One of our KPIs for the show is to keep the Topic Repeat Rate (TRR) below 1.2%. From carefully monitoring our show dashboard, we had an actionable insight: we could finally revisit episode #002. Conveniently, the topic of that show was dashboards, which explains the self-referential stemwinder of a description of this episode. That show was “a long, long time ago. We can still remember… when the dashboards used to make us smile.”
0:00:05.9 Announcer: Welcome to the Analytics Power Hour, analytics topics covered conversationally and sometimes with explicit language. Here are your hosts, Moe, Michael and Tim.
0:00:22.0 Michael Helbling: Hello, this is the Analytics Power Hour and this is Episode 178. A long, long time ago, I can still remember when the dashboards used to make me smile. And I knew if I had a chance that I could make those widgets dance and maybe there’d be insights for a while. But BigQuery made me shiver with every drill down I delivered, bad news as the chart stepped downwards, I was verklempt. I can’t remember if I’d spied a data viz that surely lied. But something touched me deep inside the day the dashboard died. And they were singing bye-bye Power BI, I took my Excel to Domo and I don’t know why. Them good old boys were using Tableau and Five…tran, singing charting graphs till the day that I die, charting graphs till the day that I die. And if you… Well, okay, no, we’re not gonna do that.
0:02:02.7 Tim Wilson: That was pretty bad.
0:02:05.7 MH: Alright, hey, welcome to the show. And Tim, how you doing? You’re pretty good on guitar.
0:02:13.6 TW: I’m afraid people were thinking they’re gonna do the whole song.
0:02:17.0 MH: That’s right, I’m like only nine more minutes folks, it’s just 30…
0:02:19.4 TW: Plus 30, plus 30, plus 30.
0:02:22.2 MH: That’s right, just keep hitting it. Moe, welcome back.
0:02:26.8 Moe Kiss: Hi, that was the highlight of my week, gotta say.
0:02:30.0 MH: There you go. [chuckle] Yeah, so we like to sing, and I’m available for any singing engagements, bar mitzvahs, weddings, whatever you got. Now you’ve seen a little sampling of what I can do out there vocally, just hit us up for the… No, but we went back in time and I think, did this come from a listener, Moe, or is this something you were thinking about?
0:02:52.5 MK: No, it’s been like burning in my brain for months on end, this topic.
0:02:57.8 MH: Yeah. We talked about dashboards in episode two of the podcast and then we’ve definitely covered things, have touched on but we wouldn’t have come back, and honestly, the world has changed dramatically as it pertains to dashboard. So Moe, kick us off here, we’re gonna spend the episode just talking about what’s the new rules, the modern dashboard dilemma, it used to be so much clearer because dashboards really came from not so many places and now they seem to be spread thin again, I don’t know the right way to say it but Moe, talk about why this came up as topic for you.
0:03:34.2 MK: Okay, so I suppose, I don’t know, I guess I take for granted that data viz is something that I care about and I think is really important and I feel like there’s lots of people that have completely amazing training on this topic, Tim being one of them, my sister Michelle, Lea Pica, there’s all these people that are just so into the topic and it just, I don’t know, my team…
0:03:58.7 TW: Of data viz, not of dashboards, just to be… [chuckle]
0:04:01.4 MK: Yes, but I… Yes, but I think the two are so connected and I think that if you don’t get data viz, you don’t understand how and why the way that you build a dashboard matters and I don’t know, it just seems to be something that I think my team kinda don’t really give a shit about. And that’s not to throw them under the bus, ’cause there’s lots of people in the team that do care about how we present information in dashboards but there’s lots of people that don’t, and it just keeps churning around with me being like, why is this happening, what’s going on, what’s with all the tooling changes, what is creating this like… I thought we nailed this in episode two, I thought people got it, you know?
0:04:49.2 MH: I don’t think anyone listened to episode two, so that’s probably the problem, Moe, is that we just didn’t have the listener base.
0:04:57.2 TW: I questioned the premise of the… I feel dashboards get treated as being… There is an attempt to make dashboards do a million things, that dashboards are the worst possible fucking tool for the job and frankly, since you’ll throw your team under the bus, I will throw aspects of my team where we sell that to a client and it’s like “Hey, we’re gonna build some dashboards for you.” And guess what, they build a bunch of dashboards and yeah, data viz is kind of crappy, layout kind of crappy but that’s not the issue. The issue is that they see the dashboard as literally being the be-all-end-all, we’ve got all this data coming from these different… And I feel like the analytics space, the tools are doing the same thing. Look, we brought in all this data together and then the next step is ergo, if you put them all in a dashboard, your stakeholders will be able to self-service and drill down and slice and filter and gain glorious insights. And Tableau, great tool, but it’s not gonna do that. Power BI, great tool, it’s not gonna do that. So I feel there is a conflation of take Analysis Workspace in Adobe, which…
0:06:19.5 TW: Is not a dashboarding tool, it’s an analytical interface tool and in the hands of an analyst or engaged stakeholder, yes, you can use this to explore your data, but I’ve literally never seen and plenty of people will take issue and say, “Well I have” and they’re kind of full of shit. Dashboards get built that are kind of, this provides all the access to all the data, and this is the source of your insights. I see dashboards as limited, some degree of filtering, some degree of drill down. Great, but they’re very best at illustrating performance, measuring performance and showing you, are you hitting… They are rarely the effective source of insights, but that may be a rant from number two.
0:07:13.0 MK: Yeah, I mean, there are so many directions I could go off your soapbox spiel.
0:07:19.9 MK: So I’ll pick just one of them, which is that idea… And I think we went through this right? Where we were like, let’s build all the dashboards with all the KPIs that matter to people and people will start going to look at them because why wouldn’t they. They’ve got the KPIs on there, and they don’t. You don’t have senior stakeholders being like, “Oh, I’m gonna go check Looker every day.” Some of them do. And the ones that we do, I love. But they still don’t… They don’t. The idea that you put it somewhere and people just self-serve… What are we missing? Because, you talk about all the tricks of making it someone’s home page or it auto comes up when they log on to the computer. I feel like we think of everything, but do people just not… Maybe stakeholders shouldn’t check dashboards, maybe that’s the moral of the story.
0:08:12.6 MH: I think as a listener you can probably tell we get pretty wound up about this topic.
0:08:18.0 MH: But it’s funny because I remember I probably had much stronger opinions about dashboards when we first did this and I think now I actually want the simplest thing. Honestly, it still needs to have the right data visualizations and sort of the ink to white space ratio and those kinds of things. That’s probably more important to me now but the amount of coverage, I really think people try to cover way too much and that’s why people don’t look at them like a very simple, very sparse piece of information. So there’s maybe five or six things you need to know every day or once a week or once a month. And I think the other thing is that we used to do this in one-offs, right? And so we’d make one of these and then we’d make another one of these, and now the tools like we’ve been mentioning… Or I’m sure we’ve mentioned a lot of them, Looker, Power BI, Tableau. They let you make a bunch of them all together, and so we create or conflate operational reporting, for lack of a better word into dashboarding and then we make 40 dashboards which are actually just operational reports but then it’s not the same thing. Not even close to the same thing.
0:09:34.1 MH: And I think that’s the thing. I honestly feel like we gotta go back and say, “Hey, this is what a dashboard actually is, and this is what you should actually be using it for, and then there’s other places you should do things.” like Tim, you mentioned Analysis Workspace in Adobe, which happens to be a thing I love, and actually Google Analytics 4, there’s the Explorer function, which is basically the exact same thing again. And it’s great, but that’s not a dashboarding interface, you do not create a dashboard out of that.
0:10:00.9 TW: Well, but that’s where people started… There were people who’ve said Data Studio and Analysis Workspace, and I’m like, “No, no, no, two different things.” Data Studio, actually much richer visualization capabilities, layout than Analysis Workspace but not something that you wanna try to build the infinite exploration…
0:10:17.9 MH: Yeah, you can’t build a lot of exploration capabilities into it. Yeah.
0:10:24.0 TW: So getting to that sparse, and this is… It’s gonna be well over 10 years ago, I did a three-part series, I remember it ’cause I was trying to get what is the platonic ideal of a dashboard. And I actually got down to where the ideal one would actually show nothing, it would only show the KPIs that were outcome-oriented KPIs, not bullshit operational metrics, not anything, not all the clutter and junk that’s easy to pull in. And it matters, but it’s not what… But it would even only show your KPIs if they were drastically exceeding or underperforming against the target that you set.
0:11:12.1 MK: Ooh.
