’s another really very technical detail there

64:22 really very technical detail there that’s also a big time savers as well. With that being said, I think we’re at time. Tommy, this has been a really good episode. Hopefully, some of these techniques that we’re using and talking about here, you either have used them or you’re going to use them., we recommend you just,, take these as a grain of salt and figure out how to incorporate these into your workflow. That being said, Tommy, where else can you find the podcast? You can find us on Apple, Spotify, wherever you get your podcast. Make sure to subscribe and leave a rating. It helps us out a ton. If you have a question, idea, or topic that you want us to talk about a future episode, head over to powerbi. tips/mpodcast.

64:54 over to powerbi. tips/mpodcast. Leave your name and a great question. And finally, join us live every Tuesday and Thursday, a. m. Central, and join the conversation. All PowerBt tips social media channels. Awesome. Thank you all so much. Appreciate you very much. And we’ll see you next time. [music] Be it high. Tommy and Mike lighting up the sky. Dance to the day to laugh in the mix. Fabric and A. I get your feels. Explicit [music] measures. Drop the beat. Now kings feel the crowd. Explicit

65:26 beat. Now kings feel the crowd. Explicit [music] measures. Explicit measures. [music] Drop.

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.