00:00:05.73 [Announcer]: Welcome to the Analytics Power Hour. Analytics topics covered conversationally and sometimes with explicit language.

00:00:14.17 [Michael Helbling]: Hey everybody, welcome. It’s the Analytics Power Hour and this is episode 276. Business Intelligence. It’s sort of like an oxymoron, but BI tools, I mean, they come and they go and somehow we’re still rebuilding the dashboards for the third time in three years. Well, there you go, Moe, fixed it for you. You know, besides providing analytics engineers somewhat Sisyphean job security, I don’t know why anyone, I don’t know if I know anyone who’s actually truly happy with their BI tool. I mean, are we expecting too much? Are we trying too hard to make them fit every possible use case and leads to a graveyard of unused and obsolete reports from the last time someone got hired in as the head of data and tried to build what the org was asking for? I don’t know, maybe newer concepts like semantic layers or AI are going to usher us into a golden age of self-serve data brilliance. Speaking of data brilliance, let me introduce you to my co-hosts, Moe-Kiss. How are you going?

00:01:19.27 [Moe Kiss]: I’m going great and super pumped to talk about this topic.

00:01:23.34 [Michael Helbling]: I know, and I’m excited to ask you how you’re going finally because I keep wanting to with the other co-hosts and it doesn’t really fit.

00:01:29.62 [Moe Kiss]: It doesn’t work, no.

00:01:31.27 [Michael Helbling]: And Tim Wilson, glad to have you.

00:01:34.53 [Tim Wilson]: I’ve shown remarkable restraint already up to this point.

00:01:38.34 [Michael Helbling]: Yeah, I am glad you’re here, Tim. I know you probably don’t have any strong opinions about this topic, but glad you’re here. And I’m Michael Helbig. We also wanted to get on a guest, someone with deep experience navigating these challenges. Colin Zima is the CEO of Omni, a modern business intelligence platform. Prior to that, he was the chief analytics officer at Looker and helped lead that product through its acquisition by Google. And today he is our guest. Welcome to the show, Colin. Thank you for having me. Colin, why do we struggle so much? with BI tools, maybe close it down a little bit more.

00:02:14.52 [Colin Zima]: Yeah. I mean, I think it actually comes down to a very simple idea, which is everyone at some level can do things with data. Some people are just calculating the change on paying for money at the store, and some people are doing hardcore data science and machine learning. But at some level, data is something that everyone is doing. Everyone takes math. And the challenge with building a business intelligence tool itself is that you actually need to build a product that is used by that entire spectrum of users. So you have a CEO that is expecting perfect visual reporting and clear data and just stuff that looks and feels amazing. You have a data science team that probably hates the tool that you bought and wants to run something in a notebook on their desktop. You’ve got a marketing team that just doesn’t even want to use your tool, but is just trying to get their thing done. And then you’ve got your finance team that’s just like, why isn’t this Excel?

00:03:11.79 [Moe Kiss]: Why? Where is the download to CSV button?

00:03:15.68 [Colin Zima]: Exactly. That doesn’t want your tool either in a different way. And I was mentioning this to Moee a couple of weeks ago, but one of these things that you see is with most business software, so if you take Slack or email, for example, Everyone uses it the exact same way. Like you open your emails, you send emails. It doesn’t really do anything else. Like there’s no automation on top of it. With a BI tool, you have a variety of users that has completely different desires in the platform. So some people are frustrated if they can’t get SQL. Some people are frustrated if they do see SQL. And I can’t even enumerate the number of times where I’ve gotten requests on the same day for two opposite opinions that are both antithetical to each other. It’s like, why is this the default? No, like why isn’t the inverse the default? And I think this is actually the core challenge at Building BI is like, how do you make a tool that is perfect for everyone at all of these different levels? And we’re working on it, but like, I think that’s actually the core problem.

00:04:13.49 [Tim Wilson]: But the premise of BI is often that you, of a BI tool is that you just have to, I like the framing of the different users and it’s like the data team or the BI team says, we just need to get everybody to their starting point based on who they are. The marketers, we just need to give them just the right dashboard and then they’ll have interactivity and flexibility. And we need to give, and it just, it doesn’t, It literally never works, and it just instead winds up being, well, it’s the next day, and there’s one thing this doesn’t do for me, and now what do I do? And then it seems like it forks into, it’s the person who’s a hacker who goes in and figures it out, but then it’s the other overwhelming majority who say, well, now I gotta go back to the team that was trying to offload the work from their plate and ask them to do more for me.

00:05:09.50 [Moe Kiss]: Is it about the tool or is it our expectations often of like, I don’t know, like the dashboard?

00:05:17.24 [Tim Wilson]: It’s definitely not about the tool, but I’m just going to say that I’ll put that on.

00:05:22.24 [Colin Zima]: I can give you an example where I do think the tool doesn’t help it. And this is one of the most visceral ones for me, which is if you build a BI tool, it usually is writing SQL. So SQL is a core language of the tool. The whole back end of the tool is writing SQL at some level. And then you need to build these front-end layers on top of it, like something that’s called table calculations or some sort of post-processing language. And inevitably, the users that you’re exposing that language to are not SQL people. They’re Excel people. But if you look at every single tool’s implementation of that final language, it’s always this weird hybrid between Excel and SQL, because you’re trying to bring a back end that does one thing and deliver a front end that does another thing. And it’s one of those examples where if you try to thread the needle and create this half language, it’s like you’re back to the XKCD 15th standard of a new language. And I do think this is one of those tools where everyone thinks they can do it better and they try to reinvent. And again, I’m mildly talking my book because we made our front end Excel and our back end SQL. But I do think that sometimes there’s a learning curve associated because the builder is trying to solve a problem. And in solving the problem, they create new things. And then there’s a learning curve for the user that they need to come along with that makes it hard. And that’s why Excel as this lowest common denominator is always sort of the release valve for every single tool in existence. It’s like, I don’t know how to do this, but I can make it make tables. And then I can go put it in this other app that I actually know how to use. And I do think that at some level that is a failing of the tool stack because the user is saying like, hey, I can’t quite understand this, but I know I can understand it over here. I do think that underlying all of this is like, you know, business is hard and they change and it’s hard to make like perfect metrics. Like the reason that marketers dashboard doesn’t work is because it probably worked and then they changed the definition of a thing. Oh, yeah.

00:07:21.78 [Moe Kiss]: And now like.

00:07:22.54 [Colin Zima]: Now, we’ve got two time series that need to attach to each other. And that’s just hard.

