It Depends – Ep. 203

The fastest way to get stuck in Power BI is to treat every tip you read as a universal rule.

In Ep. 203, Mike, Tommy, and Seth unpack a great article from Data Goblins that reframes a common frustration: best practices are broad, generally safe guidance—while optimizations are situational techniques that only make sense once you know what you’re optimizing for (and what you’re willing to trade away).

If you’ve ever heard “it depends” and wished someone would just give you a straight answer, this episode is the missing follow-up: it depends on what? Requirements, model shape, scale, refresh patterns, and the outcome you’re trying to improve.

News & Announcements

• It Depends (Power BI Best Practices and Optimizations) — The companion article that sparked the episode, focused on critical thinking and separating ‘always do this’ from ‘do this when…’ decisions.
• Explicit Measures Podcast — Subscribe and browse the full episode backlog.
• Submit an idea / topic — Send a question or scenario you want the team to break down on a future episode.
• Mike Carlo on LinkedIn — Follow Mike for modeling, DAX, and delivery lessons learned.
• Seth Bauer on LinkedIn — Follow Seth for BI engineering and measurement strategy perspectives.
• Tommy Puglia on LinkedIn — Follow Tommy for the latest platform findings and field notes.
• Argus PBI — Friend of PowerBI.tips.

Main Discussion

Topic: “It depends” — best practices vs. optimizations in Power BI

When someone asks “What’s the best way to do this?”, experienced builders often start with “it depends.” That’s not a cop-out—it’s a signal that there are multiple valid solutions, and the right one changes based on constraints (scale, model design, the report experience, and how much time you want to invest to chase the last bit of performance).

Using the Data Goblins article as a guide, the team draws a line between:

• Best practices: broad patterns that usually improve maintainability, clarity, and long-term success.
• Optimizations: targeted tactics (often learned through hard-won experience) that can make a big difference in the right scenario—and waste time or even backfire in the wrong one.

They also emphasize a practical engineering loop for performance work: define the problem, try more than one option, measure outcomes (e.g., DAX Studio / Performance Analyzer), and only then lock in a change.

Key takeaways:

• Treat “best practices” as default guardrails—not absolute rules.
• Optimizations are conditional; always ask what are we optimizing for (speed, simplicity, cost, maintainability)?
• “It depends” should be followed by clarifying questions (data source, scale, cloud vs. on-prem, refresh needs, audience expectations).
• Concepts scale better than memorized tips—understanding model behavior beats collecting one-off tricks.
• Test changes like an engineer: hypothesis → multiple options → repeatable measurement → pick the winner.
• Know the 80/20: the last 20% of performance often costs far more time than the first 80%.
• Requirements aren’t static—get something usable in people’s hands early so the real requirements surface.

Looking Forward

The next time you’re tempted to copy a “best practice” from a forum post, pause and ask: what’s the context, and what tradeoff am I accepting if I apply it here?

Episode Transcript

0:02 [Music] foreign good morning everyone welcome back to the explicit measures podcast with Tommy

0:33 the explicit measures podcast with Tommy Seth and Mike good morning hello hello we’re back to an official real live in person all errors and edits have to be taken as it is yeah this show is edited this question ask this question if we’re live we’re actually live far away we are alive maybe we lost everyone that goes live Greg baldini was saying I think you

1:03 live Greg baldini was saying I think you have to say like a special word last time it was kumquat so kumquat is the special word Greg you have to send us a new secret words we can say that we are officially live good morning everybody good morning a lot to happen in the last couple days weeks I guess we’ve been talking it’s been a minute two weeks yeah it feels like it’s been forever a little bit too long I know you guys go anywhere for spring break things

1:34 things for your family oh yeah I did yeah oh you want me to embellish sure yeah you don’t have to like cool good job all right good story let’s talk about 100 in my pocket yeah second you’re heading down into the Tampa region in Florida oh nice and it was so of all there’s this red tide thing I don’t know if you guys familiar I’ve heard I know about it in the ocean it cleared out the week before nice coughing breath so for those who don’t know what red tide it’s like this I

2:04 know what red tide it’s like this I don’t know bacteria algae no it’s an algae yeah I think that kill ends up suffocating fish along the shorelines so like it creates a bunch of dead fish yep yep roll up on the beaches and then apparently it can irritate the lung style yeah yeah like so you have problems breathing and yeah it’s not fun we were reading for months before going it’s just like oh man like what’s that gonna do to the vacation nope clear it out good kids had a great time anytime you go to the beach where the kids can

2:34 you go to the beach where the kids can just run around and play and have fun yeah it was wonderful we we also did a beach thing we went down to Florida as well and we we thought we sunscreened them day one day one is usually the worst day so we apparently we ran them all out of the ocean and them running out of the ocean and the sunscreen like just poof gone it was like evaporated so everyone’s crispified everyone’s like my top of my ears hurt my neck hurts everything’s it was it was a bad burn and I don’t know maybe there’s so much more intense we’re not

3:05 there’s so much more intense we’re not used to going down around this time of the year we usually go much later in the year so so maybe it’s just not as intense at that point but anyways it was crazy a lot of fun and super enjoyed I love putting my feet in nice soft sand that was that’s awesome where’d you go indeed dude I sit here here Tommy went to a Yankees game I watched a lot of YouTube so oh Florida red tide and sunburned I think if you’re coming from Milwaukee you don’t do sunburn like sunburn won’t

3:36 you don’t do sunburn like sunburn won’t it’s not gonna can’t compete with Florida’s Sun oh yeah there’s no sun up here you could not get you could lay outside for weeks and not get sunburned up in Wisconsin compared to Florida like what was the what was the weather I heard it was like 88. it was like 80s all week all week nice yeah it was gorgeous mommy thinks it’s too hot yeah too much sweating I’m not a fan of sand you you were a Floridian right for a while and I don’t like the beach I think it’s completely overrated are you a mountains guy