0:11:14.8 TW: And the problem is, if you say, Well, how many of the dashboards actually, the metrics that they’re looking at have a… To me, a dashboard is an at a glance… Should be an at a glance view of how am I performing against expectations, against my forecast, against my targets. If you’re in Finance and saying you’re tracking it, that happens much better. In marketing, we skip that step, we don’t say, how many orders do I expect to get from this campaign, how many… You name it, we don’t say how many leads am I expecting to get? What is my lead to qualified lead conversion rate expected to be? If we don’t say that, but we say, just build the dashboard but then somebody looks at it and they kind of shrug. What am I supposed to take away from this? It doesn’t give me good or bad, we hide behind this, Oh, we need the context, and we need the dashboard to show the why, and we need the dashboard to show what it means. Well shit, you didn’t actually sit down and say what you’re trying to accomplish but… It depends on the timing of the business too, trying to look at a dashboard in real time, if the actual response is on more of a weekly or monthly cycle, you may still want a real-time dashboard so whenever they look at it, the information’s current, but that’s like the starting point of let’s build a dashboard and let’s pull in what data… It drives me insane.
0:12:40.9 MK: Do you know I’ve had people build a dashboard that have had a JIRA ticket created for them, and they built the dashboard and delivered it to the stakeholder without ever having a single conversation with them, that to me is a broken process. That to me is.
0:13:00.3 TW: So now, I’ll get off my, maybe, puritanical… If… That, I think, maybe in… And Michael, you kinda touched on this as well, the ease of building them, if somebody says, “Can I get my top pages and how they convert? And I’d also like to see this,” and you look at it, and you’re like, “Okay, you’re just trying to get an understanding of what channels and what pages, and you know what? I can kind of consolidate those into something that fits on one page, and that’s the easiest mechanism… ” I know a guy… Actually, when I started at Search Discovery, when you were still there, Mike, there was a monthly report that got built in Data Studio for a client, is basically like a three to six-page Data Studio report, and it would get kind of referred to as a dashboard, but it was… There were many issues with the way that was kind of framed, but I think the like… I wanna throw a few different visualizations together, even if I’m doing an analysis, and I’m doing it in Power BI, and I’ve got all my data pulled in, and maybe I even wanna be able to refresh it, periodically, I wanna provide that.
0:14:14.6 TW: Great, but don’t call that like a dashboard, ’cause I think, Moe, to your point, the dashboard implies, “This is, everytime I get in my car, I’m looking at the dashboard, everytime I’m flying the plane, I’m looking at my dashboard.” Well, if you’re running your business, then what do you really need to see everytime. I’m on a tear.
0:14:39.5 MH: Let’s take a brief pause for a word about our sponsor, ObservePoint, a company that offers data professionals tools to give them confidence in their data and insights. Provide a personal anecdote about a time you didn’t have confidence in an insight, Tim, do you have such a personal anecdote?
0:14:56.5 TW: I’m perpetually wracked with self-doubt, does that… Does that count?
0:15:02.0 MH: I don’t think so.
0:15:07.6 TW: So there was a time last year, here’s one, there was a time last year when a client’s platform, this was not ObservePoint, mind you, but they spit out that a minor change on a non-ecommerce site would result in $8 million of incremental revenue annually, that was laughable, but it didn’t stop the tools account team from putting that in a deck for the client.
0:15:29.4 MK: Eek, that sounds like a disaster.
0:15:31.3 TW: Yeah, it was.
0:15:32.6 MH: Well, I don’t think ObservePoint can necessarily solve all types of analytical silliness, but they can automatically audit your whole website for data collection issues on an ongoing basis, testing your most important pages and user paths for functionality and proper tagging.
0:15:50.0 MK: And the platform can also alert you immediately when something goes wrong, as well as track the results of their checks over time, so you can see if you’re hitting your data quality targets.
0:16:00.8 MH: So if now is the time for you to get serious about data quality on your site, and it is, head over to observepoint.com/analyticspowerhour to learn more about ObservePoint’s full data governance capabilities. Alright, let’s get back to the show.
0:16:16.3 MK: Tim, I think, maybe we’re approaching this topic a little bit differently, or you obviously have some bug bears that are unique. I think one of the things that I can’t reconcile in my head is about the tooling, so a big part of the problem I’m facing is that, I guess, people are not using good data viz practices when they’re building dashboards, and that’s the thing that’s frustrating me and I wonder how much the tooling is responsible for that and how much the analyst is responsible for that, because some of the problems are that the newer tools, like your, well, Mode has been around a while, but it’s very crass and then there’s Looker, which I mean, I do love Looker, but it has some serious limitations when it comes to data viz, even maybe a little bit Data Studio, although it has improved a lot and then there’s open source, like Superset, so of course the functionality is degraded versus like a Tableau, that’s been around forever. And sometimes I wonder if the tools are not advanced enough and the features that make good data viz easier on the analyst, or if the analyst is just being lazy and not trying to work with a newer tool to present things in the best way possible.
0:17:35.5 TW: I have thoughts on that too.
0:17:38.8 MK: Surprising.
0:17:40.5 TW: I think Tableau has sort of set a very high bar, and I am not a Tableau user, I’ve become a little bit more of a Power BI user, and I think Power BI has done a good job of chasing Tableau, but I remember years ago, Lea Pica, you mentioned her earlier, she did a blog post, ’cause she was using Domo and it was called like, “How to Make a Brain-Friendly Bar Chart in Domo?” and that was her whole blog post, was going through like, “Here’s the defaults, and here’s all the stuff you should tweak, and here’s the stuff you just can’t fix,” and that was… I think she was expressing a little frustration, I wound up taking that and saying, “Okay, let me do the same exercise in R,” and granted, I don’t know that you wanna be hand-coding your charts, but it is kind of a benefit of using… If you’re coding it, I can kind of match, I’m gonna say the flexibility of Excel, like, “How the hell did Excel manage to nail the ability to really do things with a lot of flexibility, and many of these other tools have just…
0:18:45.0 TW: And I get it, they’re having to program the widgets, they’re using little visualization plugins, they’re using open source within their open source package, but when you say that, “No, I wanna change the formatting of that number.” “No, I wanna lighten up these grid lines, but leave this dark,” and you’re stuck, “I wanna tweak this, I wanna put a label in this one spot,” and they don’t… ‘Cause I don’t think… I think that, that is the…
0:19:15.4 TW: Those platforms have fetishized the pulling the data together and doing a slick demo over actually recognising, and to Tableau’s credit, bringing Stephen Few on and saying, “This really matters. Help us, coach us.” And you can still do crap stuff with Tableau, but Tableau has that flexibility. Power BI has a lot of flexibility, but it’s frustrated me still a little bit.
0:19:45.5 MK: But maybe it is a shared responsibility in terms of the failure, because one of my sources of frustration is like… And I really do apologise to my team ’cause it’s not everyone, but there are… We have the best practices for code reviews and standardisation across our programming. It is phenomenal. We did the same thing. So we have a SQL style guide, we did a dashboard style guide, I actually created it, big shock, and a peer review of dashboarding. No one does it, and this is in a team where code reviews are ingrained as something you do, but when it comes to best practice of data viz on dashboards, and the guidelines we should use, and dashboard reviews, people are like, “Is it that important? I built the dashboard, I churned it out.” And it kills me. It really kills me.
0:20:38.2 MH: It’s the same thing. It’s sort of like when people make a product, they think obsessively about who’s gonna use that product every day, and what it’s gonna be like, and I think it’s that mentality we have to sort of embrace when we’re building something for people to consume as a dashboard or some kind of reusable piece of data information, where you have to obsess about how it’s being used, what do you do with it, when do you open it, how often do you look at it?
0:21:05.5 MH: Even in the way that I work now, I go back and look at dashboards as something that we should come back and look at as we refine our approach, because there’s some stuff that we’re like, “Hey, where are you going? We are never talking about this section of this dashboard ever. Let’s take this out or do it a different way. Let’s swap it out. Everybody still looks at this, and it doesn’t necessarily have to be the world’s prettiest thing, as long as it’s useful.” I know that’s where Tim and I kind of diverge because I learned a lot of dashboarding at a time where my entire team demanded all the data and tables, and it’s like, “No!” But it is what it is, right? And at the same time, that’s what I feel like…
0:21:45.3 TW: They’re getting it in tables or they’re getting shit visualizations. Yeah, they’re saying, “Give it to me in tables so I can take it and pull it out… ”
0:21:51.2 MH: Listen, I’m totally on the same page with you in terms of what makes a great data visualization, and certainly as the quintessential analyst, I would defer to you on data visualization questions.