00:07:28.41 [Tim Wilson]: But even going from Eminem Excel, one, I think, fair dividing line is Excel, people who are comfortable and regularly used to the tables and people who don’t. And the concept of metrics and dimensions and aggregation and group by, I’ve getting somebody from If they get pivot tables, which has a deeper level of I get the manipulation of data and I’m like, oh, well, you actually kind of get some intuition around a group buying sequel. You get the idea of a dimension with dropping a metric on it. I mean, there’s a ton of people who don’t get that. And yet they’re jumping into a BI platform. And I don’t know that it’s a tooling. I think you kind of nailed it. Like the promise of the BI tools is we’re going to be everything to everybody. And that just winds up being a feature bloat and you’re imperfect for everyone. And no one’s really even defined what The focus becomes on if I get the right thing in the tool, and I’m just going to obsess about that, and it misses the step. I guess, Moe, this is to my answer to the emphatic, no, it’s not about the tool. They’re not actually going in with clarity on what they’re trying to do.

00:08:51.31 [Moe Kiss]: Sorry, the person using the tool you mean doesn’t have clarity on what they’re trying to do. Is that what you’re saying?

00:08:55.63 [Tim Wilson]: They’re saying, I’m supposed to be like, I kind of want to know how that campaign did. What does that really mean? Well, I mean, I guess like, how many registrations did we get? Where can I go get the number of registrations? There’s already that’s kind of broken because they’re kind of setting off for sort of aimless wandering in the data with the hope that some useful insight will emerge. And where they hit the most tangible blocker is they have some sort of frustration or limitation with the tool. And those could be a million different frustrations. And then they start to say, well, my problem is that the tool is not giving me this. And I think, in my experience, the problem is often, no, you’re just kind of trying to wander through the data and find something.

00:09:43.54 [Moe Kiss]: They’re trying to do exploratory analysis, right? No, not while I take issue with your… Fine, whatever you want to call it. They’re trying to wander through the data, but potentially don’t have the skill set to do that in a structured way that they don’t end up meandering. And so then the question is, instead of trying to teach people to use a BI tool, do we actually need to teach people how to do analysis, how to answer a business question?

00:10:08.98 [Tim Wilson]: Well, how to validate that it’s a good business question in the first place, right? I mean, because even that, not to a pedantic, like, well, if somebody in the business asks the question, isn’t that a business question? Definitionally, yes. Is it useful? Many, many times, no. They’re, yeah.

00:10:24.89 [Moe Kiss]: I underestimated your level of soapboxness on this topic. I do think there’s another layer of this that I’ve experienced in the past.

00:10:41.20 [Colin Zima]: I’ve managed data teams and been sort of like like disconnected enough that I’ve asked for things from the data team. And I actually think one of the other challenges, and this isn’t a product problem either, it’s a people problem, is the language of data people and how data people think is actually quite different from how business people think. And the translation can actually be very challenging. I’ll give you an example. We did this at Looker where I had left the data team at this point and was sort of doing product stuff and we were doing a giant repricing. And I went to the data team and I said, hey, would love to do some analysis. We’re going to do some repricing. And we cut the data a bunch of different ways. But they essentially came back with a dashboard that had 30 tiles on it. And my first question was, great, this all looks good. What should we do with pricing? You did all this analysis, what should we do? And their answer was, I don’t know. We cut the data a bunch of different ways. What do you think? And I was like, guys, The point of this was not to make the charts. The point of this was to get the conclusion. And the charts were for you, and I guess for me, to help get to the conclusion. But really, if you had just shown up and been like, these are the pricing tiers, that would have also been equally good. And I think this, like because the BI tool is the vehicle for communication between these two teams, teams being everyone and the data team, you get these sort of lost in translation conversations where, like, the data team might be concerned with sort of like pedantically correct, well-structured semantic layers and models. And the business is just like, hey, I’m just trying to go spend some money right now on marketing. Like where should I go put it? And I think that is actually one of the big problems as well.

00:12:20.43 [Tim Wilson]: But I think it’s a great point that everyone’s well-intentioned. Everyone has good intentions and is very capable. But if you have the business saying, I’m trying to speak the data team’s language, they wind up speaking in that language and saying, like, I think we need to kind of slice it a bunch of different ways to see if we can figure out what this is. And what the data team hears is, oh, the ask is to slice it a bunch of different ways. And then the business will be able to look at it and the answer will emerge and materialize.

00:12:54.15 [Moe Kiss]: Yeah.

00:12:54.35 [Tim Wilson]: And you wind up with both being saying, well, that’s what you asked for. I mean, the classic, I mean, the line I like to use is like you have all these dashboards and the business at the end of the just destroys the data team when they say, I know you’re doing everything I’ve asked and it’s a lot of stuff. And I basically understand what this is, but what am I supposed to actually do with it? And it’s like, it’s a dagger to the heart of the data team that then gets frustrated saying, what the hell? You just said I did everything.

00:13:24.10 [Moe Kiss]: Tim, that happens at mature companies. I think what happens is they like you do all the things and then they go, oh, can you just have this filter or I just need to have this view or like, oh, what if we just tweak it this way? And what they’re actually saying is this dashboard is not answering my question. Like I can’t and they just keep adding to it. And then you end up with this hot, hot mess.

00:13:46.45 [Michael Helbling]: Hey, you listen to the analytics power hour so you know that data is everywhere and it is growing fast. In fact, we just generated some data together since you’re listening right now. You know, you put all the world’s data together and we generated 149 zettabytes last year. And next year it will be 181 zettabytes. I mean, that is a lot of data. Thankfully, you don’t have to manage it all, but you do have to manage yours, and that’s why there’s 5TRAN. It’s your one-stop platform for all your data sources with over 700 managed data connectors, 5TRAN seamlessly centralizes your ever-growing data streams, and now it’s even easier with the new connector SDK. Data engineers can use simple Python scripts to build custom connectors into any data source. That’s worth saying again, any data source. So forget the data chaos. Centralize effortlessly with Fivetran. All of your sources, one solution. Simplify your data life. You can go start a 14-day free trial today at fivetran.com slash aph. That’s F-I-V-E-T-R-A-N dot com slash aph. All right, let’s get back to the show.

00:15:11.01 [Moe Kiss]: Colin, I need to just revisit Tim. Go ahead. The topic you were just kind of on about that specific example with the pricing. I’m really curious because full disclosure, we are doing a lot of thinking internally about BI tooling and semantic layers and all that sort of stuff. It seems to be the topic of the moment because we’ve had a BI tool for three to five years. Obviously, that’s the thing that we’re looking at at the moment. That’s what you’re blaming at the moment.

00:15:37.97 [Tim Wilson]: That’s what’s being blamed at the moment.