4:06 okay so anything else not Beach yeah interesting I don’t think I’ve ever found someone who doesn’t like the beach well you do you do a podcast with them so true well there we go you learn something new every day obviously you’re on vacation were you working did you see that there’s a March update did you see all the amazing new things that Microsoft keeps literally turning out oh totally yes it definitely was taking a peek at the March update

4:36 taking a peek at the March update I’ve been trying to so I was partly working I’ll say that so I was definitely was not fully the kids have off school and I don’t have off work work still happens people still got to get stuff done so I was still working on things ways of owning your own company right yeah well it comes with it comes with the double-edged sword you don’t get to actually officially ever walk away from stuff you got to keep people busy so you’re not relying on just yourself to feed you got to feed other people too man so so that is one

5:06 other people too man so so that is one of the things I I enjoy working for an organization because my I I built a team and these these guys are rock stars right so I just Tee It Up and literally I I left on a Friday and Thursday I was like okay guys like let’s walk through this again and like if you need anything here’s here and I basically got the shut up we’ve got it

5:37 check out for for a good long portion of the vacation to be able to do that is so refreshing because it’s just like you you don’t realize my wife was saying this too it’s like you don’t realize how high the stress levels get or how how like you just accumulate this burden right that all of a sudden just can be released as you go on vacation and just like reset a little bit and it’s it’s nice because you can come back and for me anyway you see things differently there’s you look at the business in different ways and figure out different strategies

6:07 ways and figure out different strategies that you may not have seen before and yeah good stuff I had a great vacation excellent very good moving on to some other things that we had some Community articles come up that were pretty interesting Tom you want to pick up the one that you you picked up here that was pretty interesting yeah so this is actually our very own James Bartlett on basically I think he’s doing it’s a new community thing so we’ll send the link but it’s called the data dojo and it looks to be a just another great

6:37 it looks to be a just another great Community Center for people I think it’s going to have contributions in terms of putting workshops together I am all about stuff like this and I I think that user groups should do hackathons again I I think though those are so underrated working with people and trying to figure out those types of problems I know it’s not as easy with something like power bi because only one person can be in it but at the same time I am all four things where there’s

7:08 I am all four things where there’s workshops where there’s the contributions like with the the Wows workout Wednesdays but especially when people are working together so looks like a cool thing so definitely check it out yeah I I liked it from the standpoint so this is this is James’s Journey right that’s just started with him creating a community of practice right within his his College organization and well let’s be clear he’s not in college he works for I don’t know correct right faculty The Faculty right exactly yes those that he’s

7:39 right exactly yes those that he’s working with anyone who’s working with data things which is really cool he’s going through his experience so so these like this is I think the first and and then in a series what I love about these series type things where a lot of what you read in even articles we write come from an area like we’ve solved this problem here’s how it’s done with the people problem right which is power bi adoption and a lot of road map stuff and things that we’ve talked about there there’s when you introduce people and environments things change right so

8:10 and environments things change right so the fact that he’s taking people along for this journey it’s not like hey we figured it all out and here’s how we did it he’s just ER it’s the early preview right like as they’re going along he’s he’s describing what they’re attempting to do how it’s working out and teasing into the next like next one already so I I love it looking forward to how it goes for them and also being part of their learning process and being able to adapt for are you style

8:41 being able to adapt for are you style your use cases so good stuff yeah I really liked it and I think his community pieces here the couple couple let’s call it rules of the road he’s he’s at an interesting term for his community he’s calling it that did the data Dojo which I think is fun a neat little play in words there as well and then it’s because some of the values here just very fun about community and I love how you’re pushing this right anyone interested in learning you’re you’re a member right come on in if everyone’s welcome attend as many or as few as you like

9:11 attend as many or as few as you like again the idea here is I think if you’re adding value people will show up so if you just regularly add value and people enjoy working around with you makes total sense a couple other points here towards the end there are no instructors and there are no students we’re all here to learn I think it’s a great value I think it’s really a good way of approaching it and then the only well everyone says oh this is a stupid question I would every time someone says that in a medium like I have a stupid question like there’s no stupid questions everyone’s everyone was there at some point

9:41 everyone was there at some point everyone doesn’t know something let’s help you clarify let’s get your mental model right but there’s there’s the only stupid question in the data Dojo is the one that goes unstated which I think is also really good or unasked yeah unassed that’s I like that I think that’s a really good way of approaching the situation so James great article really good I put an article in the link here in the chat so if you want to go read more about what is happening at the Des Moines Iowa user group and how they’re doing community go check it out it’s a really good read and I think there’s going to be more coming I think this is like part one of a couple parts that James been

10:12 one of a couple parts that James been formulating in his mind on how to build more of a Community Practice and actually James I Gotta Give mad

10:18 and actually James I Gotta Give mad credit James has been a pivotal person on helping out with the business Ops tool that we develop and hopefully we’re cleaning up the tool again adding some more stuff and there’s going to be another release Here coming out hopefully in the next couple weeks so working towards another awesome release on that moving on ahead let’s talk about maybe our main article for today oh it’s speaking of which I’m actually doing uninstall right now

10:50 I think the next months of desktop just released something April 6th it’s probably one of those minor updates what do you think oh it’s got to be I got it I was in desktop this yesterday afternoon and I was just playing with some data getting some examples going and I just had a little screen pop up that was that was saying hey update your little the window in the bottom update your desktop like really am I that far behind maybe I’m still in February I don’t know I’m updating now I’m not sure how far behind I am so I’ve got to go check it out and see what happens with this new install

11:20 see what happens with this new install yeah there’s a a minor update on 4-4 but it’s I don’t think it’s like the April release gotcha okay so they didn’t get it out early for some reason yeah okay I’ll check it and I’ll let that being said sorry one more thing I have been have you guys played around with the new data Gateway experience is it it’s newer because it’s been around like that for about almost a year with the new permissions and the new it’s like unified between both powerapps

11:51 it’s like unified between both powerapps and now power bi okay so we just did it we just did it in a hot minute since I’ve installed a new Gateway well this is part of the Gateway this is the channel well the Gateway the gateway’s always been there but I think now Microsoft is unifying a bit more of the experience around how it looks so if you go to the data Gateway in powerbia. com it looks different than it used to there’s a bit of a UI change and now that UI matches what’s already in power up so maybe they started with power apps

