Build a Customer Dashboard in Tableau

Hello everyone and welcome to today's session.

My name is Rhys and I'll be your moderator for today. We are going to get started very shortly in a couple of minutes. We're just waiting so everyone has a chance to join. If this is your first time at a DataCamp live session, welcome. We are going to be starting very shortly. We've got a few bits of accompanying info for you if you want to code along with us live today. So I'm going to be sharing those very shortly. We've got a dataset and we've got some setup instructions for you. I'm however, they are, uh, simple enough. So yeah, we'll share those via link very shortly.

You're—

we are going to be working with Tableau Public today, so if you don't have that, uh, set up already, then, uh, I'd recommend getting that set up now. Um, I'll be back to repeat these messages very shortly, but if you haven't done so already, make sure you register for this session by scanning the QR code on screen. You can head over to datacamp.com/webinars, and I'll also post a link, uh, as well to that in the chat very shortly as well. Um, that way if you register and you need to jump out, we can send you the recording as well as all of the accompanying info as well. We are also gonna be answering your questions for the last 10 minutes of the session, so make sure you stick around for that too. Hello everyone and welcome to today's session. My name is Rhys and I'll be your moderator today. We are just about to get started. We're just waiting for the last few people to join. If you want to code along with us live today, I've just sent a link in the chat to the resources document. That's going to contain a link to where you can get the dataset and the setup info for what going to cover today, as well as a few other useful links as well. So please do check that out if you want to join in live today. However, if you just want to watch along and catch up with the recording, make sure that you register. Uh, you can do so by scanning the QR code that's on screen now, or you can head over to datacamp.com/webinars. And I'll also put the link in the chat for you now as well. But hopefully everyone who's here is already registered, so hopefully That's a little bit of a useless message from me. Regardless, welcome and thank you for joining today. We're going to be getting started in about 30 seconds. I will just run through the housekeeping once more before we get going. Please do register if you haven't done so already. If you want to code along with us live, please do check out the resources document that I've sent in the chat. That's got the link to the dataset and the setup info that we're going to be using, as well as a few extra bits in there as well. And we are going to be answering your questions today. So if you have any questions or notes at any point during the session today, let us know and we will get back to you during the last 10 minutes of the session. So yeah, make sure you stick around for that.

Brilliant.

I think that's everything from me. So now I'll hand you over to your host for today's session, Richie. Richie, please take it away.

Hi there, Data Scamps and Data Champs. Welcome back. This is Richie and I'm very keen to be Yeah, getting on with some Tableau today. So if you were here on Wednesday, we had a really nice session learning about dashboard design best practices. So today we're going to put those skills into action to build something. Obviously, you don't need to have been here for Wednesday's session. This is all going to be standalone. But if you want to check out the recording, yeah, we'll send a link for that in the chat. All right. So today we've got a really nice case study on building a dashboard for customer data. Obviously, it's an incredibly important use case. You need to make sure you understand your customers. Data is the best way to do this. Now, I have not one, not two, but three guests for you, and they're all data analysts from the same team at the telecoms provider Optimum. So first up, we have Liz Singh. Hi Liz, welcome. Hi, great to have you here. Next up, we have Abe Korteng. Hi Abe, great to have you here.

Thank you.

And last but not least, we have Rishabh Shah. So, uh, welcome Rishabh. All right, now I feel like, um, oftentimes when I've got guests on, I'm like worried like maybe they have a bad internet connection or something like that. I think today no problems, we got that covered hopefully. Uh, so yeah, uh, no problems on the technical front. I've probably jinxed it there. Anyway, I want to call like the three of you like the Beastie Boys of data analytics, but I was thinking actually, no, Beastie Boys are a duo these days. We need a different band, maybe like the Yeah Yeah Yeahs of data analytics. Wonderful. So anyway, all three are Tableau and reporting experts. I'm going to let you take it away. Liz, I believe you're going first.

Yes. Okay. Oh, welcome everyone. Really glad you're here today. Um, so what we're going to do in the next 45 minutes is walk, walk you through a real-world business intelligence use case start to finish. We're talking about a problem that every major telecom company deals with every single day, which is customer churn. Telecom is one of the most data-rich, operationally complex industries out there. Your customer's experience is shaped by dozens of factors all at once. The quality of their network, their billing experience, whether a technician showed up on time, whether a service issue got resolved, all of that data lives in different systems, different tables, and even different teams. And it all has to come together if you want to actually understand what's happening with your customers. That interconnectedness is what makes telecom a fascinating and challenging space to work in from a data perspective. Which brings us to the churn analysis. Churn is essentially customers canceling their service, and it's not just a metric to monitor, it's a real business problem. And here's the part that most companies don't really talk about is they're usually more focused on acquiring new customers that they miss what's happening on the other end. If you're bringing in an X amount of customer a month and then also losing that same amount within the month, you haven't grown at all. You're actually just running in the same place. And in business intelligence, we want to look past those surface numbers because churn is not random. And what we can do is go into the data, look at service history, billing patterns, how long a customer has been with the company, and start to identify whether there are patterns that explain why customers are leaving and probably even when too. And that's exactly the business case we're bringing to life today. We put together a simulated telecom data environment, synthetic data structured to mirror what a real telecom data warehouse looks like to walk you through this end to end. So I'm going to pass it over to my coworker Rashad to take you through the data schema and show you how it all connects.