00:15:39.65 [Moe Kiss]: That was sarcasm, Tim. One of the observations I’ve had. There’s no secret in this, but we have been a big user of mode for a very, very long time. I pointed out to someone the other day, one of the challenges that I feel that we have is it is used as an exploratory tool by data scientists a lot. It is very good. You can make hacky stuff, you can quickly get to an answer, do what you need to do, write SQL, awesome. But then there are other people in the business that have really been trying to use it as a dashboarding tool. that is stakeholder facing and does all the things in my view, not super well, because it’s really ugly and I like beautiful things. Then ultimately, you end up with this tool that is like data scientists trying to build hacky exploratory shit, and then you have trying to have business stakeholders use it to go to look at dashboards for specific things. You end up with this massive amount of bloat. because no one can find anything. It just becomes unruly. Are we getting to a point where we need to think about this from a more mature point of like, lots of businesses now have multiple BI tools for different purposes? Is that just the evolution that if you get to a big enough stage, that’s kind of what you have to do? Or is it, again, back to that, we’re trying to use it for too many things?

00:17:06.70 [Colin Zima]: I mean, I’d like to think that someone could solve this. I think the reality is like when you start building different tools, I actually, not to answer your question directly, but I’ll give you like a direct analogy, which is Looker had semantic layers, you know, like, I feel the burn of it. You’re aware of them. DBT didn’t exist when Looker started. At some level, you’re always trying to turn this knob between self-service and control. What you’re talking about is this knob between self-service and control. It’s like, either people can do stuff and they make a mess and they’re wrong or they can’t and they get frustrated. I’d like to think that at Looker, we democratized a lot of sort of getting it data because we had the semantic layer and people could do things on top of it. But in some ways, a lot of people thought that Looker’s semantic layer was even too open and too many people were touching it. And for a lot of early customers, DBT was an interesting option just because they might have only had four people that had access to the DBT layer. And thus, there was a new gate and a different level of control and a different version of self-service where Looker was even abstracted away from the transformation in the warehouse. And I think the mode looker balance that you have is very similar to like the Excel Tableau balance or like pick any pair of tools. Like there are progressive layers of freedom that you get the more disconnected you are from the core business system. And in some ways, it’s nice to have a disconnected tool because you can just point at it and say over there, we don’t trust anything, but you can do whatever you want. But over here, we have control. The downside of that philosophy is like I’m coming up with terrible analogies as I talk, but it’s like if you had like a crime-free zone and a no crime zone, like I can’t remember the purge, like if you go crazy on the purge day, then like, yeah, you’ve made a mess in one day, but like the other days are clean. And all you’re kind of trying to do is like side pocket and figure out how much of a side pocket you want. So like maybe the purge is a good structure for your data org, like chaos and then order. Or maybe you should try to create order every single day and create more balance. That was a bad analogy, I was testing that one.

00:19:22.86 [Moe Kiss]: I was going to say, we often talk about the path of there’s a structured path to go down that has been built. If you want to go off road, it’s fine. But then you’re responsible for your four wheel drive and your own safety.

00:19:33.65 [Colin Zima]: Exactly.

00:19:35.01 [Moe Kiss]: One of the things, I’m really curious to hear your view on, is we also are users of Looker. We have built LookML. This is my personal observation. I’m sure there are people that potentially don’t agree. I feel like what we’ve done with our LookML, and this gets us very much into that hot topic at the moment of semantic layers because it seems to be all anyone’s discussing, I feel that our LookML essentially replicated what we had in our data warehouse of basically like report tables, right? Like super lean, but very structured. And what it meant is like we have 70,000 tables. In snowflake in our data warehouse and we duplicated that in looker where like you don’t have a view for. Users you have a view for users by country you have a view of users by platform you have a view of users by marketing channel by users by and you end up with this. What seems to a data person as a very beautiful, like look ML layer, but to a user, they’re like, sorry, there’s 50 tables. It’s like 50 looks at say user, which table do I use? I sometimes feel like it’s not even like it’s the data team wanting to structure it. to be perfect from their view of like this is the perfect architecture, but then it creates this usability problem. I’m so curious to hear your views on that.

00:21:03.54 [Colin Zima]: I think this is why we bounce between these centralized and decentralized platforms. So like a lot of my thesis is essentially like people appreciate business objects and like very centralized data teams. And so they were like, perfect semantic layer, data team controls everything, great. Quality very high, agility very low. And then someone gets their hands on Tableau desktop, starts building things and everyone’s like, oh wow, that side of the world looks really good. Ditch the business objects, everyone runs over there. That builds for five years, maybe three years. And you’ve got all this reporting and then you’re like, wait a second, Like we just did a meeting where eight people brought eight different metrics. Like this is like, oh wait, look at Looker over there. Like that looks pretty good. Like let’s go run over to that side of the house. And I think you end up with this oscillation back and forth. And like maybe you settle with a little bit on both sides. I’d like to think that the ideal version of this is you’re doing it in a bunch of different layers, and you’re sort of differentiating between what happens and what layers. So the example I always give, and this is a pretty trivial example, is like five trans sucks data out of Salesforce. For whatever reason, Salesforce delivers deleted records. No one ever needs that. Like clean that up in the warehouse, make sure that no one can ever touch a deleted record. Like I’ve built up on to reporting on deleted records. It’s very frustrating. In the inverse, if you’re building like a daily, weekly, monthly table and telling your user to go navigate the BI tool to go figure out like daily active users and they’ve got to go like exploring for 14 different cuts of users, I think then you’ve probably abstracted too much into the warehouse and you’ve not left enough in the BI tool. It’s really hard to actually strike a proper balance, which is why everyone now is like, hey, maybe AI fixes this and like, I can just throw a text box on top of everything and like magic beans, like everyone can get the reporting they need.

00:23:01.18 [Michael Helbling]: Because somehow the LLM will know which of those user tables is the right one.

00:23:05.12 [Moe Kiss]: Exactly.

00:23:06.26 [Tim Wilson]: It does feel like, I mean, that’s everyone says the requirements are very simple for my BI tool. I just want a tool that’s intuitive, easy to use, gives me access to all the data and I won’t get in trouble with it. And that is the behavior that people act under. I will call out that we had Ben Stansel on episode 190 to talk about metrics layers. Remember when those were kind of a hot, and I think he’s gone beyond it. I still am firmly in the camp of when people show up with those different reports, different numbers, different revenue numbers. Sometimes it’s not particularly material. Sometimes it’s not even a metric that fucking matters. But in, we have conditioned people to say, ah, we have found the discrepancy and everybody feels actually good about going to track down why these metrics are different. And another way to like solve for that is to look at a lot fewer metrics when you’re sitting in a meeting. I get that there is a tooling desire, and I love the way you’re articulating the range of, it’s a balance, and there’s never going to be a perfect balance. How do you handle it? To me, it feels like it’s It’s kind of going to be an impossible fix as long as businesses jump to the data. Everybody wants to come. We’re supposed to be data driven. So let me bring my 75 charts. If everybody brings their 75 charts, we’re going to find charts to disagree and we’re going to have an argument about which chart. And then somebody is going to say, we got to figure this out. So then it goes back to the data team who they’re they’re both right. But now the data team is putting together an explanation of reconciling these two different revenue counts. And that’s the next meeting. And nobody’s saying, what the hell are we doing? Like, this doesn’t matter. But everybody’s felt good. They’ve been they’ve had activity and they’ve discovered things. But OK, I’m sorry that. Sorry, sorry. It’s going to happen every seven minutes mode. They’re just going to be like a little pressure valve and then I’ll go quiet again. Will that work? No, I’m here for it. OK.