0:22:04.7 MH: Part of why I wanted you to come to Search Discovery in the first place was to elevate the level of the work that we were doing, and you did, so I’m a genius really.
0:22:17.4 MH: ‘Cause I can’t do it, so get Tim Wilson in here. But that’s the thing I think that is missing a lot of times, like to your point, Moe, it’s sort of like, “Let’s just do it. Let’s just get it out and get it done.” It’s like, “No, actually, it’s sort of… That should be something you craft, really, really… ” I feel like we don’t put enough work into them, is was probably what I would say is the biggest thing I feel like is a problem.
0:22:42.2 TW: But I think there’s two… And to be clear, I’ve done a CXL course on data visualization, data storytelling, I’ve got multiple clients that I’ve delivered it to, I have a 20-part modularised thing internally, so I might have a few thoughts. And having presented on it, it’s to your point, Moe, that people don’t believe that it matters, and so I’ve gotten to where I pretty much refuse to start with “These are data visualization best practices.” I start with, “Let’s talk through why it matters,” and it kinda comes down… What I typically do is I wind up talking through… And this, almost all of this came from Stephen Few’s information dashboard design, although I will now… Storytelling with Data by Cole Knaflic is my, hands down, if anybody will read a book on data visualization, that’s the one to read… But Few talks about the three types of memory, the iconic memory, short-term memory and long-term memory, and the whole Miller’s Law with short-term memory is the human brain can only hold seven plus or minus two chunks of information in their brain.
0:23:51.2 TW: So if you’re building anything, a dashboard or a slide or a report where you’re saying, “This is an engineering challenge to get all the information represented,” the short-term memory, which is where the processing happens, can’t take in nearly as much as you think. So that’s like point one. Point two is the curse of knowledge, where when the analyst is looking at the data and it makes sense to them, they’re super familiar with it, so we can talk through the tapping game and other ways that the curse of knowledge is terrible. And then the third is this idea of cognitive load, where… What I love to do, and this is in Few’s book… I think he may have actually pulled it from somewhere… Is this, it’s just like four or five rows of numbers, so it’s data but in kind of an artificial way, and they’re kind of grey numbers just randomly covered, and if I put that slide up and say, “Count the number of fives that are in this. And take your time. There’s no trick here. Just count the number of fives. Okay, what do you think?”
0:24:58.2 TW: And I think the right number is like six. And then we flip to the next slide, exact same thing, except the number of fives, every five is a certain colour, a different colour, and it’s bold, and I say, “Now, how many fives are there here?” And you instantaneously can see that there are like six, and then I point out, “How many pixels did I actually change on this display? Very, very few. This was an incredibly subtle difference that I made, but your brain had a wildly easier effort when it comes to interpreting it.” And then I tend to say…
0:25:37.3 TW: That is experiencing cognitive load like you literally just… And if you’re not thinking about that, when you’re producing your data visualizations, you are doing the difference of that, and then I walk through other… Other examples are kinda ripped off from Steven Few, and then Michael, this will be my second, the other point when people say, “Why do I need to put the time into it?” I think there’s another fallacy. People tell me, “How do you… You took that chart that I’ve been labouring over and you in 10 minutes made it so much better, that’s amazing!” I’m like, well, guess what, I’ve done that a thousand fucking times. And you know what? It took me a long time, and so, you kinda have to build up that muscle. And guess what? When I produce a chart, my starting point is much better. So if you believe that it’s important and then you say, “I don’t have the time to do it,” well, you gotta make the time, ’cause you gotta build the muscle and you’ve gotta learn, and then it gets to where it’s not taking you any more time. And I feel like we don’t recognize that either. If they never start practicing, they say, “I don’t have time to do it.”
0:26:43.6 MH: Yeah, and I think that’s a really…
0:26:45.5 TW: I’m just gonna take a break. You guys just, you know, talk amongst yourselves.
0:26:48.0 MH: No, that’s a really good point. ‘Cause you’re absolutely right, the more you do it, the more you practice, the last time it’ll take. I mean, take the time to do a lot of communication back and forth between the users and you. So, take the time to refine the dashboard over time as well. But your point is really good, which is, yeah.
0:27:10.1 TW: That wasn’t what I was saying.
0:27:11.3 MH: No, no, no.
0:27:11.5 TW: It’s a good a point, just not really…
0:27:14.5 MH: I’m amending my point slightly to more adhere to yours, Tim, is what I’m doing. Okay?
0:27:19.0 TW: That’s fair.
0:27:19.6 MK: Okay. What about the tool though? Because, one of the things that I find is a barrier, for example, I had this weird nuanced example where I think the date format should be the same on all of our fucking dashboards. Because it makes it easy for someone to understand when they’re going from one dashboard to another, one graph to another, very easy if all analysts just use the same date format. It is simple, right? But every.
0:27:46.8 TW: And as long as it’s not the American format, which is just stupid. In the…
0:27:51.8 MH: Month, day, year?
0:27:52.6 TW: Why? Yeah. Yeah. Sorry.
0:27:56.4 MK: I quite like day, day, MMM, year, year. Because I feel like that, I don’t know why. Anyway, that’s my personal preference.
0:28:06.2 TW: That’ll screw the Americans. So, I do year, year, month, month, day, day. ‘Cause then even the Americans don’t get confused. Yeah.
0:28:12.4 MK: Okay.
0:28:12.7 TW: But carry on.
0:28:13.3 MK: Well, anyway. The tool that we use, the last time I checked, you have to go in every single graph and change that. You can’t set default, all graphs to this date setting, right? And I see that as a massive barrier to the team. ‘Cause I get it. It’s a pain in the ass changing out in every graph that you ever build for every dashboard, and we do have hundreds of dashboards, whether that is right or wrong is another discussion point. And so I do wonder… I don’t wanna… It’s not about pointing fingers, but it’s like you talk about an analyst taking time to do things right. Well, in this case, is that up to the analyst or do you think that the tools need to evolve and understand the requirements of analysts better so that they build those features more quickly?
0:29:01.7 TW: Well, I’m leery about throwing the tools under the bus, as you know.
0:29:07.1 MH: Uh, yeah.
0:29:09.2 TW: No. Curse every tool? No, that’s a great point. And the tools should do better.
0:29:18.5 MH: Yeah. I agree, Moe. I think there’s some anchoring in a dashboard that should be common, like the same date format, probably the same location, same thing with the title, and maybe a couple other things that I’m not thinking of right now, but that probably are important, that need to be in the exact same location, no matter what you’re looking at. It’s the same reason we go and look at a style guide to look at how we wanna maybe do colour schemes and complementary colours because you wanna tie it into markers that connect people with whatever it is they’re looking at, right? So, it’s like you’re looking at, this is for our company, so it’s gotta have some of our brand’s colours in it. And those kinds of things, I think make a ton of sense. But yeah, the other thing I think is nowadays, in the modern dashboarding and data building time that we’re in, you don’t have just analysts building dashboards, you have engineers building dashboards a bunch of the time too. And I think it’s a very different skill set to assemble the data, put it into a view for a dashboard, and then create a visual on top of that, that’s multiple disciplines at play, and I think we’re not being really honest with ourselves about, a lot of times who we’re asking to do what. And I think that shows up in the work a ton.
0:30:38.9 MK: But I don’t think there’s an issue with an engineer or a PM building a dashboard.
0:30:42.0 MH: Oh no! I’m not saying they can’t do it, it’s more…
0:30:44.4 MK: Yeah.
0:30:45.1 MH: You’re asking them to do something that’s not common in their experience, right? So they have to learn that skill set just like an analyst does.
0:30:52.3 MK: One thing we have done, which I think is a really smooth move, is that you…
0:31:01.8 TW: Your team’s not listening anymore, they’ve all…
0:31:01.9 MH: That’s right.
0:31:03.0 TW: Oh, she’s just, it’s too late for a recovery.
0:31:05.4 MK: We do encourage people. I do think that people should build their own dashboards. I have dashboards, one’s called Moe’s Cheat Sheet, ’cause it’s like, I have to do these monthly numbers, so I build something and then it has all the graphs that I want and I can just plug it into something for a monthly report. So you want to encourage people to build their own things. It’s when those own things get shared out, it starts to become problematic. So what we started to do is in the title, we have a little image, which is basically a data analyst certified dashboard. So anyone looking at it knows if a data analyst has QA’d that dashboard or not. And I think that is a really good move.
0:31:41.2 MH: That’s a really good move, yes.