12:21 power apps managing gateways in the service yeah yeah okay and out of curiosity for almost a year out of curiosity does your managing gateways say when you go to managing gateways what does it say in power bi. com service for you does it say manage gateways or does it say manage connections and gateways connections and gateways and then the data preview okay because I had someone in our class say it was they didn’t have managing connections and gateways so it indicated to me that something was changing and

12:51 to me that something was changing and they were doing something different there well there is a tenant setting actually that says allow Microsoft to basically experiment and do a b tests yeah so yeah I think that’s maybe what I’m partially seeing there because I have the when you go to the manage connections and gateways I have a now a new window that’s called Data preview which is interesting because I think that’s new I think that’s been around for almost a year this is what I’ve been working with

13:23 I feel like yeah maybe they’ve been doing some tweaks to it next time hey maybe I should do more games how often does it happen all the time right like how often yeah you cycle around different parts of the product and it’s like hey it’s new to you like wait a minute something looks different here yeah I just I just I just one of the articles that I hadn’t seen before on the blog pin workspaces dude that’s awesome yeah it’s a simple feature but I love it I’m very happy for that one scroll scroll scroll through to all those workspaces and now I can just

13:54 all those workspaces and now I can just like 100 give me give me my my top ones that’s a that’s a quality of life Improvement there I’ve been making something changes too to the service like these like slow little updates so all right so I just finished my install yeah it is March 20 to 23 but it is a newer minor version of March 23 so they must have been fixing some bugs in there or something oh I got excited Tommy Tommy right again

14:26 right in what I’m thinking this is the podcast Tommy’s wait it’s just become a change is an assistant theme Here Tommy is right again I’m just here yeah that’s amazing excellent all right let’s move on to our main topic for today so I was at when I was at SQL bits just so happened my schedule rolled out where I was up early just how just how my mind working I was just busy doing things again always working

14:56 doing things again always working because I’m a business and that’s kind because I’m a business and that’s how you were saying sadly you’re of how you were saying sadly you’re never really off you don’t really take a break so I was always doing stuff in the morning but while I was there there Bueller was actually at the guy who runs data goblins was always there at the conference Pro it just so happened our schedules seemed to work out where we were both upper and usually eating breakfast together so a lot of videos a lot of talking early in the morning so with that we’re jumping into his article today which is called it depends power bi best practices and auditions

15:29 power bi best practices and auditions so let’s kick it off here Tommy what do you think what do you you hens a lot we talk about things we’ll take this one today well first off the term it depends goes on our official podcast t-shirt right next to a game changer so Game Changer and we use it all yeah we we’ve used it all the time but I love this article and

15:59 all the time but I love this article and honestly I could really see this being it’s gonna be a very hard to even break this down in an hour but really so what Kurt really broke down was we know what best practices are we’ve talked about best practices and no calculated columns and have a star schema and like the the process that you do both from as the developer but also what you would do in the organization from adoption here’s your best practices and really Kurt kind of I don’t want to say throws a

16:30 I don’t want to say throws a wrench in it but he’s like okay yes there’s best practices but we’re dealing with a lot of different situations where there’s also something that optimization techniques where we need to find a solution here where it’s maybe putting a filter on a card rather than putting the Dax into it which serves the same purpose or just doing in a sense the the things that you’ve learned over the experience that can clean something or fix something or solve something but it wouldn’t be considered best practice and there’s a

17:01 considered best practice and there’s a large distinction between the two and I think the the other point that I think is the the hot I don’t say the hot take but the where I see the rent here is listen just because something’s best practice doesn’t mean that it always must be followed and I think especially in our world it’s not always the same input it is completely different when it comes to our models and the situations so we have to be willing to adapt so that’s my take from it I am very

17:31 that’s my take from it I am very interested in what you guys first got out of it I over overall I think this is a lot of times I on a really high level I I love this article because one of my biggest pet peeves or you call them like things where he flies off the handle or something where people overuse the term best practices

18:01 overuse the term best practices right like everything’s a best practice everything so I stopped using it like because it was just not everything is a best practice just because something works it it doesn’t mean it’s going to work everywhere but I didn’t wrap my head around like separating out okay what what is a best practice versus something else and what what Kurt’s bringing for is like it’s an optimization right it’s Unique to a certain a specific use case or there are methods that you can go about and figure

18:32 methods that you can go about and figure out optimizations that are going to be different whereas best practices typically apply to almost everything right there’s a general rule of when you do these things there’s a benefit to doing them and that’s why we do them across the board versus something that is is objectively valuable to perform on a report but it’s only specific to your your use case and for me just reading through that and having

19:03 through that and having like following along with his thought process in breaking it out logically was worth the read in and of itself because I’m like what this makes a ton of sense and now I can not not have this meaningless aversion to best practices I now have this thought around making the distinction between what is actually a best practice versus an optimization and and Kurt brought me along for the ride so great great article and and

19:34 ride so great great article and and that’s like on a high level what I really enjoyed about it so I I really like so I think these I think your point there so I think I really like your idea of best practices are overused because there could be a ton of these things and they may work in certain scenarios but if you talk to Greg baldini there will be patterns that are best practices that don’t work well in all scenarios and there are other patterns that work better on larger data sets versus smaller ones and you’ll find that there are different patterns across different solutions so I

20:04 I I don’t think I would have ever abandoned fully the term best practices but I would say there’s a level of understanding that needs to occur when you’re working with data modeling particularly when you understand what the model is trying to calculate calculate you can you can you can you can see and anticipate what the engine will do and it’s like an understanding level right so I think best practices are what I would quantify as you now are starting to understand

20:35 starting to understand how the model works and how it works

20:37 how the model works and how it works efficiently right so again Kirk goes through some things here like using bi-directional relationship does not really you don’t want to use that one it’s an anti-pattern we want to have a single direction that Dimensions filter facts those things and I would also echo here especially when you’re talking about best practices star schema I think is one of these items where it’s in an ideal state it works really well but a lot of your models don’t always adhere to a full very clean star model design it’s difficult to get everything