00:25:20.46 [Colin Zima]: The what I was going to say is like I can’t remember. I think I took I think this might have been like a pop science book that I read that effectively the thesis of the book was like happiness in life is about exceeding expectations. for everything. I might have gotten the wrong message out of it, but that is what I personally took away, is like, always have lower expectations than exceed them. And I actually think this is the problem with data tools, is if your expectation is like, I’m gonna do some product analytics and they’ll mostly be right and we’ll make a better decision, you’ll exceed expectations. If your expectation is like, we’ve got a little bit of a mess over here and we’ve got some slower metrics over here, but like you’ll mostly get what you want, If the expectation inversely is, I can answer every single question I have in eight words with no context whatsoever, then you have a very mismatch expectation for how messy data is in reality. And so I think the funny thing when you think about people that really are happy with their data tools, I think sometimes it’s the most technical people. that are businessy. And the reason is because they can find the balance between these things. I don’t know Ben that well, but I bet if we had a conversation about managing product with data, both of us would be like, a little data input is good, but mostly just make good decisions and pick stuff.

00:26:42.02 [Tim Wilson]: Yeah, but he’d have like a long footnote and links to some obscure movie in a clip. He would say it much more eloquently.

00:26:50.16 [Colin Zima]: But like, I do think a lot of it is just being able to communicate what you are doing and have people understand it well. If people, I remember this at the tail end of Looker, for example, like you start getting bigger and everyone’s like, let’s make metrics based product decisions. So the canonical example is like, if we move these pixels around, do people click more stuff on my e-commerce site? It’s great if you’re Amazon because like moving things by a couple of pixels like increases, you know, click through by 50%. It’s like for Looker, it’s not gonna tell us whether we should go build like trellis charting or go work on like the SQL interface. It’s just like, be an adult and make a decision and choose between the two of them. And then like monitor and make sure that it’s doing what you want it to do or that, you know, like people aren’t getting stuck in some weird way, but like data is not going to do your job for you. And I feel like a lot of people are expecting a data person to come in and be like, go make me some money for the business. Like I gave you the database, like where’s the money now? And it’s, it’s not usually that simple.

00:27:55.07 [Moe Kiss]: Okay, so just hypothetically, Colin, I feel like I could talk to you for like 5,000 years about all of the problems that I’m trying to deal with. So I love this point of expectations and you want everything to exceed if if things exceed your expectations, then that’s great. With BI tools, fundamentally, I do think the issue is that what is sold to the business is also, we’re going to migrate from tool X to tool Y, we’re going to do POCs, people spend way too long gathering requirements, doing all the things, and then it’s sold as like, this is the perfect solution, which disappoints everybody because it doesn’t do any of the things completely perfectly. It sounds like the win is to message it differently of like, we have this tool, it’s going to be able to do 90% of what we need. And then when for some people it does 95%, everyone’s going to be like, yeah, this is awesome. Like, is it actually we just need to get better at how we message this?

00:28:55.50 [Colin Zima]: I’d like to think the answer is sort of yes. I think if you, I’ll give you an example, which is like, if we go turn an AI on our company and we’re like, this thing’s gonna make us 100 times as much money next year. Your CEO’s gonna be pretty disappointed. If you make one and you’re like, I think this will search docs a little bit better. And then you come back and you’re like, wow, this thing can kind of answer all of our support questions. That’s pretty good. I think then you’ve won a little bit. you’re probably not going to be able to go to your boss or your boss’s boss’s boss or whatever and be like, I’m going to rip out my BI tool and monopolize the data team for six months. And we’re going to get an OK product. Like it needs to go solve some problem. But I think you need to the thing that we would do in Looker POCs, for example, is we would always try to solve a really tangible problem. Like if marketing was the problem, let’s go make marketing a little bit more efficient in the POC. And like, of course, you can get sidetracked and get into like the 100 thing checklist. But ultimately, you’re just actually solving a marketing problem. And frankly, you might not have even needed to use Looker like you might have just gotten a sales engineer and whatever tool that they wanted and been able to do it.

00:30:07.83 [Tim Wilson]: The incentive structure, I mean, I think this is the challenge. I think I agree. Changing the messaging would be the solution. The challenge is that the BI platforms and their sales teams, like they, in marketing for the BI tools, RIT large and analytics platforms, they kind of have to have a, a pithy declarative statement that is the kind of extreme. And then I would claim that even when you go to the POC, when you do the demo, the demos are always simplistic. They look at the example of how this works on this one page that’s the lowest hanging fruit. What happens, I will claim, is people see that and say, that’s awesome. That’s not my exact use case, but mine’s adjacent to it, and I’m sure it will do the same thing for me. The same thing happens when picking a POC. The most solvable thing happens, and I’m not opposed to POCs. There’s a lot of value coming out of them, but the next level of managed expectations is saying, we picked a tight and defined POC that had a high likelihood of success and would let us kick the tires a little bit. Now everyone’s going to assume that that’s going to happen in an instant for all of us and neither the BI platforms nor the consultants who are being brought on to help implement have any incentive whatsoever to disabuse them of that. notion because they’re getting paid when the licensing gets signed or when it gets implemented. They’re not getting paid a year from now when value has really been delivered. I think it is a massive, it’s just a structural challenge. I think the more people who are trying to manage those expectations, it does fall, I think, to the internal person that says, Take everything that’s being said with a grain of salt. It’s shocking how little skepticism, seasoned, experienced people have. When the vendor tells them X, they just accept it. When the vendor doesn’t know the internal complexities, absolutely.

00:32:19.75 [Moe Kiss]: Moere than you think. I feel like I’m a bit of an asshole. I’m like, this could go wrong.

00:32:28.28 [Tim Wilson]: But the thing is, if you look at their incentives for many, many people, people have been beating up on the tool. So it’s either our tool or it’s my team and the process we run. What’s the easier psychological thing? Yeah, I will pile on the tool now. And guess what? If I buy a new tool, and Colin, you said it, that buys us six months or 12 months, of time where we get to tell people, I know business objects sucked, but we’re rolling out Tableau. I think the opposite.