0:31:43.3 MK: Because it encourages people to build their own. But if you’re a PM and an engineer has suddenly shared a dashboard with you, you can be like, “Oh, actually, do you mind getting a data analyst, or I’m gonna use these numbers cautiously before checking with the DA,” for example. And I really like that as a kind of a way to manage that.
0:31:58.1 MH: Alright, let’s step aside for the segment that is becoming a fan favourite, and I think we all know why, because it’s the Conductrics Quiz. So that quizzical query with conundrums for the most advanced…
0:32:13.2 MH: People, I don’t know more like que-something. Quizzical. Anyway, a quick word about Conductrics. So before we get started, AB testing vendors a lot of times promise you a silver bullet to make experimentation really easy, but it’s just not the truth. It takes a lot of hard work and you need a technology partner that’s willing to admit that that’s the case, and they’re innovative and forthright about those challenges, and that’s where Conductrics steps in. They’ve been that kind of company for over a decade, helping solve those problems with best-in-class AB testing, contextual bandits, and predictive targeting. You can actually check them out on their website at conductrics.com. Okay, Moe, are you ready?
0:32:55.2 MK: Barely.
0:32:57.4 MH: Here is… You are competing on behalf of listener Anthony Mandelli. He is a really great guy and if you ever wanted to know something about cryptocurrency, talk to Anthony. He knows everything about it.
0:33:09.5 TW: Ant man.
0:33:12.4 MH: Yeah. And then Tim, you are representing listener Tony Altic. So Tony is… So it’s kind of two Tonys, Anthony and Tony. Okay, so there you go. Alright, here we go. Moe and Tim are again, quietly chatting, drinking tea. I find it interesting to think you two just sit around drinking tea calmly. Anyways, Moe is enjoying a robust English breakfast. I don’t… And Tim is sipping a smoky Lapsang Souchong.
0:33:39.7 MH: I don’t know if I pronounced that correctly, but this is important to the question.
0:33:43.5 TW: Is it?
0:33:44.4 MH: No, it’s not.
0:33:46.6 TW: Okay.
0:33:46.7 MH: Again, I somehow, am always sort of the perplexed person barging into the room, so I will calmly walk into the room with a calm and healthy demeanor on my face. That’s not how it was written at all, I’m supposed to be super panicked, but you just go with it. My client wants an opinion if they should offer free shipping. The costs only makes sense for them if less than 12% of the shipments weigh more than 20 pounds. The average, the mean shipping weight is 14 pounds with a standard deviation of two pounds. So do you get those numbers 12%?
0:34:26.2 TW: Yup.
0:34:26.1 MH: Okay.
0:34:26.2 TW: Wait, yeah, so less than 12% would have to be…
0:34:28.1 MH: Less than 12% weigh more than 20 pounds. Yup. Mean shipping weight is 14, standard deviation of two. Okay, I told them to go ahead with the free shipping because less than 1% of the packages will be over 20 pounds. I was assuming that the weight of the shipments was close to normally distributed. Oh, I’m supposed to say that in a panicked voice.
0:34:52.1 MH: I’m not going to say that in a panicked voice.
0:34:54.4 MH: The problem is I just found out that the distribution doesn’t look normal at all.
0:35:00.7 TW: Oh come on.
0:35:00.8 MH: It is symmetrical and it looks like the data is multi-modal with multiple clusters of packages of similar weights that aren’t close to the average weight. I have no idea how many packages are going to be heavier than 20 pounds. Now, here’s where the story gets interesting. Both Tim and Moe put down their tea cups, do a quick mental calculation and say simultaneously, “Don’t worry about it.” Me, Michael, mystified as usual, “How can you be sure?” They look up and smile at one another and say, and this is the question, what did Moe and Tim say at that moment? Was it A, Gauss Markov? Was it B, Chebyshev’s inequality; C, Cramer-Rao bound; D, Jensen’s inequality; or E, Lipschitz continuity?
0:35:56.3 TW: You wanna run through those one more time, it’s not gonna help me?
0:35:58.9 MH: Yeah. A, Gauss Markov, B, Chebyshev’s inequality, might be Chebyshev’s, I don’t know. It sounds like my car manufacturer. C, Cramer-Rao bound, D, Jensen’s inequality. And E, Lipschitz continuity. I have this other note from Matt Gershoff saying, this one is actually insanely useful and every analyst should know it in his opinion, so.
0:36:27.1 MK: Do you know who… Okay.
0:36:27.2 MH: That’s just backing information for you.
0:36:27.5 TW: That’ll disparage your sponsor.
0:36:31.6 MK: Since I have no chance of getting this correct…
0:36:35.0 MH: I feel like he makes this harder because people like it. It’s like he just makes it more difficult.
0:36:40.3 TW: I mean, we can start with how many have I even ever heard of, not that I can describe any of them.
0:36:46.7 MK: But can we talk about the hilarity that as you were reading that question, and I noted down the weights. In my head, I wrote down kilos, even though I knew you were saying pounds. Is that not problematic in and of itself?
0:37:01.6 MH: Well, that would definitely if I had 14 kilos, I would be going over the 20 pounds regularly, yes. But anyway, let’s just assume the problem is in kilos for our Australian listeners or even European listeners.
0:37:15.0 MK: It would also be helpful if going forward, we went back to four choices, so we limit the amount of incorrect answers that we can give.
0:37:25.5 MH: Your input will be conveyed to the quiz creator when they hear this episode probably. Alright, so how do we wanna proceed? Does someone wanna lead out with a guess?
0:37:36.4 MK: Or do you wanna… I mean, can we eliminate any two?
0:37:40.3 MH: You have to tell me which ones you wanna eliminate though.
0:37:46.1 TW: I wanna eliminate… I am gonna dive in and eliminate A.
0:37:50.1 MH: Gauss Markov.
0:37:52.4 TW: Yeah.
0:37:52.5 MH: Alright.
0:37:52.6 TW: Am I still alive?
0:37:52.7 MH: You’re still alive. So A has been eliminated, Moe. No more Gauss Markov, just Chebyshev’s inequality, Cramer-Rao bound, Jensen’s inequality, or Lipschitz continuity.
0:38:07.5 MK: I just have no idea.
0:38:12.4 MH: No, don’t…
0:38:13.9 MK: I feel like when it’s… When I have no chance of getting it right, I don’t know, do I just try and pick one as… I think it’s this based on no rhyme or reason, or do I try and eliminate one? I feel like I should try and eliminate one. That would be better, but…
0:38:31.8 MH: Yeah.
0:38:32.0 MK: Okay.
0:38:33.5 MH: Eliminate one.
0:38:33.5 MK: I’m just randomly choosing one to eliminate, which is C. Cramers…
0:38:37.6 MH: The Cramer-Rao Bound, so that’s not your guess this one you’re saying.
0:38:41.0 MK: Yes, it’s not that.
0:38:41.2 MH: Absolutely, that one is not the answer.
0:38:43.6 MK: It’s not that.
0:38:44.2 MH: Alright. Cramer-Rao Bound is out of there. Now you’re down to three choices. Look at you two.
0:38:49.5 TW: See? And I was feeling like we had two inequalities, and I feel like inequality is kind of the… Didn’t say… He didn’t say bi-modal, he just said multi-modal, but it’s symmetric. So I feel like it’s an inequality one, and I have not heard of either one of those, but I am gonna think that it is B or D, so I’m gonna eliminate E.
0:39:14.2 MH: You’re gonna eliminate E. Okay, well, congratulations, I must tell you, you successfully eliminated the Lipshitz Continuity, so yeah, now the choice is basically just between B and D.
0:39:25.9 TW: The pressure’s on.
0:39:26.8 MK: Okay, so I’m gonna choose one now.
0:39:29.6 MH: Okay.
0:39:30.3 TW: Or you can eliminate one, either way.
0:39:32.0 MK: Either way.
0:39:32.5 MH: No, ’cause there’s only two answers left.
0:39:34.8 MK: I’m gonna choose one.
0:39:35.5 TW: This is like the runner up. You know, this is like when you crown Miss America.
0:39:39.6 MH: That’s right.
0:39:40.4 TW: Yeah, yeah.
0:39:41.1 MK: I’m gonna choose D.
0:39:42.6 MH: Who will be the baby of the year?
0:39:45.7 TW: She chose D.
0:39:47.1 MH: Chose D. Okay, so that leaves you with B, Tim is that how you wanted to end up? You guys could flip a coin. Alright, so let’s talk about our two contestants that you’re playing for us. So Tim, you are playing for Tony Altic, and he is a winner because B, Chevyshev’s Inequality is the right answer.