21:09 design it’s difficult to get everything into a start model so I think there are as where you can bend the rules a little bit and I think that’s where I like where your shirt takes it for the optimization techniques because you the optimization techniques because now you’re looking at okay what in know now you’re looking at okay what in your scenario I’m trying to optimize what this model is doing and so it’s a bit more of a narrowly focused problem and solution I don’t know what you guys think I I think there’s two areas here when when we say best practices maybe and

21:41 when we say best practices maybe and Seth I’m with you in terms of I’m using best practices less and using more principles or that terminology but there’s two areas here in terms of the developer role of best practices and optimization and also from the bi adoption world the bi admin role of the center of governance that that curve goes into where in both situations there are there’s a there’s a Microsoft road map a documentation Mike

22:11 road map a documentation Mike you run the world’s best governance program and possibly the only one but but no but like you have an incredible course that’s all around best practices and really like learning the knowledge which is so needed but and obviously if you can get certification in power bi or you can read all the books all those are about best practices but when we’re in the situation and we’re dealing with either the variables

22:41 we’re dealing with either the variables of the people or the variables of the model and this situation we need to solve usually it’s never going to fit or does the majority of best practices don’t necessarily always fit that box from from Soup To Nuts so to speak maybe there’s like a base you can start with but I think what I’ve seen a success is being flexible and adapting so I like later on in the article where

23:11 so I like later on in the article where Kurt talks about some possible examples of optimizations right and these are some patterns that you find or or things that work to make things things optimized right if you’re going to use a distinct count or distincts on top of columns sometimes that’s slower than just using a sum X instead right and there’s also other things I think Greg baldini has done some you think Greg baldini has done some Dax hacks as well like if you use a know Dax hacks as well like if you use a divide function versus some other functions I can’t remember what the calculation was Greg you’d have

23:41 what the calculation was Greg you’d have to correct me if I’m wrong here if he’s online but if you use some other functions it actually is faster to iterate through the math than it is to do a distinct or unique count across the various elements in the model another example here is using IF eager instead of just if or or disabling cross filtering in visual interactions right so turning some of these things off optimizes what the report’s doing so that it makes it a bit easier to work with and you can actually speed up the performance of the report you’re optimizing a bit more those may not be best practices though

24:12 those may not be best practices though these may be more of like the a little bit of a hack to get it to work correctly and this is maybe to some degree when we were making entire navigation panes on the side of our report and we were using like that was a bit of a hack right that was an optimization to get a navigation element into the report there was no way to do that but you had to make all these bookmarks all these other additional Elements which were slowing down the report to get the functionality that we wanted right so that was maybe an optimization for user ux UI but yeah it was it was it’s not

24:46 ux UI but yeah it was it was it’s not really a best practice right now we have tools like that has I think by people building those things that has forced the hand of Microsoft and now Microsoft says hey this is actually something we can solve by the way here’s a bookmark Navigator here’s a page Navigator so now you have these navigation Elements by default I I was literally just dealing yesterday with a project and we’re dealing it’s a template from a data provider and they’re like could you add a button to the page like another button because I’m trying to add this page like well

25:16 trying to add this page like well we can’t they’re like why I’m like because that those buttons are actually an image so yeah and so I there’s all these again calm hacks but I think it’s there’s underlying principles that obviously should be applied but there’s so much that whether it’s just the need or the necessity that we have to adjust to let’s be real these these aren’t hacks these are Creative Solutions to

25:47 hacks these are Creative Solutions to problems that like we have to solve that the tool doesn’t support right exactly I like to think about those in in that regard I I like what Kurt does in terms of like defining best practices what what those are at the beginning of the article defining what an optimization is what he says is best practices are methods to address a problem recognized as Superior to Alternative approaches right and and the way I read this and he’s got a whole chart of different areas of examples of the two different

26:19 areas of examples of the two different flavors of best practices that he defines and I and I like this as well which is the do and then or the the pattern and then the don’t like the anti-pattern of things like things we don’t want to be doing what’s what’s interesting to me and makes a lot of sense is the best practices patterns that he is either saying are good patterns or anti-patterns are on a higher level which I would expect them to be because if we’re going to make generalities and say these are some really good things

26:49 say these are some really good things you should almost always do right or all of your reports should follow these processes those are have to be on a higher level at least from my perspective because when we start getting into specific optimizations right or specific things that’s where it falls into this optimization right and that’s where maybe some of my knee-jerk reaction has always happened previously because a lot of the times we fall really quickly out

27:19 of the times we fall really quickly out of best practice patterns right into very specific use cases or patterns that only work within the context of my model or the reports that are connected to the same model and in that case I think he even opens this door which is applicable which is sometimes these optimizations are actually against a normal best practice right so there’s this Nuance between defining what you

27:50 this Nuance between defining what you this Nuance between defining what we would say is a best practice and know we would say is a best practice and and just some of his examples keep it simple so three bullets in there use straightforward clear and concise models and code simpler Solutions are easier to maintain and understand Occam’s razor seeks Solutions with the least number of elements right yeah I like that high high level but at the same time like these are great framework things that you would want to or I should say you should spend time ensuring you’re walking through

28:22 ensuring you’re walking through or or have this list of to Do’s when when building reports or models and if you’re going to deviate make sure it’s a specific deviation right like in and or I should say an intentional deviation away from a best practice or something that we would typically say you should do everywhere but it may not work in this case because maybe there’s some requirement that forces us to get what I just say first is us to find

28:53 what I just say first is us to find Creative Solutions right and I that’s a really good point because I think I ha I think there’s a difference between hack and optimization I think hack is you’re occurring a bit of tech debt because it might not be completely tested in in more than one situation honestly I think optimization anything that you do from the optimization point of view is at least follow principles of it’s got to be scalable it’s got to be reliable and up and it’s got to be optimal