Hey. Hello, everyone.

Can you hear me?

So just to switch, as Liz mentioned, like the customer data or the data in the telecom industry can be quite complex. Whatever you are seeing here is just a kind of a small mock dataset which we created for the use case of, for the use case of this specific exercise and to showcase how complicated that data can be. So let's just start with a very Basic layer we have currently at the bottom. We have network layer. Network layer is a store where store, we store all our circuit information and all the location where we have installed the circuits. And there can be multiple more datasets on this, but just for our simplicity, we're just considering these two tables for today. On the top of network layer, we have a kind of a sales layer which allows us a user to sell that services, which allows our sales rep to sell that services from network providers and we can provide the customer data, what services they're enrolled into, what type of products they are receiving, and what type of billing. And from the sales layer, we go also— it's all the sales stuff is connected to billing addresses or billing and orders so that we have different type of logs like what type of subscriptions they have, what type of service order they were subscribed to. The main difference between sales and the billing layer is like the sales might have an X amount of data, but billing orders always will have less than X amount of data. In a real world, there might be multiple sales layers showing like there is multiple points of sales. In this specific case, we are just trying to simulate stuff, so we only have a single sales layer. Moving forward, as a data analyst, we do most of the stuff, I would say, at the end, at the pre-production of the free transformation of the data. So we do have some analytics table which can help us to identify from the logs of disconnect and installation service order, like how we can analyze and make our Tableau more powerful. Moving to a more deeper dive, like how these tables are connected and what are the views you can see in this table. So this is a kind of a detail of like what we have used to connect this table, what is a primary key and a foreign key in this table. And this will be shared after the session, I would say. And if you can see, like, this is an interconnectivity, like, from— there is a table where we get a geographical context, there's a table where we get an installation status, there's a table which maintains who, uh, where we, uh, handle the sales, what is the loyalty of a customer, or how long that customer has been a part of us, all those type of different stuff. With the help of this data, we can also enrich and see how our disconnects are happening, which are our market at risk, what are the market we should go into. All type of analysis can be done. For this purpose, what we have done here is like we have created a kind of an executive analysis, analysis dashboard for you. So like, how important is that you understand data and create the analysis and then improve the visualization. That's a 3-step process. So this is something after giving some thought and time, what we can do is like we can create a detailed analysis of this. If you see there's some kind of an executive analysis where you see what's active customer count, what are disconnects specifically to focus on a churn type of data. We created this dataset, this QuidWeb, where we are losing more customers than we have active customers. So if you see, we will have a high negative revenue and all the stuff. What is the tenure for that SKU? We have set up twice like 2 to 3 years. Something, some can be more than that. Now if you see what is a disconnect trend at the base of the dataset, you won't have the reason categories. You need to identify how disconnect reasons can be categorized together. Are they because of some kind of controllable impacts? Are they because of some non-controllable factors? If there's something which can be controlled, like something called final pricing or something, or if it's something, uh, if some competitor is trying to poach our customers, then we should focus and create strategy around that. So what we are showing is like, uh, disconnected reasoning based on what can be the forecast disconnect based on the prior data. And what should be the steps for the strategy can be decided further. And if you deep down, this is a proper grouping which can be said like what's the reasoning group for the disconnects. This shows the count. The another point of view is also to analyze like where are we losing our customers? Are we losing for some specific apps? If yes, then why? What's the reason? While talking about the just losing of customer, we don't want to see the number of losses. We want to see number of losses in compared to what they have, a number of accounts they added. So if you see, if anyone says, sees this data and say, okay, this person has a 54% disconnect rate, but I would say this person is performing way better compared to this person having just 52% disconnect rate because the— this person is handling 345 active accounts compared to this person handling somewhere around 245 active accounts. So this all become— becomes a kind of a perspective what you want to show. And this is also one kind of a high-level analysis. You can see like for each state, where are we losing like money, from which state we are losing money, or who is a primary, I would say, point of, point of contact where we are losing money. So this type of analysis can be done with this type of dashboards. There's even more you can do them into this, this sample dataset. Like you can even create geolocation maps, you can create multiple more things. But for this code along, I will pass to Abe who will show how you can set up and connect this data into, to show some nice visuals.