0:40:10.8 MH: Chevy… Chevy… I don’t know.
0:40:13.1 MK: Chevy Chev?
0:40:14.2 TW: So this is where matches wants to make you pronounce things that…
0:40:16.1 MH: Yeah, exactly. I should probably like try to find a little YouTube pronunciation. Anyway, let’s just explain briefly, ’cause we’ve heard from our listeners, sometimes they’d like a little better explanation. Chebyshev’s Inequality holds that no more than one divided by K to the second of the data will lie K standard deviations from the mean for an arbitrary probability distribution. So in Michael’s case, no more than 11.11% of the packages will be greater than 20 pounds. Chebyshev can be very useful for the analyst to determine worst case bounds, and there you go. Just plot that on your dashboard next time, everybody. You don’t have to remember it. Alright, that’s been the conductrics quiz. Always a great time. Thank you, Moe and Tim, for being such great sports, and thank you, Anthony and Tony, for being our contestants. And let’s get back to the show.
0:41:15.3 TW: Back in the… I think we had Cognos Power play and Oracle Discover at the time. Mostly it was Excel basically, but what we did was we had our team, the BI team. We actually did have a style guide and we had enough people who were into it. We had kind of a review process, and our contention was we’re gonna be sharing stuff out in Excel, it is going to look so clearly on brand, and that was corporate colours, best practices. I mean, we basically did a book group on information dashboard design, Stephen Few’s, one of his books. And that was the same like we are going to subtly train the organisation that if they see one of the things coming, it’s gonna look like it came from the BI team.
0:42:00.3 TW: In our BI team, we’re the analysts, and if others won’t be able to… So it was kind of a similar… And we actually had a buddy system of a review. We had two or three people who could go to look at stuff and that we had templates, and back to your point on the tools, even in Excel, there was stuff you could do to set up default, it was a little clunky and somebody got a new machine, you had to go and put a new default that you know, Dot Excel T, or whatever it was.
0:42:32.9 TW: So it was kind of a pain in the ass to set up, but it gave everybody that jump forward, that’s the flexibility and power of an Excel and the challenge of some of these other tools where it’s like, Nope, it’s just… And those may even be features where you’re… I guess like, I know you’re using… You know Plotly behind the scenes, and that’s a feature in Plotly… But you didn’t… Your tool disabled the feature to change that ’cause it caused some other problem with your release and then you’re just stuck. You’re like, this is as close as I can get. And now I gotta either re-create it or copy and paste it if I want it to look actually… You know… Good. But I would also let me just quibble quickly with your… You build your own dashboards, that’s using the… To me, that’s using the vendor’s language that they say, This is a dashboard…
0:43:29.1 MH: Yeah.
0:43:29.5 TW: You’re not actually building a dashboard…
0:43:31.0 MK: Yeah, I agree.
0:43:32.0 TW: You are building a set of reference visualizations that will auto-refresh that you can then pull in…
0:43:35.7 MK: Yeah, 100%.
0:43:36.7 MH: But… That just doesn’t roll off the tongue.
0:43:40.3 TW: Yeah, but I think that’s part of… Right? That’s like the dashboard becomes the catch-all. Yeah, we want people to… We want people to find efficiency. I’ve got a co-worker now and she’s a big Power BI user, and she’s like, “Look, I just… I built these 20 things, I don’t know any given week or month, which one of them I’m gonna use. I’m gonna build them all. I can refresh them all automatically, and then when I go to put this report together, I can look at them all and pull the two or three that I want.” That’s just efficiency. It’s Power BI, which you call it. I am now, I’m blanking on the… Man, they change their names often enough… Yeah, it works the… I don’t know… Whether that she would call that a dashboard or not, but yeah, end of rant number 47.
0:44:25.6 MK: Yeah, I audit… I mean it’s confusing, and I feel like you’re really focused on how that term dashboard is used. So like take…
0:44:35.4 TW: Words matter, Moe.
0:44:36.6 MK: Technically…
0:44:36.7 TW: Words matter.
0:44:36.9 MK: Yeah, I know, but technically right…
0:44:40.0 MH: Listen, little lady. Why did we have to sing that old song at the beginning, Moe, if we’re not gonna talk about dashboards.
0:44:45.6 MK: But technically, I mean, what I created… Is a bunch of tables, right, with gradient colors? That is what it is, but in Looker’s tooling I built it on a dashboard.
0:45:01.7 TW: Yeah.
0:45:02.5 MK: And that’s the thing, is that’s actually why I like Tableau, because Tableau does have… What do they call… They have some other things that are called like works. I can’t remember, but some other things that are not a dashboard, but, and you can basically build a table on, that auto-refreshes, and I think that’s a nice add-on. So I’m just kind of wondering in my head how much, maybe, I need to reclaim the word dashboard in the organisation to start this journey in the right direction. I don’t know.
0:45:33.9 TW: But it’s tough. If you’re using Looker, and they use it to refer to a set of functionality, it’s gonna be an unwinnable thing, if that’s what the tool calls it, and it says file new dashboard, then you’re like, “Yeah.” But that’s not really a dashboard, that’s just a Looker. I mean, it is tough. Yeah, you…
0:45:53.4 MK: So don’t get me started on Mode then. So is the thing that probably worries me…
0:46:00.2 MH: We weren’t trying to.
0:46:00.7 MK: But okay, I know… I feel like Tim’s gonna have a real… I don’t know. This, I really struggle with. Okay, so query from stakeholder, why has sign-up right declined? And someone will build a dashboard that has 30, or 40 things to do with sign-up right, and they will give that dashboard to a stakeholder and not even like, “Here’s why it declined,” just like, “Here is a dashboard on sign-up rate. This will answer all of the questions.” Okay, so there are times where someone’s been like, “Hey, Moe, a dashboard was built for this three years ago. This will answer your question.” And then I go look at the dashboard and I’m like, “Fuck, this actually really was really helpful. It did answer my question.” But what pisses me off is that the last time that dashboard was refreshed was three years ago. And what happens then? I feel like there is some kind of responsibility for when you build something, if it’s not getting refreshed, and getting used, it should be… We got to this stage with Mode, where we had thousands of dashboards, and no one could find a freaking thing, ’cause their search functionality sucked. And it’s like these dashboards just die in this pile, but no one does anything about it, because people are also using them to present analysis, which I have very wary views on.
0:47:23.2 MH: Yeah. I feel like that before Tim gives us the real answer, I’ll give a middle answer.
0:47:30.8 MH: No, ’cause I… Honestly, Tim, I kind of think you do have all the right answers for most of this stuff.
0:47:35.3 TW: Yeah. [chuckle]
0:47:36.3 MH: But I agree, Moe, I feel like there’s this other layer which is… And the way the world has changed, I call it analysis, but maybe that’s not appropriate anymore, but it’s going in and using the data to take a narrative to people of understanding, right? So the dashboard kinda tells you what question to ask, and the research tells you what… The other combining factors that do this, and then what can we do about it, or where should we go from here? And so it sort of becomes a sort of a step process. But that’s not a dashboard that you use for that. That delivery has some other mechanism. I don’t care how people wanna deliver it, you can deliver it as a doc, you can deliver it as a deck. I am a consultant, so I’ll do a lot of that delivery in Google Slides, or PowerPoint, but…
0:48:23.8 MK: Or Canva.
0:48:25.3 MH: Or Canva. Yeah, no, I’ve never done it with Canva, but…
0:48:29.1 MK: Great charting functionalities.
0:48:29.1 TW: Already lost your team, you’re recovering…
0:48:34.7 MH: You know what?
0:48:35.2 TW: Yeah.
0:48:35.4 MH: I don’t think you wanna invite Tim into Canva to figure out whether or not he feels like the charting is up to stuff, because that’s gonna be a scary day for Canva, and we wanna stay positive. Okay, but… No, but the point is, is like, yeah, don’t hand a dashboard over to someone asked a question, hand over an answer to their question where you’ve gone and explored the data to give more perspective. ‘Cause obviously, the question should be coming from the fact that they saw some data, or understood some data, and it drove a question to them.
0:49:04.9 TW: I mean, to me, it just gets back to the kind of the labeling. So definitely, it is not a new problem that once somebody has built something, and it’s automatable, there’s zero cost to keep it on…
0:49:18.8 MH: Yes, yeah.
0:49:19.4 TW: Even if it’s sending out an email, a daily email, right? And analysts, they’re like, “I’m done,” and nobody stops and says this has been going out for years.
0:49:27.5 MH: I would like to stop those emails, yeah.