29:23 and it’s got to be optimal and I think that’s the difference with a hack where it’s maybe this one use case and you’re just trying to figure something out and getting back to it but optimizing something where way where it may not actually fall into any documentation it’s still something that you can trust to run trust that if it can be flexible with a model changing and you can rely on it and I think that’s the difference between hack and op and optimization I I go back and Kurt had some other

29:54 I I go back and Kurt had some other really interesting points here I think to along those lines that you guys are both talking about here he talks about a section here about challenges for Learning and teaching and I thought this was a really insightful piece rules are easier to memorize than Concepts concepts are harder to detect and understand as the paraphrase that I’m reading there and I thought that was really interesting Nuance is frustrating when there’s nuances to things and I think that that makes sense that’s why there’s a lot of answers like hey it depends it’s it is you

30:26 hey it depends it’s it is you know if you do it this way you would want to design it in this format if you wanted to do it that way you design that for it so I think that’s the idea here is understanding conceptually what’s occurring makes more sense because then you can handle more of the nuances as you’re designing measures or Pages or whatever right so conceptually again conceptually right we want to use Gestalt principles right conceptually we want to have less things on the page because too many visuals

30:53 on the page because too many visuals make us two smaller people get overwhelmed right conceptually we have a direction on what we’re going to do for design but when we step back yeah okay we put an extra couple of visuals in here fine right but but generally we know what the concept is so that that gives you a little bit more flexible like every report doesn’t only have to only have three visuals on a page right right like you can you have that the ability you understand the concept and then you’re able to pair that with what you need to get

31:23 that with what you need to get accomplished so then again it does provide that little bit of extra flexibility I thought that was really good that was a really neat point I I agree with you where my mind was going in several different directions because business intelligence and like report building and the things that we do is anything but paint by numbers right and and for folks that come in with very specific questions all the time

31:53 specific questions all the time expecting that it’s going to be this repeatable thing that they can just like copy and paste into different areas it’s not the case right and and especially with with understanding what the model looks like how did you set things up do you have an appropriate Star schema all of those are the consistent questions that are asked on the community like I I go back when I when I was heavily involved in answering questions in the power bi Community I literally had the same set of questions every single time because it set the base

32:23 single time because it set the base context for me to understand like where the question was coming from but without those I would be lost I wouldn’t know right and and I think where where my thought here in the Nuance is frustrating part would be was around like certifications right and at first I was gonna be like this is why certifications are bad but at the same time maybe this is why certifications are good because if I think about my own experience related to SQL power

32:54 experience related to SQL power bi was slightly different because I was so ingrained in the material taking certifications like okay none of how the problem it’s a checkbox like however we did it right but however for folks who are are using power bi but don’t know the scope of the program don’t know all these other things and are just chewing off little bits and pieces at a time a certification or like the process to Go Achieve one I think gives a goal for somebody to start consuming

33:24 somebody to start consuming more information that gives them the context where they start to understand the Nuance what I’m saying so because it’s different I’m always a proponent of it’s different to go just solve the single singular problem that you’re trying to solve to build a visual and then you go to the next one and then you go to the next one and then you go to the next one and you don’t understand the context of what you’re trying to create with the Dax measures or the model itself Etc whereas

33:54 measures or the model itself Etc whereas learning in a different way along with the specifics is always a good thing to do right and I’m a big picture thinker so I have to get that big picture first right just so I know where I’m landing in the ecosystem and then I go drilling typically right but a lot of the times we can’t do that and that’s where I’m saying a lot of this Nuance can be frustrating but you you mitigate that with trying to increase your knowledge base around everything that we can be

34:25 base around everything that we can be doing because then it opens these doors for oh I’m learning about this Dax calculation and it’s solving this problem but I understand that it’s it’s specific to this model like this person’s problem but at least it’s in the back of my head then that I have a pad a new pattern that I saw somebody use and I like I can adjust it to my own use case as opposed to what I see a lot of people doing that are newer which is

34:56 of people doing that are newer which is I saw I’m I asked this question it’s very similar and the answer on the community was this and it doesn’t work for me why right and that’s where this new answer it depends or a lot of the optimization techniques and I think a lot of the decks functions and how people are trying to help others solve those equations and functions without that knowledge base is where people get a little frustrated so look yeah I love like that the funny thing is we’ve been talking about like a paragraph in this very long article for

35:27 paragraph in this very long article for like eight minutes here but yeah I love current stuff well Seth the document starts with literally the words critical thinking and I think everything you just said is revolves around that where what we like part what separates I think a skillful developer and power bi with someone who’s real maybe relatively new is you and I may come to the same number in a visual in a card but mine I can trust mine that can be adaptable and flexible to that model and I know that I

35:59 flexible to that model and I know that I can entrust it for probably not to break based on things changing and also if you do certain filters that I can let it run and it can be reproduced I can also explain why that number is doing what that number it is and I love your point about going through the the community but right all we’re doing is just adding to our tool belt it’s like okay there’s that technique okay how how does that work where can that work where does it not work and I think that is part of again that what distinguish really

36:30 that what distinguish really separates separates I think expertise in the space well in knowledge base sure yeah yeah well I think it’s the critical thinking with Concepts maybe that’s everything but I’m going to change I’m gonna change direction here just a little bit I’m I’m thoroughly impressed with Kurt’s his section around performance best practices practices the the this just blows my mind like this the

37:00 this just blows my mind like this the the thought and he puts into these sections is just so good so I’m really enjoying if you scroll down to the section on the radical where it starts talking about performance best practices and starts talking with like the process right so again that’s what you’re talking about earlier right going through through these testing and optimizations and how the best practices work I think a lot of times we try to get into these reports we we lose this idea of we never really Define what was the problem what was this what was the thing I’m trying to solve for particularly when it’s like tax problems or something like that and a lot of

37:30 or something like that and a lot of times Seth you and I have talked about this a lot extensively and I think we’ve talked about in the podcast as well when you get to a point where you don’t know what the Dax is doing it helps to go away and find a solution or remove yourself from the actual model because there may be bad data there may be other weird things go find a way where you can test repeatedly different solutions and I think that’s what what Kurt here is refining Define the problem in its context research and understand all your available options build a solution test