Thank you, Rishabh. Thank you, Liz. Welcome, everybody. So I'm going to share my screen really quickly. Let me just move this away and then share my screen, entire screen. Please let me know when you can see the screen. We should be good. All right. So like Liz and Rishabh said, let's set up the case. Rishabh gave us a roadmap of as to how we put this together. What I'm gonna do now is to give you a detailed view as to how we can use Tableau for the analysis. So I'm gonna just give you, I'm gonna, so this is, I'm just gonna set up a new dataset. I think you have the dataset. You also have the instructions. So what you can do is follow along if you can. Otherwise, after this, you can, you'll be able to go back and review it. Now, what we're going to do now is to bring the data into Tableau, connect them, set up relationships, create calculations, create the sheets, and then put them together in a dashboard, put everything together in a dashboard. But we're not going to be able to do all of it for obvious reasons in the interest of time. So I'm just going to run through it very quickly and just share this with us. So the first thing we're going to do with this is to connect the data. And then we're gonna select the particular file where we wanna bring in, which is this one that we, we put together. Once this load, what we are gonna do now is to obviously wait for it to load. Now we're gonna bring the tables in. What we are gonna do is we're gonna bring the disconnect. This is gonna be our primary table. It's going to be Disconnect Log, bring it into the table pane. Once we have this in, we're going to now connect the tables in and then we're going to bring the Disconnect enrichment. I could just simply type it in here to bring it as two is a table we want. It's going to be Disconnect enrichment. We're going to put this right on top of this and then set up the relationship. Now if you're lucky, The connection is already established. We don't have to do anything, but this is the connection already established here. We have the disconnect ID matched to the disconnect ID. The next we're going to do, we're going to do the geography or the disconnect reason, I should say. This is going to be this table right here. We're going to connect that to the enrichment. Again, we have a match reason code to reason code. That is good. The next one is going to be the geography, which is this one. We're going to connect that to the enrichment table as well. And then we have geocode matched to geocode. So far so good. We are tracking in the right direction. And then we're going to want to connect the users because we want to know which users, as Rishabh said, are affecting the bottom line. So we bring this in here. And now we need to just connect this. For this one, we're going to connect the user ID, which is the user handle by ID for this table. And then we're going to connect the user ID. We have a match there. So the next step or the next last step we're going to bring in is the account because we want to know which account. We're going to tie this one to the disconnect log. So we want to know which accounts are active, which accounts have disconnected. And then with that one, we're gonna connect the account number, which is gonna be this guy right here, and the account number, which is gonna be this guy right here. We have it. And now we're gonna now go ahead and build the calculations. Now, if you're following along, this should look similar to what you have on your, on your, the worksheet. So we have everything connected. We're gonna go now build calculations. Now what I like to do is to, if you're looking along, you're gonna see that some sections are marked in green. So we're gonna, we have part 3 now. We have part 2, 1 and 2 done. We're gonna do part 3. So I'm just gonna set up the calculation, copy this, click on calculate field, put that in there. Make sure we don't have any space at the end of it and then bring the calculations in. All right, we have no errors, so we're going to chug along. We hit OK. The next calculated field we're going to create is the disconnected account. Again, we're not doing every single one of them. This is just a show and tell, so we're just going to go through it. So once we bring that in, again, make sure that we don't have any space at the end. Bring the formula or the calculation in there. All right, so this is showing us an error. We want to figure out what the error is. It said bad character. So let's see what's happening. Uh, it doesn't like this one, so I'm going to remove that one off the All right, all right. So we have these two set up. We're gonna do the next three, which is gonna be the active customer count. We wanna know how many active customers we have. So we put that in here. We bring the calculation. All right, everything looks good with that. Also make sure there's no space at the end. We are good with that one. Now we're going to know— now that we have the cust_active account, we want to know how many accounts we have that are disconnected. So we do that as well. We count the extinct, the ID, we have that calculation done. Now we want to know the disconnect ratio because we want to know how many customers we have versus the disconnected. So we configure the ratio between the active account versus the disconnected account. These are all insights that we think is necessary to have so we can run a better business. So we put that in there, make sure everything is good. We are good with that as well. And then, like, let's say we want to risk the revenue loss because we want to know of all the customers that disconnected, what revenue we lost if we had retained them. This is money we could have been making, but because they disconnected, this is money we have lost. So we're going to look, figure out what the error is. Says reference. Okay, so that means one of these fields. Let me just fix that. Let's see what it says. Aha. All right, this is very good. I'm glad we have this error. What this means is that this is coming up as a string, but we want it to be Not a string, so we're going to look that up. We're going to bring it up. We're just going to change that, the data type. Because it's revenue, I figure we're going to do