0:49:29.7 TW: And literally, people are… “Oh, can I click on… ” They don’t even click on subscribe, they’re just like, “Yep, Oh, I feel like I’m staying on top of my inbox, I get to delete that every day.” But I think a one-time analysis, which sometimes would be a “Oh, we’re gonna iterate a little bit, or we may add some logic.” So it definitely is nice to be able to build it in a platform where I can say I may iterate a few times to really explore this back and forth with my stakeholder. They’re like, “Oh, shoot, I forgot, could we try that again for like a year ago? Was it the same thing?” There’s value in that, in having the tools that are wired up. Or I’m running a campaign, the campaign is running for three months, I wanna be able to monitor the performance during that three months…
0:50:13.4 MK: But that’s the keyword, monitor the performance, monitor, someone’s gonna be looking at it.
0:50:18.4 TW: Yeah… Well, but then it’s like… Then the campaign’s over and well, now we got to do the campaign read out, so we wanna use the same dashboard. It is a problem. There is, I guess… I think there is a governance issue, and that’s, again, an opportunity for tools to say… I mean, my to-do list application, I can go archive stuff. I can archive projects. And then, you know what? If that client comes back, I can go and un-archive it. So it’s a…
0:50:45.6 MH: That’s a good functionality. I wish for that.
0:50:47.7 MK: That is a great functionality. Damn, we need a dashboarding PM to listen to this.
0:50:51.7 MH: Yeah.
0:50:55.6 TW: ‘Cause you don’t want it to go away. And you actually… If somebody has a direct link to it, you want it to… I just had this with a co-worker, she’d done something and she had… Not a dashboard, in Analytics Workspace, and she sent a deck. Somebody was like, “Have you done anything like this?” She was like, “Hey, here’s this thing, this is maybe what you’re looking for,” and he was like, “Whoa, how exactly was that set up? Do you have that workspace?” She was like, “It’s actually linked. On the first slide there’s a link to this Analysis Workspace.” I was like, “Oh, Theresa, that’s amazing, you literally… ” And she could have… Analysis Workspace doesn’t do a great job at that governance either, where is the place I can go put it to say, “As far as I know, no one’s ever gonna wanna look at this again, but if somebody does, I haven’t completely killed it.” And the usage of it will show me all the… If it’s in Mode, can I see all the ones that haven’t been used?
0:51:50.0 MH: And give you good metadata for finding it later, Moe, that kind of stuff.
0:51:55.4 MK: And what about… It’s pretty common practice now, most SQL tools are starting to build in some data viz capabilities. And Tim, you made a point as we were pulling out thoughts of this topic around the blurring of tools and how it could be contributing to the topic. I can see a time and a place where some of the SQL tools start to go further into this territory. And you have it already, right, lots of the data viz tools are now starting to do data prep, and that’s becoming a really big, I guess, feature of their product. Is it gonna go… Do you guys think it’s gonna go that next step where SQL tools start to try and become data viz platforms and… I’m not gonna lie, it kinda scares me.
0:52:46.2 MH: The answer is yes, and it hurts, because everyone’s doing such a terrible job, and it’s mostly because there’s so much ground to cover. I have… I don’t know the right way to say this, this blanket view of a lot of the startup space in analytics and even data viz just generally, which is, might be cool for one or two things, but almost 100% of the time they haven’t thought through the problem set sufficiently to build the things that they need to do to actually solve real problems. And so it’s like, okay, that demo is really great and people are buying it, just ’cause they’re not thinking through the things they actually need to have this tool to do to function well. And so we end up with this really… And this is one of those areas where people just… There’s not enough of Tim Wilson to go around to… No, I’m saying this only because…
0:53:44.2 TW: You’ve hit your…
0:53:45.1 MH: Listen, Tim, I’m not trying to be snide, I’m serious. There’s a lack of education and a lack of sophistication when it comes to how we integrate data into everyday usage in things like a dashboard, so that people can ask good questions of potential vendors to get to the actual ways that data is brought in, understood, and acted on. And the reality is, is people are very thrown off by, like, “Look at this cool Sankey chart I can just whip up,” and it’s like, “Who cares? Sankey charts are terrible, no one should almost use them ever.”
0:54:24.2 MK: Yes.
0:54:24.7 MH: It’s like, “Okay, you made a Sankey chart, now go get me another visualization that shows me something I can do with that data, because none of that is gonna be worth a damn.” And I think that’s probably my biggest complaint. And so there’s a million, there’s a million… If we could sit here for the rest of the show and just name vendors one after the other, and probably come up with ones, probably 50 more that none of us have ever heard of before, and they’re all doing the same thing which is they’re solving one slice of this, and maybe they’re getting that slice really good, but there’s the other 95% of the pie chart… See what I did there? That they’re totally not addressing.
0:55:10.0 MH: Anyways. Okay, I’m gonna stop, ’cause that’s… That’s where I struggle. And that’s where I feel like Tableau, who’s been around a really long time, Power BI, who has the benefit of owning Excel and learning all the lessons they’ve learned from that community over the years, and… Well, actually, I like Looker too, although I feel like Looker has a little bit of gaps in this regard, in that they don’t let you organise the information the way I’d like to see it organised, but it’s very powerful…
0:55:35.7 MK: I will… I disagree with that. I think the one thing I do love…
0:55:38.9 MH: I’d like more flexibility… I don’t wanna get into the details of the product as much. I like it too. The point is, is there’s some vendors who are definitely crushing in the space, and there’s more that I’m sure I’m not getting to. But the point is, is 95% of them are a complete waste of your time, from my perspective.
0:55:57.0 MK: But one thing I do wanna draw attention to, is that I think tools like Looker, that have thought about how the data is structured before you even build a graph or then put that on a dashboard, have probably learned something from the Tableaus and whatever, which basically were like, “Throw in any data that you want and we’ll find a way to visualize it.” I feel like Looker has actually transformed that way of thinking, and instead gone, “We’re gonna be really rigid about how you pull your data in and the structure.” They have this thing called LookML, which is basically its own mini SQL to how you set up the data before it comes in, and even though it is a right pain in the ass, it was the best thing that ever happened to us because it made us be really thoughtful about every single metric, how it was calculated before we pulled it into a chart, and that actually is something that maybe is fixing some of the problem.
0:56:58.0 MK: And there were issues, I know, in previous iterations with previous teams where people were making their own calculated metrics and people were doing it incorrectly, and there was no way to QA that. That then got fixed, and then you go in this cycle where then, now they’re like, “Actually, we’re gonna let you do calculated metrics,” and it’s like, “Wait, you fixed that problem really well in your product, and now you’re re-introducing it because it’s something that people are used to.” So… Ugh…
0:57:25.5 MH: Yeah, that’s… Again, it’s the one of the ones I like, I like Looker, so it’s ones I… I’m definitely down with it.
0:57:31.7 TW: Well, but it is a tough challenge. The products are trying to do, they’re trying to… They have to appeal to the market where the market is. So I think, to Michael’s point, if I could flip a switch and make the market demand using analytics in a better way, that would help ’cause that would then drive it, so they have to respond to where the market is, and then the other is, they are all trying to grow. Half of them are venture funded. How do they grow… To your point, Moe, they go down market and up market. They look at the entire… They all basically are trying to march towards, “We’re the single tool that you want.” Adobe is kind of an extreme, although oddly, Adobe doesn’t have a dashboarding solution, right? They don’t have a…
0:58:16.8 MH: Love to see them come up with one. Well, maybe I wouldn’t. I don’t know.
0:58:20.4 TW: But all of them are trying to go like if Tableau said, “No, we’re pure visualization,” you get it, but no, Tableau started getting into where we’re gonna hook into the data, we’re gonna do…
0:58:29.6 MH: Yeah, they started pushing down.
0:58:30.9 TW: Yeah, and so I think it’s going to happen. It is kind of nice, like I mean, with one vendor. It’s not a visualization or dashboarding tool, where our clients have enough clout catching that vendor at the right stage, that we’re saying, this is the process that makes the most sense for the market when it comes to managing hypotheses. We can actually help influence the product roadmap, and we have enough scale with our clients who are doing it that they say that makes sense, and we’re gonna listen to you. But we don’t run into that with the… I mean, a lot of them, these kind of fly-by-night, the whipper snappers who are like, “I’m gonna solve analytics with my tool, it’s just open source stuff I’m gonna cobble together.” And it does, it takes years. And that’s a great point about Micro, and Microsoft went Excel, Power Pivot, Power BI, like… Yeah, they’ve… And oh, by the way, they’ve got Azure. They’ve dealt with really big data, so they actually, as much being a laughing stock…
0:59:36.0 MK: You are really into Power BI now. Like this is… You’re on the train.