38:01 available options build a solution test it multiple times and he refers to a Dax Studio where you can write DAC statements and test it three four five times run it multiple times and you get more of a a regular distribution of what what that test looks like change the approach test it again now I I love this idea but the challenge here for me is I don’t have enough time I feel like all the time to go into the nth degree for all these testing things but what impressed me the most is when Kurt goes down here to the bottom part of that article

38:31 to the bottom part of that article you start talking about career performance result of test number one he’s got a little simple bar chart great then he goes in query performance test of all 20 tests and does a rain plot distribution with bars on there for I thought wow this is really good and he he spent the time to go grab the rain cloud plot from deneb from Kerry kozlow kolosko sorry I said that wrong and and adapt her visual for the article and

39:01 and adapt her visual for the article and then again he does another one where he has a violin plot showing the same thing on on performance analyzer on the page render time like the the visuals he’s putting the effort he’s putting in this is is really awesome because I really am enjoying the visuals and the attention to detail on okay I have a whole bunch of tests and I need to think about how do I distribute or display that information in a graphical way so I can actually see the distribution of what the content’s doing he’s doing

39:31 what the content’s doing he’s doing another one query performance with changing relationship data key points right or a date versus an integer versus a string and he’s actually doing some performance measures there and I really appreciated the article and the really nice visuals that he’s putting in place here that are helping you visualize and see how he’s performance testing the data from his various objects I thought that’s really cool I I I agree I think it’s great I guess I have a question for you guys

40:01 have a question for you guys how how how much time all the time how much time do you guys invest in testing or see that organizations want to have you invest a lot of time in testing ah well this is a good one I think this for me it’s like law of diminishing return return right if the if the report’s fast they don’t care what you’re right right it could it’s not they’re not looking for you to shave off like if you if if things return within a half a

40:31 if if things return within a half a second second it’s they don’t want you to spend time making it down to a quarter of a second right like the gain on that problem isn’t an issue when you go to a page and it takes 15 seconds 20 seconds to render that’s when people notice like hey this page feels a little bit slow and I think at that point there’s like a threshold right there’s a there’s a threshold on whatever you’re rendering on the page where like hey you’ll you feel it this it’s a little sluggish at that point in time then they’re willing to spend the money or time to

41:02 willing to spend the money or time to optimize and build that stuff I think over time when I observe fewer developers of power bi newer developers of power bi tend to

41:13 bi newer developers of power bi tend to run into these problems fairly frequently and they build bad dacs right away and there’s usually a lot more pages that are just slow in nature because because why are you writing the Decks that way that’s that’s very inefficient right you’re not using an if statement you’re not you’re not cutting out the data you don’t need you’re over filtering or you’re under filtering and doing too much calculations so I think when I look when I step into organizations where we have people that are less experienced in Dax particularly it’s very easy for me to go through and make a handful of changes that immediately speeds up the report so in those cases it makes a lot of

41:44 so in those cases it makes a lot of sense sense I’d also Echo usually when I’m coming into these organizations the data model is also a mess we have like 45 50 90 tables in a model there’s all kinds like again this is one of these it just worked and we kept adding things to it and here we are a year later where we have this Rat’s Nest of tables and relationships and nothing’s designed or planned out so you nothing’s designed or planned out so we have to come in and say okay know we have to come in and say okay really what are we trying to accomplish here what pages what value is coming out

42:14 here what pages what value is coming out of this report and can we simplify that model down and that’s where the real value is comes from so I think to answer your question it depends Mike hold that thought because I want to get to that sorry no no no no there’s something there that I think is very interesting with our conversation today but so when we’re going to say testing are we just saying process only opt like like in in a sense of the like loading time or are we talking

42:46 like loading time or are we talking about accuracy are we talking about a feature functionality when when you say test yeah it was a broad question that I think as Mike was describing providing a good answer I think I would narrow it down right like if narrow down to per the performance okay right and and I’ll I’ll stop there and let you come in yeah honestly with new models and ex it’s actually very similar to what Mike said where it’s going in you go whoa what

43:18 it’s going in you go whoa what looking at it’s like okay where are the problem Childs here what do we need to keep that’s what needs to get removed and then it’s it’s just kind get removed and then it’s it’s just going through that and called the of going through that and called the process but yeah it’s really usually with with models already created unless it is something large that I’m dealing with that it you notice it’s going slow and generally you can pick that up but it’s not it’s not a part of every process no yeah so to me this strikes so

43:49 process no yeah so to me this strikes so this is very Dev focused right like this is part of a Dev best practices thing as we’re implementing what I what I I what I like about it is not having a test process in here leaves you wide open to deploying a change that has unintended consequences right and I guess my initial question was like do we do organizations really spend a ton of

44:19 do organizations really spend a ton of time testing across the board all the scenarios some do I I would argue more often than not like you’re not investing tons of time in terms of like performance tweaking and testing just making sure things operate in such a way that it’s not customer impacting whether those are internal or external right but at the same time from a from a developer perspective or whomever is building the report report understanding these changes and how they’re impacting right are very useful

44:51 they’re impacting right are very useful for your team to understand right because if it’s hey we need to make this modification or we’re having challenges around this particular Dax function or measure having test like doing a benchmark first and then testing the new change to understand the impact it is I think extremely valuable when you can fit that into the cycle and if you’re making a lot of changes I would argue that for sure testing has to be

45:22 argue that for sure testing has to be part of that and it also makes sense like I’m I’m I’m trying to like find this realm where I’m like ah sometimes I do it and sometimes I wouldn’t but at the same time like it does this mic make sense where where no I guess it wouldn’t because each model is different where I’m trying to think of like is there a standardized way in which you could create some test templates or test things that you would automatically 100 sorry every model I think you should definitely do that I think I think

45:52 think I think I would also Echo okay we talk about this a lot right is something governed or something not governed right so in a governed data set something that’s going to be critical used at the higher levels or you’re like certifying content coming from that that body or that that report I think in those cases you’re doing a bit more rigor what you’re describing right there’s a bit more process there’s a bit more rigor around those things other things that are maybe team or personally based reporting I probably wouldn’t spend a whole bunch of time noodling on a lot of extra