00:32:56.21 [Moe Kiss]: Well, you’re wrong.

00:32:57.39 [Tim Wilson]: That is a thousand percent what happens.

00:32:59.74 [Moe Kiss]: I know that that is what happens, but that whole… I fundamentally find it so weird how people are like, we’re going to do this migration at six to 12 months. I am like, I get mortified when I need to tell the business that of like, we’re going to spend 12 months migrating something. Like that’s unacceptable. Like I find it weird that people would want to use that to pad. Like, I’m like, that means we’re not moving fast enough. We’re not delivering value.

00:33:27.88 [Tim Wilson]: But once they get over that hurdle, they’re in this like glorious window.

00:33:32.63 [Michael Helbling]: The overarching point is, We take a process problem and we try to slap a tool on top of it as opposed to reevaluating the process and its lack of functionality within our organization. And that’s kind of, I think, Tim, what point you’re making, whether it’s three months to replace it, whether it’s 12 months, it doesn’t kind of doesn’t matter. All right, I’m going to mute Tim in motion, just a second comment.

00:33:57.81 [Colin Zima]: You and me.

00:33:59.12 [Michael Helbling]: No, I’m just kidding. But I do wanna just switch gears just a little bit because we’ve talked a lot about a lot of the challenges in our industry around BI, but there’s lots of things happening right now in our space, specifically AI. And one of the things we all wanna do is get analytics people to sort of think where we’re headed with this stuff. And so maybe you can start to share some of your perspective on, okay, there’s a lot of talk and hype around AI, but when the rubber hits the road, what’s real and what’s the hype cycle?

00:34:37.72 [Colin Zima]: I think it’s actually this conversation even magnified. I remember when I joined Hotel Tonight, there was almost this expectation that was like, hey, you’re a data scientist. We don’t have one of those. Go make us some money. Find us the insights. Yeah, exactly. It’s just like, isn’t that what data science is? It’s like, you take data and you make, and I feel like AI is actually that more magnified. It’s like, people are sort of like, you know, why can’t I just hook this up to my database and have it optimize my business? I’m pretty far along the side of like, I think that that’s what humans are for, is like interpretation. And if you have a true optimization problem, it’s because you’ve cleaned the data so well that you’ve created essentially an optimization problem, not actually a business problem. I think at the same time, it is remarkable the things that can happen with so little human input that AI can do. The examples I always cite are Like you can go hand at your database and it’ll go write you like a query with eight CTs and go spit out some sort of cohort analysis like something that would have taken a user hours to do. And it can be right. I think the flip side is that the control of what AI is doing I think is actually really important. And the way the human fits in the loop I think is really, really important. The examples I like to give for stuff like this are I think that if you look at the AI use cases that are most well adopted now, it’s writing and it’s coding. And the reason is because they are so heavily human in the loop. Like if it writes a block of text, you don’t just press enter and like get a picture of a block of text, you can touch it, you can feel it, like you can manipulate words. Similarly with code, it’s not like go build this app for me, though there’s obviously stuff that is doing, trying to do things like that. It’s much more like I took a cut at this, you can pull in the things that are valuable and throw out the things that aren’t. When I see it applied to data, I think that there’s a pretty hard fork of AI that will write SQL. So truly like black box, the stuff that Anthropic and OpenAI are spending a lot of time working on, like I’ve heard they have a couple hundred engineers that are working on text to SQL. or sending it through a semantic layer. And again, this comes back to governed analytics, the classic BI concepts. I think that people will have to make a decision. Implicit is obviously, I think semantic layers are really, really important because I think if you cannot maintain control and tie it back to UI, it’s going to be impossible for users. I think there’s still going to be a place for text to SQL. If I do need to go write 200 lines of SQL, it can be really valuable. But I think that the control level in manipulating SQL is just not high enough to address all of the users that we’re talking about. And so I think for it to apply well in data, it’s going to need to attach to UI. So what I mean by that is you’re going to need to send it through some sort of semantics or some sort of intermediate layer that can let a user touch and feel the results, touch the filters, understand the subqueries, understand what the aggregations mean. And what I would say is, I think when I do see it applied in those types of contexts, it’s actually unbelievable how good it is. We turn this on our own Salesforce data, and I routinely now, instead of building a dashboard that is like an opportunity lookup, I say, Give me some information about opportunity XYZ. It just picks out eight random fields or whatever, but it finds me who the sales rep is, who the sales engineer is, whether there’s three opportunities, and it does it in four words. That kind of stuff, while it feels like a very low bar, no one would call an opportunity record lookup data analytics. I actually think those are the areas where end users are struggling the most very frequently. Show me the Zendesk tickets for customer XYZ. That’s not data analytics. It’s data retrieval, basically. It is. But that is actually, I would argue, the most impactful version of data analysis. What I know is the least sexy thing that anyone would ever say. But looking up stuff is kind of hard. And it’s all sitting in the warehouse and it’s all attached together and you need to be able to link it all up. And I think AI is actually unbelievable for those types of problems because it can get things mostly right. And then the human can say like, oh, actually I want these two fields or I want to tweak the filter or I want to change the sort. And so I’m pretty excited about those things. I think the sort of next layer is, can it ask the next question or sort of point you at follow-ups? And I think it sort of can. There’s just always the danger that it does the trivial. And like, I see this a lot. It’s like, oh, I can cut this by time. I can cut this by region.

00:39:27.38 [Tim Wilson]: It can cut it so many ways that it can, it’s going to find anomalies, just like mathematically, it’s going to, I can slice, if you had to slice it, every time and look for something that would slow you down. And if you found something you would have had, it would have had your logic behind it. If you just let the machine slice it a thousand ways, it’s going to pop up like 20 things. And that’s just statistics. Like that’s, yeah.

00:39:53.03 [Colin Zima]: We have 300 customers now. Like I don’t need it doing statistical analysis on our opportunities. Like they’re coming from just like the growth of our business and like the signal means nothing.

00:40:03.16 [Moe Kiss]: You have such a wonderful perspective of this. I’m curious to understand how you’re balancing it, though, because I’ve been looking at lots of BI tools and it feels like everyone is trying from a product perspective to sell the dream of just natural language questions with an interface. Anyone can ask anything and like, how are you balancing that? Because that’s what is being sold to execs that we’re going to be able to do this in months.

00:40:31.12 [Colin Zima]: I mean, we do a little bit of it too, don’t worry.

00:40:35.39 [Colin Zima]: I think that I’m trying to encourage people to climb the slope more gently, which is not always the most appealing statement. But again, going back to the Salesforce or Lookup example, I truly think that Lookups are the most killer use case for natural language right now, immediately.

00:40:53.11 [Tim Wilson]: But is that genuinely what people are struggling to do without a simple, where is this data? It feels like the most solvable with traditional BI.