0:59:41.6 TW: Well, I mean, give me any tool, I’m gonna get into it and play with it and have… But even then, I manage to run into visualization frustrations, where I’m like, “Really? I can’t… I don’t have enough fidelity over the size of that font?” Which I guess that goes to me… Kind of when you were talking about Looker, I was thinking like, oh, well, that is with R, I love the fact that I can blow through and transform the data to be exactly how I want it to be structured to visualize it, and now I’m visualizing it with code, which means I can tweak, and I know that is not the answer. The answer for data visual… I mean, people will talk about building shiny dashboards really quickly to meet their client’s needs. And I’m like, that can’t be the answer either, that can’t, ’cause then it’s back to needing engineer, you need coders who can also build a good interactive visualization.
1:00:39.8 MH: And that’s one of the big drawbacks of Looker is you can’t build anything without some SQL capabilities, right? You have to have some SQL skills. But you’re absolutely right. There’s layers to these tools, and basically they’re BI tools, I guess, and it’s sort of like you have the visualization layer, which is sort of where all of the consumers, the people we’re gonna push this out to, they have no idea what’s going on under the surface. But that really matters to the rest of us ’cause of like to your point, Moe, some of the things you were talking about was sort of how maybe you can insert a calculated metric and maybe it’s not standard, so it’s happening a different way over here than it is over there. And you’ve gotta go there, but if you just look across at like the visualization, I mean for a long time, Tableau was the stop. It was the best you could do visually with anybody, and it was just sort of like but you had to work, work, work, work, work your way up to…
1:01:30.6 MK: Yes.
1:01:30.8 MH: Getting your data to the point Tableau could accept it. And so that’s what created the opportunity for everybody else, and then Tableau was like, oh, we gotta do data prep, so let’s move down into that space as well. But yeah, anyway. Okay, so we’ve been just rambling quite a bit and going over, so let’s do this, let’s…
1:01:50.6 TW: Ranting?
1:01:51.1 MH: We’ve been ranting, yes. Ranting.
1:01:53.0 TW: Yeah.
1:01:53.5 MH: ‘Cause frankly, we’re frustrated, we’ve had enough analytics community, and it’s time for you to step up, but here’s the things we want you to do. Moe, what’s the first thing we want people to do? And you’ve already mentioned a bunch of things that are really positive, so just remember a couple of those and tell ’em. Tell people.
1:02:06.4 MK: So sorry. The first thing that I want people to do with their dashboards?
1:02:08.5 MH: That’s right. To make a better dashboard world out there, what’s the one thing, Moe, that you can think of?
1:02:14.6 MK: I believe coming up with a few standardisations across your dashboard. So, like date formatting, for example, if you have multiple products use the same blue every time you’re presenting data on a specific product. A few things so that when people are swapping between dashboards, they don’t have to do the cognitive load of what colour is this, and what does this mean? Like yeah, a few standardisations.
1:02:40.4 MH: Nice. Tim, what about you? What’s one thing somebody should do to make a better dashboard world out there?
1:02:47.6 TW: I want them to plan their dashboards and really lock in on a structure that has category objective, whatever with KPIs, and make fewer metrics with richer context around the metrics.
1:03:02.0 MH: I love it.
1:03:03.0 TW: What about you, what’s yours?
1:03:03.2 MK: Gosh. We didn’t even talk about wire frames.
1:03:06.5 MH: Oh yeah.
1:03:06.9 MK: I’m a big proponent of wire frames.
1:03:07.5 TW: Love me some wire frames.
1:03:09.0 MH: Honestly…
1:03:09.6 TW: Did a deck on that too. Got presentations on that too.
1:03:11.9 MK: Of course you do.
1:03:12.8 MH: What I would say is do… Basically try to do usability testing on your dashboards. Like actually go talk to the users and see how they’re actually using it, what is going into it, what’s not and then review and refine. Alright, what’s another… But we’ll go with Tim. Another tip? Another important thing?
1:03:32.2 TW: Well, the downside of going and asking what they want is they will keep adding stuff, so actually when they were being very clear that it’s not the… ‘Cause that is the slippery slope. “Oh, can we get it?” Like you have… It’s a one-time question, we should not add it to the dashboard, every one-time question.
1:03:47.9 MH: That’s very fair. It’s not what do you want on the dashboard, it’s I wanna see how you’re using this dashboard. Different question, yeah. Tim, you have another… We’re gonna do two or three more of these tips just because I feel like if we compress it in, then maybe they’ll forgive us for ranting for an hour.
1:04:07.0 TW: I will say Cole Knaflic’s Storytelling with Data book is… Read it. I would say get that. That’s a self-educating thing.
1:04:17.1 MH: Moe, get another one?
1:04:18.3 MK: I mean, I feel like this is a little bit repetitive, but I’m like don’t build something without talking to your stakeholder, like… My favourite process is chat to the stakeholder, build a wireframe of what they want. And like in Canva, we actually now have a few templates that we’ve created so that when we’re doing a Looker dashboard, we can be like bam, bam, bam, this is what it’s gonna look like. QA it with your stakeholder and make sure it’s got what they need, and I agree with Tim, that’s not an opportunity to be like, “Add this, add this.” If anything, it’s an opportunity to cull and then build it.
1:04:53.7 MH: Yeah. Use brand style guides, read Information Dashboard Design by Stephen Few as well and then take Tim’s class.
1:05:02.2 TW: Yep.
1:05:03.7 MH: You have a CXL… It’s CXL course now? Is that the one…
1:05:07.9 TW: It is a CXL course, yeah.
1:05:08.9 MH: Alright. So there you go. That would be my other tip. You actually seriously, you should take Tim’s class ’cause that… If you wanna get better, ’cause what I found is I have some skills, they degrade ’cause I don’t do this kind of work that often and so when I need to step back in, I have to revamp, and so I need to go back into the source literature and re-read stuff. ‘Cause I… Like Tim, I don’t know how you keep sharp, Tim, that’s crazy.
1:05:30.9 MK: Wait, I’ve got one more.
1:05:32.7 MH: Go. Yes?
1:05:33.7 TW: I’ve got one more too.
1:05:34.1 MH: Both of you.
1:05:36.8 MK: After you build something, take a look at it and see if there are three to five things you can take off, because I think when you’ve finished it, you’re often just like, okay, I don’t need that or this… Rather than what can I add at the end? Say what can I remove?
1:05:57.5 TW: I believe Chapter 3 of Cole Knaflic’s book is remove the clutter. I mean, there is maximizing the data pixel ratio, which I don’t know, Michelle always was like, “Why don’t you just pretend your printer is running out of ink?” Which is actually saying the same thing. Knaflic Cole says to declutter and that… Oh, know the tool. Whatever tool you are using, actually learn how to hack the crap out of it, ’cause you look at Tableau, Power BI…
1:06:26.2 MH: Data Studio.
1:06:26.9 TW: Domo.
1:06:27.4 MK: Looker.
1:06:27.5 TW: There are people who have been…
1:06:29.9 MH: Yeah.
1:06:30.8 TW: Yeah, they all have all sorts of… People have figured out when you’re like, “Can I really do this? Nope, I can’t do it.” It’s like Google it and then get to where you know all of the… I mean, I’ve done stuff with Data Studio where I’ve got the same metric represented three times kinda layered on top of each other, which hurts the performance a little bit, but actually it gives me the context that I want, so that’s the… A lot of the tools, yeah, make sure you really are up against a brick wall, not that you just haven’t figured out either tool functionality or a hack to get around the tool limitation.
1:07:06.0 MH: And then lastly, don’t beat yourself up, just work on the next one and get better. Alright, there you go. ‘Cause that’s more for me, I just have to give myself that ’cause I get stressed about… Because think about it, everybody, like if I create a dashboard and Tim Wilson sees it, you know how much shit I’m gonna get? It’s not even close the amount of pressure you might feel in your job, let me just tell you that right now. No, I’m just kidding.
1:07:29.6 TW: So let me just go ahead and say it ’cause they do listen. Sam, Julie, Neyha…
1:07:34.7 MH: That’s right.
1:07:35.5 TW: I’m sorry, you’re doing a great job.
1:07:38.0 TW: You guys are the real all-stars.
1:07:39.8 MK: Well, on my team, Arth is the only one who listens and I don’t think he’s part of the problem. So keep up the good work, mate.
1:07:47.3 MH: Well, I’ll give a little special shout out to Hector then.