46:24 of time noodling on a lot of extra effort there to make sure that those are spot on yeah right it’s so to your to your point right but I think there’s a huge Miss in the market right now there was really there’s really not a lot of tools today that we could use that enable us to write tests quickly and easily across all these things inside the data model not for teller model no not for tablet models like the word data science would be golden yeah yeah and and it it also Echoes I think further Upstream so I

46:54 Echoes I think further Upstream so I find a lot of challenges are the data model is doing a lot of business logic to fix data problems that are coming from the data warehouse somewhere else so I feel like there’s another challenge here that’s also occurring that is you here that’s also occurring that is the data model gets overly complex know the data model gets overly complex because of we don’t control the Upstream data engineering to some degree and you need a little bit more Upstream data engineering to clean out and fix things which then simplifies your model which means less tables which then simplifies your Jacks all the way down stream so it really it helps to have that

47:24 really it helps to have that architect level view of Upstream the model and then what it’s doing in the report and I think Kurt also put out another article more recently around where does the measure need to live does the measure need to live in the model or should we be building more measures in the thin report when and when’s the right appropriate time to to migrate measures that are in the thin report back up into the model again I think there’s some really good thoughts and capabilities around this the technology I think is changing our

47:54 the technology I think is changing our habits on how we think about these things man you’ve just literally put two great points so I I wanna I wanna say one aspect to that you highlighted or you mentioned about testing that I think is very particular to what we’re talking about is you go in and you automatically notice a bad model too many tables Etc I think we could all at least agree and I know we’ve been saying best practices are not what we thought they were but there we can all agree there’s

48:25 were but there we can all agree there’s at least some definite Universal best practices it’s a base things that we can do it doesn’t necessarily apply to the entire solution but call it building your foundation right and I think that’s really important to know and what we’re in and especially what we do where there’s those principles that we should definitely say are best practices where in terms of how we’re loading what we’re loading the relationships we’re creating but that being said Mike and you just said it again so you’re like setting me

48:55 said it again so you’re like setting me up the tool is constantly changing yeah the features are constantly changing so it’s almost impossible to do to implement best practices when the the feature completely changes it doesn’t mean it gets deprecated but that might be a better way yep look again yeah page navigation like we talked about earlier right we had this practice of how you build page navigational there are tool changes boom now there’s a now there’s a single button that you use that does it automatically for you

49:25 use that does it automatically for you like great the March update had query reduction so if you are dealing turn that on yeah exactly yeah all right and so I think there there are underlying best practices absolutely the foundations obviously but I think we’re also in a not only are we dealing with different data sets with different like logic that we have to deal with because this like you said it’s not Upstream but the tool itself is constantly changing whatever principles or processes we’re putting in place

49:57 or processes we’re putting in place so I don’t know what do you guys think in terms of or the old best practices if you were dealing like in a SQL Server database back in the day things didn’t really rapidly change but how important do you think it is for us to adapt to those changes and and also begin to roll those out out I think by Nature we adapt but like the only thing from my perspective that a rapidly changing tool like power bi

50:27 a rapidly changing tool like power bi does for trying to confine best practices is make them more generic right and I and I think if you look at all the do’s and don’ts of best practices that Kurt goes through in the article they’re all very high level and by Nature that’s because the tool changes a lot but if it didn’t maybe you could have more specific best practices what I’m saying like where like even in five like one of his filter early and only keep keep the needed data one of the bullet points prevent loading

50:58 one of the bullet points prevent loading excessive data in power query transforms okay that that’s that’s a functional technology problem we have right now is that something we’re going to have a problem with in a year I don’t know right so those those sorts of changes in the tool absolutely do I think change the best practices or recommendations which also creates some of the problems around like hey the community’s been around for six seven years now and people are still going in looking things at in 20 20 or where the tool didn’t

51:31 20 20 or where the tool didn’t have some to Mike’s Point right we built things differently because the tool didn’t have a lot of the features and functionality yeah totally yeah I would agree with you like yes we have to be cognizant that not only best practices but a lot of materials and answers to questions evolve over time and I think maybe that that is one of the challenges for newer users of power bi is understanding that this tool when it came out was on such a a rocket path of

52:03 came out was on such a a rocket path of of innovation and and rapid changes that once you come like once you look at like holistically like one two three years those first three years years thousands of changes thousands right and it’s slowed down right so it’s like hey if it’s within the last year you’re more likely to find the right answer or something but be cognizant in those forums of when somebody answered the question because it could like today you might walk in there and be like yeah

52:34 you might walk in there and be like yeah but all I have to do is this right yes now yeah right yeah yeah and I wanna I wanna we’re probably winding down here just a little bit for the end of this episode so I think it’s about time to run it for chat GPT what’s this what does chat GPT say about this topic so the question I asked it was what is the difference between a best practice and an optimization so hopefully that’ll that’ll cover part of the topic here and the response is a best practice

53:04 and the response is a best practice refers to an approach or methodology to perform a task or achieving a goal and I like this definition next it’s a proven technique that’s been developed and refined over time to achieve a desired outcome efficiently and effectively best practices are generally accepted and as the most appropriate and successful way of doing things based on a past experience I thought that’s a really good summary right people have tested it and they’ve figured things out and and now this is like

53:34 this is like multiple people have noodle on this idea two years roughly what we’re seeing as communicated as a best practice on the other hand an optimization involves improving efficiency or effectivenesses of a process a system or product it’s the process of making something better faster or more efficient and identifying eliminating inefficiencies redundancies and or bottlenecks the optimization involves analyzing data identifying areas of improvement and implementing changes to achieve the desired outcome and that’s exactly

54:04 desired outcome and that’s exactly what’s in Kurt’s article like halfway down he starts doing this like here’s here’s the process he literally outlines Define the problem statement come up with Solutions research Solutions test multiple Solutions Define which one performed the best like that makes total sense so I this this really I like this part because this really resonates with my engineering mindset like I love the idea of like a hypothesis here’s what I think is true do some things figure out some stuff okay what of the three or four things you tried which one’s the best way to to wrap it up