00:41:01.19 [Michael Helbling]: I think more than you think actually. I see it a lot because a business user doesn’t know how to address the underlying data in a way that’ll get that for them. So like being able to just ask it and the AI kind of know like, okay, this table, this table, this table. It’s sort of like a little bit of a analytics engineer in a box. So instead of prioritizing in a queue where I now need to wait a month until they can get to my report, if they ever get to it, I can just have it right now.

00:41:27.94 [Colin Zima]: Yeah. And the other example that I think that people, and again, this is going to sound trivial, but the idea of reverse value lookup, I can’t tell you how many times I’ve walked into a tool where someone’s like, I need to filter for customers in the US. That’s not a complex query, but to find US, you need to know whether that’s region or country or geo. And I know that sounds incredibly trivial and every tool should be able to do that.

00:41:53.78 [Moe Kiss]: No, I hear you. I deeply hear you. And did they write US or United States?

00:41:59.37 [Colin Zima]: Or USA, or yeah. And I think those are, again, in terms of expectations, I know that’s solvable. And I think I see our sales reps literally using the product more because they can do stuff like this. I think the follow on is like, if you get really good at this thing, like you unlock the next thing and the next thing and the next thing, I think it’s less of like a a regime change stepwise, like everyone’s great at data now. I think it’s that people get a little bit more comfortable, like in the same way that instead of doing a lookup on Yahoo message boards, now you went to Google and you typed in search and like, you know, ThoughtSpots had this message for 15 years. But I think in many ways, like that has become trivial for every BI tool to implement and will now become much more native in all products that we are using.

00:42:50.78 [Tim Wilson]: This is kind of going back a little bit, but I would have loved to have you just riff a little bit. It’s Microsoft, so we can say good or bad things are huge. They’ll be fine. We’re going to have no influence. But as you were like the Excel to power pivot, to power query, to power BI, DAX or no DAX in I know this is still bugging me from when you were saying it. Is it fair to say that was Microsoft’s attempt to address that balancing act that you’re basically there’s like a Peter principle for the business user that they’re going to progress Now, if they got to go to actually Excel to a pivot table to then power pivot to power query, that’s a complicated and he hit the limitations in Power BI of which desktop versus window who has what access, what data can I pull into. So you wind up with all the other challenges we’ve been talking about, but was that kind of a viable and intentional approach by them?

00:43:49.20 [Colin Zima]: I think they are the only company I cite that is trying to do what at least I say that we’re trying to do. I think they have tried to layer everything into one place. You can layer in PowerPoint and presentations on top of that as well. I think that because they have such a wide customer base and they’re very good at producing product, they’ve realized that the semantic layer is different than the spreadsheet. is different than the halfway point in between them. And the products evolved probably a little bit more naturally. So the way that they link together is maybe not as elegant as it would be if it was like centrally planned. But I think the reason that all of those exist is because of these gradients. Like DAX is really just a natural evolution of Excel. attached to sequel, it’s their intermediate language halfway between sequel and excel. But like it does a thing that neither of them do. And so like, yes, I think it’s actually the best done version out there by a single player.

00:44:49.16 [Tim Wilson]: If they had just stopped at Excel 2003 and then started building instead of getting all the bloat and Excel trying to make Excel do, I mean, actually that’s interesting. Excel tried to do too much. Now all that legacy overhead is still there before they built up the other. Okay. Oh, okay. That’s helpful.

00:45:07.42 [Moe Kiss]: Yeah. I mean, that was actually going to be my next question was just on that topic is like, there’s seen, there does seem to be more like strong newer players kind of in this space at the moment.

00:45:19.55 [Tim Wilson]: And part of me is wondering like, they’re also garbage charlatans who are AI hype monkeys.

00:45:25.42 [Moe Kiss]: Totally. Totally. But that’s always the case, right? For some, um, it, It does seem like there is like this not race, but like maybe a bit of a pushback from us. And I mean, I work in tech. So, you know, we have a very different perspective on tooling and what we’re willing to try or like appetite to try new things, maybe then some more traditional businesses. But there does seem to be this thing of like some of the more legacy BI tooling, like a bit of pushback. And I just wonder, like, I feel like so many companies have tried some of those tools and it hasn’t worked. We need to go to something that’s a bit more new and innovative and thinking about this problem with a different perspective. I’m curious to see what you’re seeing. Is that a fair observation or is it like, no, this is just the standard run of the mill and people are still choosing the legacy tools?

00:46:18.26 [Colin Zima]: Yeah, I think there’s two things at play. So one is, I can’t tell you how many people I talked to that are still using micro strategy, now strategy and business objects.

00:46:26.91 [Tim Wilson]: They’re just in it for the crypto.

00:46:28.19 [Colin Zima]: Yeah. Well, maybe they like Michael Saylor and his business strategies. But like, I would argue that a lot of those are there purely because they literally just do what they need to do. Like, to your point, they could go transition off of them. I think that over the infinity of time, they will all be deprecated, but maybe they do need to produce 12 spreadsheets a day and send them to S3 and let the team pick them up. They are happy with that workflow and it exists. I think the challenge is a lot of modern tools didn’t pick up the functionalities that business objects and micro strategy have. So like a lot of the hardcore legacy modeled BI tools, I think actually still don’t have like 2025 comps that can replace them, ourselves included, like we have work to do to do some of those things. So I think that’s one piece of it. I would say the mediating factor on the other side is I’m also talking to a lot of like CIOs of big companies now that are looking to buy any AI tool under the sun to POC it because they’ve got to go buy some AI. Because they have a mandate. It’s a weird dichotomy of, like, I’ve got MicroStrategy producing PDFs over here. We talked to one customer that was like, I need to spit out 200 pages of PDFs of all the products that we sold yesterday, every day. Like, I need your tool to do that. And we were like, well, we don’t do that yet, but we will do it. But it’s like they have business objects doing that, and they decided they need it. But at the same time, they’re like, also, we want to replace our whole front end with natural language. And so you have this weird tension between the dream and the reality that people are trying to navigate. And I think that’s what’s making it a kind of strange time, because you do have the YC two-person companies doing text to sequel that are able to go talk to Fortune 500 companies. And then you’ve also got a 22-year-old business objects deployment.

00:48:28.20 [Tim Wilson]: Moe made the comment a while back. I’m going to do another callback to mode and maybe the not the most attractive visuals. So broadening it, I’m not into beautiful. I’m into effective. But it’s not about making the data pretty.

00:48:42.68 [Moe Kiss]: Beautiful can be effective. Pretty means being understood.