1:07:49.8 TW: I should also say, Dan, you’re doing a terrible job, just ’cause…
1:07:54.5 MH: Hector, you’re doing a great job. Appreciate everything you do, buddy. Alright, we’ve gotta start to wrap up. This has actually been really fun, we haven’t… This felt frenetic to me, so we’ll have to see how it sounds when we actually come back and listen to it, but I just think it’s an area we all have a ton of passion for, it touches our work so deeply, and it matters so much because of all the people who need to use it. So alright, let’s go around the horn and do some last calls. So I think I don’t have to explain to you two what last call is, so since we’re guest list this episode, although it sounds like we should have gotten Cole Knaflic now. We’ll have to work on that. Okay, Tim, why don’t you kick us off with the last call.
1:08:36.0 TW: Well, I’ve got a couple of options.
1:08:36.7 MH: Uh-oh.
1:08:37.7 MK: Of course.
1:08:37.8 MH: He’s reaching for the guitar.
1:08:41.9 TW: We’ll see, let me take this out.
1:08:43.3 MH: Okay.
1:08:45.7 TW: Well, when you’re sitting there in your mesh back office chair, talking to some clients on the phone. I’ll be at my desk all day with my nose stuffed in GAs.
1:09:11.1 TW: And a cup of joe to keep myself awake. Test it now on the home page, test it now. I know you think that button should be round.
1:09:28.6 MH: Round.
1:09:29.4 TW: I can send you some data in the morning. I can send you a dashboard by email.
1:09:43.7 TW: I can send out some insights to your VP and I won’t forget to get credit for the team. I’ll stop there. You may cut that one out.
1:10:01.8 MH: That’s outstanding.
1:10:02.8 MK: Oh, I love it.
1:10:03.9 MH: No, I think we’re gonna do that because… I think what we’re realising is that art is a big part of analytics.
1:10:12.5 MH: And Tim, you’re an artist.
1:10:16.2 TW: I didn’t get to the R and Python references in the second verse.
1:10:20.6 MH: In the second verse. That’s awesome.
1:10:22.1 TW: When you’re sitting back in you’re brand new Cadillac that you bought with money my improvements made, I’ll be in my basement cube with a red bull and Hadoop SMR or python.
1:10:29.6 TW: The swords of my crusade. Co-writing credit to Matt Cohen there but I actually did have a quick last call that was a causal inference cheat for data scientists. I think the for data scientists may not be the best, it basically just gives… If you’ve gotten into causal inference which clearly I have a problem with. It really just goes through four levels from an experiment to a statistical experiment to a quasi experiment to counterfactuals. I don’t think it’s the super cleanest thing but it does kind of show where your strength of evidence is, so if you’ve gotten into trying to wrap your head around why pure experiments are the gold standard and kind of how you start to lose strength of evidence and things get more complicated. That’s a cool read.
1:11:20.6 MH: Nice.
1:11:21.3 TW: I’m gonna write a song about it.
1:11:22.6 MH: Yeah, wrote a song about it, here it goes. All right, Moe, what’s your song that you’re gonna sing? No, I mean your last call.
1:11:29.1 MK: No one needs to take care of that. Mine is really weird. It’s got nothing to do with data but I’m really digging it and I feel like, I don’t know it’s gonna be an acquired taste. But the ABC in Australia is like our National Broadcasting thing. And there is a woman that I’m totally obsessed with, Annabel Crabb. Oh God, it nearly escaped me, but I’m a very big fan of Annabel Crabb. But she’s done a TV-series called Ms Represented on the ABC and I have confirmed by googling I think you can watch it overseas but Ms Represented is a history about Australian women in politics. And she just nails it like everything Annabel Crabb does, like turns to gold. And it’s actually available in podcast form too. But I highly recommend the TV-series on ABC iview, Ms Represented, it’s just phenomenal history of women in politics like I… Yeah, it’s one of those shows you just watch. And you’re like, “Damn, she did it.” Like they did it so well. Anyway, I was very into it. Helps.
1:12:38.6 MH: Very nice. Alright, my last call is an article, actually, I saw, I think first on the measure Slack somewhere. And I looked it up and read it. And it’s written by a guy by the name Steven Finkelstein, Finkelstein, I don’t know how to pronounce his name ’cause I never met him. But he wrote this article at the tail end. I think he’s doing his master’s of analytics at Georgia Tech. And he wrote this article about his journey to getting a data science job offer, and he quantified and calculated all the things that did in that process, how many resumes he sent out. What kinds of coding technical sessions he had to go through and assignments. I just thought it was a really well done for anybody in the job market or thinking about going back to school and getting a Master’s or if you’re in school right now and you’re thinking about, “How do I get a job?” I just thought it was a really practical article, really well-written and well-researched. So it’s called The 100 Hour Journey to Getting a Data Science Job Offer. He said he did get a job. I don’t know. I didn’t look him up on LinkedIn to see where he ended up working. But anyways, I thought it was just a great read. So great job, Steven.
1:13:44.3 TW: I think he’s working at Looker, I think.
1:13:47.3 MH: Oh is he?
1:13:48.1 TW: No.
1:13:48.1 MH: Okay [laughter]
1:13:48.6 TW: No, no.
1:13:49.1 MH: After all this. You know what’s funny, so Tim, you and I have a former co-worker who did go to work for Google. And I reached out to him at one time and was like, “Hey, I’m doing this Looker project. Do you guys use Looker over there at Google?” and they’re like, “No, we just all use Data Studio.” I was like, “But you own Looker, you supposed to be… ” anyways, it’s funny. Not every department at Google is even using Looker. Alright, that’s my last call. Great job, Steven. Alright, you’ve probably been listening and saying “Moe, Michael, not you, Tim. But Moe and Michael, you got it all wrong. And here’s what you really should be thinking about dashboards.” No, but we’d love to hear from you. Do you have great examples. That’s one of the things like sharing great data visualizations, techniques you’ve used with different tools. Those are great things to share with the community and they build us all up. So feel free to share if you shared on the measure Slack or LinkedIn or on our Twitter, we’ll be sure to kind of try to retweet it or like it or whatever we can do to share it out with others too. So we can all gain in our knowledge.
1:14:48.3 MH: Alright. Well, no show would be complete without chatting a little bit about our good friend Josh Crowhurst, who is our producer who also follows in Tim’s footsteps in creating data visualizations and dashboards that are extremely high quality or so I assume, I don’t know if I’ve actually…
1:15:05.8 TW: And I think he has gone well past either of our footsteps when it comes to actual musical.
1:15:10.3 MH: Yeah, yeah, no, not even like…
1:15:13.1 MH: That’s… Yeah, we’re not…
1:15:15.8 TW: He’s he’s gonna be crying by the time he gets to…
1:15:18.6 MH: We’re not “competing” on everything over here. Anyways, we’re very thankful…
1:15:23.3 TW: Josh, I’m sorry, your ears are bleeding.
1:15:25.3 MH: For all the talents you bring to the podcast, Josh, thank you very much. And listen, I know that creating dashboards can be a trying process and be frustrating because not only the stakeholders but the tools themselves and even our own skill level. But no matter what you do and what you’re trying to do with dashboards, just remember, on behalf of Moe and Tim, my two co-hosts keep analyzing.
1:15:50.6 Announcer: Thanks for listening. 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 Crowhurst.
1:16:08.6 Charles Barkley: So smart guys want to fit in so they made up a term called analytics. Analytics don’t work.
1:16:15.3 Thom Hammerschmidt: Analytics? Oh my god, What the fuck does that even mean?
1:16:21.6 MH: Blah-blah-blah. A long, long time ago.
1:16:29.6 MK: That was amazing. Oh my god. That was amazing.
1:16:40.3 MH: I think that was not good at all, though.
1:16:49.3 MK: You guys have reached your creative peak.
1:16:55.3 MH: Yes, you know what? I think that is an accurate statement.
1:16:58.8 TW: That’s fair sort of an end of our rope type situation. It’s sort of like “Well, what else are we gonna do?”
1:17:05.6 MH: Do you want me to drop it down?
1:17:12.8 TW: Well, yeah.
1:17:13.3 MK: What’s drop it down mean? You guys with this whole music lingo.
1:17:17.6 MH: Change the key, Moe. The artists are working here. Okay, so…
1:17:24.3 MK: My old piano chords are not useful here.
1:17:33.1 TW: I’ll let you go ’cause I’ll just, another stem winder, gonna be longer than American Pie.
1:17:41.8 TW: Rock, flag and Dashboards.
This site uses Akismet to reduce spam. Learn how your comment data is processed.