54:35 best way to to wrap it up and then lastly it says here in summary the best practice is a proven approach to achieve a specific goal optimization involves finding ways to make best practices even better by making it more efficient or effective I thought that was an interesting way of putting it they’re they’re kind putting it they’re they’re interrelated of interrelated any reactions what would you say I thought that was a pretty good answer from chat GPT I would say that’s probably about a seven or eight solid oh hey there you go we should start rating that’s right

55:05 that’s right I give this one it’s it’s it’s a seven seven point four yeah point four sorry seven point four we I I’m creating that model so you go we’re gonna start trading that data set I’m doing that today today is this solid I’ll get a six six only a six there’s a little generic but yeah that’s good okay all right it is generalizing the entire internet to give you the answer so come on yeah well I’m not I’m not giving it a pass so would you give it a six

55:37 pass so would you give it a six wait that’s a lazy man’s Lazy Man’s rating just a solid number okay 6. 2 I’m sending you guys a giant article about why the rankings between one seven is ultimately and efficiently better than one to ten so I would oh that’s something podcast episode I don’t know why how to do power bi but yeah sure we’ll put another job

56:07 because I really really want to ask this idea of it depends is so subjective in a sense because it really is based on that particular person saying that for everyone listening what would you guys say when your ultimate experience and your knowledge how do you navigate or how would in terms of making the right call So when you say it depends you probably have more than one option so they said how do you actually have the skill to be able to say you’re optimizing you’re not packing so I think I think the it depends question comes from I didn’t get enough

56:38 question comes from I didn’t get enough input on the question to understand exactly what your answer should be right you usually it depends upon and then I give you like okay it depends upon these additional factors right it depends are you talking about data that’s on-prem are you talking about data in Cloud because there’s two different solutions from my experience I think your experience pulls on the right solution with the right context the it depends statement comes from I didn’t understand or I don’t have enough input information to accurately give you like

57:08 information to accurately give you like what I think is the best solution of all those those at the end of the day most of my question or most of the stuff that we design is it usually boils down to money right time and money right you can build the most amazing thing you want but if you don’t have enough time or money you can’t get it done or or you’re going to spend too much so most of those decisions the right solution is the one that takes the least amount of time and costs you the least amount of money to run or build so I think that’s what you try to focus on and that I think covers if you use those this is the other scenario we used to hammer on a lot in

57:39 scenario we used to hammer on a lot in the podcast was if this report isn’t making you money or saving you money why are you building the report it’s a similar concept what is the end goal of what you’re trying to accomplish and so I think if when I’m saying it depends it’s usually because I need more information yeah and I would I would say anyone who’s ever said to me it depends Falls it had like has a follow-up right yeah that’s not the answer the answer is that you’re never going to run into somebody and say like it depends and that and

58:11 and say like it depends and that and they walk away the clarifying thing I’m saying it depends because if you’re saying that you want the fastest report in the world like how long is that going to take it it depends it depends on how much time and money you want to invest in here like we could get you something that works exponentially faster right by investing 80 percent of the time but then I think to Mike’s point that he makes all the time like that extra 20 is gonna is gonna take four times the time it took me to get you eighty percent so it depends on how much

58:41 eighty percent so it depends on how much you want to invest right so a lot of the questions that come up require that like require the individual saying it depends to have the experience and they know multiple different answers could could be what you you’re you’re searching for I would say you don’t hear it when people don’t have that experience because somebody would ask a question and they’ll answer it with the only experience or the knowledge that they have which that’s a good point could be the right answer but it also could be the wrong one right and it’s also where

59:11 the wrong one right and it’s also where you get into this cycle of somebody asked a very specific question it gets answered and then they come back and be like it’s not working for me right you see that a lot in forums I would agree with that and another one Mark is is saying something here about it depends on better requirements and I think a lot of times when we walk into reporting pieces the user asking for the report doesn’t necessarily know all the requirements or no or maybe knows them but doesn’t know how to articulate them to the report Builder so that’s why you

59:42 to the report Builder so that’s why you get into this like it depends solution right hey you want the button on this visual yeah it it depends why don’t you just click this other button that’s already there that like blows the visual up you can see all the details like like what I find is if you spend too much time trying to get all the requirements up front you’re wasting time building things they don’t need so I feel like it’s better to get something in someone’s hand quicker and sooner for them to react to because the requirements change as they see and React to what you’ve built and I feel like that’s really the value

60:13 like that’s really the value go fast I love it add value new work item all right I love it thank you for a great article yeah it depends this is this is a great go-to in in highlighting a lot of the delineation between best practices and optimizations it was great read for me a lot a lot of great critical thinking through here and really appreciate the the efforts that we could we could talk about today super good super good article go check it out read it out read the article it’s really

60:43 it out read the article it’s really really well written and has great visuals and graphics and the Goblins are always funny to look at and interesting so great great job on the artwork as always thank you all very much we appreciate your time we hope you enjoyed the podcast we hope you had some fun thoughts thinking about how this it depends topic works in your work World whether you’re doing best practices or optimizations I only ask is if you share this with somebody else if you like the content you had here please let someone else know you enjoyed this this topic if there’s someone who says it depends well then give them the link to this

61:14 well then give them the link to this episode and tell them yeah we know all about that so James you just got here yeah we just talked about your article James so welcome James and okay so with that Tommy where else can you find the podcast well James you can listen now on the audio which will be available every Tuesday and Thursday subscribe on Apple and Spotify make sure you can follow all of our 203 episodes by the way congratulations on over 200 guys guys or you can join us live every Tuesday and Thursday 7 30 a. m Central

61:45 and Thursday 7 30 a. m Central love it thank you all very much and we’ll see you next time

Thank You

Thanks for listening. If you enjoyed this episode, subscribe to the podcast and check out PowerBI.tips for more templates, themes, and BI guidance.