00:48:47.08 [Tim Wilson]: I have modules and classes I’ve taught on that front. That’s another reason I think people have stayed with Excel, is Excel still seems to have more flexibility on the charts and every BI tool I’ve worked with, I can’t get I can’t shift the get another couple of pixels between the grid and the label on it. I’m deep down the RGG plot world, love that, which every tool was working in that, but that would be stupid because nobody would understand it because it’s a nightmare to learn. So I put myself in the camp of the like super precious about the specifics of the visualization, the palette, the color, the font size, all the stuff to follow Steven Fuse and Edward Tufty’s best practices. BI tools, I understand it because they’re trying to serve all these masters and they have to plug in the limitations. Like how much is the front end visualization a, impossible, it has to be good enough. We’re going to differentiate ourselves for a long time, Tableau, that was the way they were differentiating themselves.

00:50:01.30 [Colin Zima]: Still. Literally still.

00:50:02.96 [Tim Wilson]: So where do you fall on that on the importance of saying, when I do a chart, I need to be able to pick it and control it. And if I want a bar with one series and a line with the other series and this on the right side, where does that fall?

00:50:17.99 [Colin Zima]: I’d say I’m closer to the most side of the house on this one. When it comes down to it, again, you have many masters here. The example I always give is if you look at the chart of sort of dashboard consumption in an organization, it is just absurdly skewed. Like the top three dashboards are 80% of the usage in the entire environment. If that’s the case, you should spend time on those three dashboards, making sure that they look and feel amazing. Inversely, for the other 2000 dashboards that exist, have an average of two views apiece, you need to optimize for how fast it is to build them, how flexible they are, like how well they attach to different shapes of data. And this again gets back to this really difficult challenge of like, It needs to be both the fastest to build in the whole world, but it also needs to have the most extensibility. So a thing that we did is you can build it in the UI, but you can also unlock the Vega spec and literally write Vega code. And if I said that to a business user, they just glaze over. They’re like, I don’t know what you’re talking about. But the point is it’s not for that person. It’s for the one dashboard where you do need to move the title 15 pixels to the left. being able to go do that. And again, like it’s so hard to do both. And the reason Excel can do so much is like, frankly, they’ve just built these features over 40 years. And it’s really hard. So like, I like to say that we want to be better than every tool at everything, except Tableau. We want to be almost as good at them as visualization, which is like, it’s kind of sad, but it’s just like, it’s hard.

00:51:50.63 [Tim Wilson]: And just to clarify, and I think that can probably go back over a decade where I think most sister and I had a shared presentation. Just to clarify, I think the visualizations that I produce are beautiful, but the big point making when teaching analysts who say, what do you need to make it pretty? There is a tendency to say, I’ll do this crap like dropping shadow or adding more color or doing all the stuff that’s additive and terrible. So I say it’s not about being pretty. It’s about being understood. Now, the fact is if you nail, they being understood, it is a more, it is a, it has a, it’s a lower cognitive load. People think that it looks good, but they don’t start off by trying to make it. So I just want to clarify when you said you came down on the side of Moee. I, there are many analysts who listened to this, who I have worked with who were like, Tim is a fucking stickler about making the stuff look effective.

00:52:44.77 [Colin Zima]: So I mean, I tried to I tried to like full band pie charts and word clouds at Looker for a while, like fully resist them. And like I still really don’t think that you should use them. But like we built both immediately at Omni because like what I realized is just like You need to pick your battles with your users. And if the person wants to build a word cloud, like I do, I can’t convince them every time that it’s not a good visualization. I just need to be like, here’s your word cloud, like good luck.

00:53:17.99 [Tim Wilson]: You should have the requirement that says if you put more than like four categories in your pie chart, it pops up an alert like buried in the ggplot documentation.

00:53:27.32 [Moe Kiss]: Don’t do this.

00:53:27.90 [Tim Wilson]: To do a pie chart in ggplot, you have to use this like change the coordinate system to pull in the help documentation that it basically says, this is generally a bad idea. Like we get that you’re trying to make, you were probably trying to do something that’s a bad idea. So that would be a killer feature for a BI platform.

00:53:46.64 [Michael Helbling]: You just pop up a nuke, form a clippy. Looks like you’re trying to commit a truck crime. You want some help with that? Okay, so I think we’ve nailed the solution here, which is… We haven’t even talked about APIs.

00:54:00.46 [Moe Kiss]: I mean, there is so much more. Anyway.

00:54:02.76 [Michael Helbling]: Well, we’re over time, though. Persona-based quizzes, when people onboard to the new tool to determine what features they get, It’s like, oh, you can’t answer these questions. OK, you get the static charts. Oh, you know a little bit about this? You get some notebooks. I don’t hate it, actually. Like, the philosophy of that is correct, I think. Yeah, because then you can say, like, hey, answer these questions, and then we’ll get you into the right part of the tool for you.

00:54:32.25 [Tim Wilson]: What’s the persistent avenue when they butt up against the limit and they have a need and they’ve gotten to that point, there’s an avenue for them to say, can you move me to the next level? I think that, I like it. Yeah. It’s a killer feature right there.

00:54:43.99 [Michael Helbling]: And you get a gold star for the day. Okay, but we do, we have to start to wrap up. Obviously, Colin, thank you for managing to get a word and edgewise around all of us. Cause like Moe has strong opinions, loosely held. Tim has strong opinions, strongly held. So great conversation and really illuminating, really appreciate your perspective. One thing we love to do, go around the horn and share a last call, something you might find of interest. So you’re our guest, Colin. Do you have a last call? You’d like to share?

00:55:19.03 [Colin Zima]: I have two quickies. One, I’m sort of embarrassed to say this. The games on LinkedIn are pretty fun. They’re like 30 seconds and you should play them every day. What?

00:55:29.18 [Tim Wilson]: I went for three or four days. I went down that I said, I need to step away. I’ve already got.

00:55:34.15 [Colin Zima]: You can see your leaderboard against like other people. It’s, it’s a lot of fun.

00:55:37.44 [Moe Kiss]: I go on LinkedIn like once every six months. Like I don’t want to, I don’t want to open the trap.

00:55:42.25 [Colin Zima]: This was their viral hook. Anyway, their games were fun. The second one, and this is like my little trick, turn off images by default in your email. And the reason you do it is because it blocks all the pixel tracking. And you can decide whether you want to turn images on, which is effectively alerting people that you’ve opened their email. But otherwise, it by default blocks all the pixel tracking. And so no one can actually track whether you’ve opened their emails or not.

00:56:07.16 [Tim Wilson]: So when you have over those 300 clients who has their email dashboard, and some jackass is looking at click to open rate, and the poor analyst has been saying, I’ve been trying, I mean, there are a million reasons that, I mean, that pixel is the most imperfect thing anyway, and it is treated as sacred. So that is, you just tried to make I mean, it’s a horrible metric to be looking at anyway. So I like the, you know, mess around with mess with them a bit more. So I endorse that.

00:56:42.21 [Michael Helbling]: All right. Moe, what about you? What’s your last call?

00:56:45.26 [Moe Kiss]: Okay, guys, you know how I go through this phase and I like get really into something. And then I like read everything on that topic and like, I’ve reached a new one. This is similar to the Why We Sleep book, where I’m going to be talking about this for the next 18 months. So prepare yourself, friends. I finished Careless People, a story of where I used to work, and the author, Sarah Wynne Williams, about her time. It was called Facebook when she joined. Holy shit, man. I actually wasn’t going to read it because I was like, I don’t want to read a worky thing. I need a bit of space. And someone’s like, oh, I actually think you should read it. It might be good for you. And I’ve discussed the book genuinely with so many people. But it’s kind of just reaffirmed for me how much my own values are important to me and my job and like, I don’t know. It’s just, it’s got me thinking a lot about like the kind of places I want to work and the kind of people I want to work with and also like, I mean, she just dropped some great tea. Like it is a good time. Careless people. Ah, okay. Yeah. So she worked at Facebook very, very early on with Zuck and Sheryl Sandberg and like, There are some great anecdotes in there. There’s apparently a word for it. The morning you get when you finish a really good book, I had that for days. I was so upset. But I was straight on to the next one. And so the summary is I am deep in my reading books that there’s lots of tea about tech companies. So I’ll show the next one on the next episode.

00:58:23.63 [Michael Helbling]: Nice. All right, Tim, what about you?

00:58:26.22 [Tim Wilson]: Well, I’m going to go with the book as well. And this is not a book. I have not finished reading it, but I’ve enjoyed it so far. It’s pretty random. It’s called Once Upon a Prime, The Wondrous Connections Between Mathematics and Literature. So it’s a mathematician. And you’re like a whole book. She’s literally, it goes down Tristan Shandy, a gentleman in Moescow, metamorphosis. Some of them are like experimental literature. She talks about easy things like haikus and there are other versions of haikus, but it is this deep exploration by somebody who loves reading and is a professional mathematician. It’s not particularly useful for anything other than she feels like the math and the arts have gotten too far apart, and she’s trying to bring them together. But it’s just got the whole concept of experimental literature that is math-based, and she’s got multiple examples of that. So it’s just kind of an odd, but interesting read. What about you, Michael? What’s your last call?

00:59:36.38 [Michael Helbling]: Well, I’m glad you asked. So MIT recently did a study of the cognitive debt when using AI assistance for essay writing.

00:59:48.80 [Tim Wilson]: Oh, dear. Have you read Cassie’s teardown of this thing?

00:59:52.11 [Michael Helbling]: No, I haven’t. So I’ll read that next. But I did read the study, and it’s not everything that the media folks are making it out to be like, if you use AI, you’re going to get dumber. But there are some interesting conclusions that they come to. And it’s, I think, worth a look at what they’re saying, because I think It goes even to something Colin, you said earlier, which is sort of like, if you just accept the response that the data gives you that it’s not ready the same with an AI, if you just accept the response without critically analyzing what’s going on or that person in the middle, you run the risk of maybe giving up just a little bit of your intellectual critical abilities and Lord knows we need those. So just be careful out there.

01:00:40.35 [Tim Wilson]: I’m just saying, Michael, I’ve already, it is 90% drafted. I referenced that study in a post that will be, will have long been out by the time we,

01:00:49.62 [Michael Helbling]: We’ll add that to the show notes then. Perfect. All right. Well, I’m sure as you’ve been listening, you’ve been having your own thoughts about this topic and we would love to hear that. Feel free to reach out to us. There’s some great ways to do that. Obviously, you can get to us on LinkedIn or on the Measure Slack chat group or via email at contact at analyticshour.io. So please reach out. Colin, once again, thank you so much for coming on the show. Appreciate you taking the time to do that and share some of your insight. And as you go through this and you’re listening and if you like what you’re hearing on the show, please feel free to drop a rating, a review on whatever platform you listen to podcasts on. That helps us out quite a bit. See, I’m still not ready for this. Like I still want to thank Josh, but you know,

01:01:41.10 [Tim Wilson]: He did a lot. You can thank him again.

01:01:43.87 [Michael Helbling]: Hey, thanks, Josh, for everything you have done in the past. So anyways, but I know that no matter what BI tool you use, I think I can speak for both of my co-hosts, Moe and Tim, when I say keep analyzing.

01:02:01.59 [Announcer]: Let’s keep the conversation going with your comments, suggestions, and questions on Twitter at @analyticshour on the web at analyticshour.io, our LinkedIn group, and the Measure Chat Slack group. Music for the podcast by Josh Grohurst. So smart guys wanted to fit in, so they made up a term called analytics.

01:02:22.68 [Charles Barkley]: Analytics don’t work. Do the analytics say go for it, no matter who’s going for it? So if you and I were on the field, the analytics say go for it. It’s the stupidest, laziest, lamest thing I’ve ever heard for reasoning in competition.

01:02:38.55 [Michael Helbling]: I think they might be going direct to video at this point or something, but yeah, I thought that was a very bold analogy and I was, I was there for it, you know. One night a year you get to kill any dashboard you want.

01:02:54.81 [Colin Zima]: Hey, you’re a data scientist. We don’t have one of those. Go make us some money. And I feel like… Find us the insights. Yeah, exactly.

01:03:03.65 [Moe Kiss]: That’s what is being sold to execs that we’re going to be able to do this in months.

01:03:08.65 [Colin Zima]: I mean, we do a little bit of it too, don’t worry.

01:03:12.62 [Tim Wilson]: There are also garbage charlatans who are AI hype monkeys.

01:03:16.37 [Moe Kiss]: Totally, totally. But that’s always the case. For some.

01:03:22.84 [Michael Helbling]: All right, I’m gonna mute Tim in motion, just the second call, just you and me. No, I’m just kidding.

01:03:29.92 [Moe Kiss]: I feel like I’m a bit of an asshole. I’m like, this could go wrong. You’re a snowflake.

01:03:36.48 [Michael Helbling]: Obviously, Colin, thank you for managing to get a word in edgewise around all of us, because Moe has strong opinions loosely held, Tim has strong opinions strongly held.

01:03:53.62 [Tim Wilson]: So when you have over those 300 clients who has their email dashboard, and some jackass is looking at click to open rate, and the poor analyst has been saying, I’ve been trying, I mean, there are a million reasons that, I mean, that pixel is the most imperfect thing anyway, and it is treated as sacred. So that is, you just tried to make I mean, it’s a horrible metric to be looking at anyway. So I like the, you know, mess around with mess with them a bit more. So I endorse that.

Podcast: Download | Embed

Subscribe: RSS