decimal. So we're good. We fixed that. Okay. So we have— now we have all the calculations that we want to— for this exercise, we want to put in. The next step now is to build the sheets because obviously we have to have the sheets. To create the dashboard. So now that we have these calculations in, this is what we want to focus on for this exercise. It's a scaled-down version of what Richab showed. Richab gave us a holistic picture, but for the sake of this exercise, we're just doing a scaled-down version of it. So we have in here, then what we want to do is now bring in the disconnect event. So what I want to do, just Minimize this so I could see the disconnect events. I'm going to put that on the text. All right, so we have 2, 400 and 4,500. What I'm going to do now is to just name these sheets Disconnect Event for reference so I know which to call when I'm building the dashboard. Now we're going to do the disconnect ratio as well. Put that in here. We're going to bring the disconnect ratio, put that also on the text. We have in zero here, so let's fix that. I think that has to do with— let me just check into this one real quickly. Disconnect. All right, we'll come back to that. Let me just build everything and we're going to come back to that and fix that. So we have the KPI Lost. With that one, we're going to drag the revenue. We— I think we're going to have the same thing here, Revenue Lost. All right, we're going to format this to currency. I'll just do standard for now. And then we want to know the active customers. So we just bring in the active customers. These are our sheets that we need to create so we can set up the dashboard. The active customers, active customer count, okay, that's 1,700. So now let's go look to see why the disconnect ratio, which is this guy right here, is showing zero. So we can look at the formula again. It looks good here. Let's go to this one. Active account. Okay, I think I need to fix something with the Disconnected account. Let me just bring this one back. Disconnected account. Yes, this might be it. In the interest of time, I'm just going to gloss over this and then we're going to set up the dashboard so we don't lose time. But once we fix this, everything should fall into place. Now with this built, what we're going to do now to see how we're going to put it together in a dashboard. We have this pre-built already. With every calculation in the mix. So we're just going to bring it in. So I'm just going to minimize this right here. And what I like to do oftentimes when I build that is I want to set up templates. So I have a template. It makes it easier for me to just drag and drop, drag and drop things into them and format it the way that I want. So we're just going to put this here. I'm just going to relabel this to churn because that's what we are trying to do. Let me just rename this churn. Oops. And then we're going to just rename this, say churn analysis. Analysis. This should be good. No, I can't even spell. Okay, I think we should be good with that one. All right, so with this, these are the insights that we want. I think Richard did show us we have this connect event. We're going to put the— I could put the disconnect event right over here, then get rid of this one. Then we're going to put the disconnect ratio. Let's fix this by making sure this is entire width. Fix that entire width. Then what else do we need? We need the lost revenue. We get rid of this. We fix this formatting. And then we put one more. What should we put? What should we put? All right, let's put the active account. Now you obviously can rearrange this however you want, but we're doing it this way for exercise purposes. So this dashboard gives us a view where we know how many account holders connected, what the disconnect ratio is, what the revenue loss is, and active customers. Now, you might be interested to know why do we have active account 1,700, but then disconnect account 4,500 and the ratio. What we did on purpose just to bring visibility and insight into this is to figure out if the disconnect ratio is at 73% because We're doing the math based on the active versus disconnected. We want to know, in essence, we have more customers that are disconnected than we have active customers. So like Liz said, we don't want to bring more people in or more customers in while we're losing a whole lot more. That is not a business model we want to follow. So we want to make sure at any given point, whatever we do, we have more customers active than those that are disconnected. So now that we have that high-level overview, what we wanna do now, bring some graphics into this. What we are gonna do, I'm gonna put the reason breakdown, the reason why customers are disconnecting, then get rid of that. So this is the breakdown of the disconnect reasons. We also wanna know, I'm gonna put the sales disconnect reason. Reason for this is we are trying to look to see of all the sales, of all the disconnect, which sales agent is predominant or is an outlier. This way we can trace back and see, is it something they are doing that is in essence the reason why people are disconnecting? Are they not giving people the right information? Are they being overzealous and just trying to game the system and try to do things to get customers in and not really follow through or whatever the case is. Because I've seen situations where a sales agent will promise heaven and earth to the customer, they get the service and things don't turn out to be the way they were promised. So we want to trace back and look to see if that was happening, especially if we have a particular agent or agents that are outliers. So we can look to see where the gaps are. And mitigate them. Again, these are all design ideas or design principles that you can follow to try to figure out. You're not just looking at data, but looking at situations of what's going to give you the better insight to identify the gaps and mitigate them. Because at the end, what we really want to do is look to see what we're doing well and capitalize on it, and what we're doing— we're not doing so well mitigate that. So with this, we also gonna probably wanna bring— let me just put this in here. We wanna do— we have the sales agent. We're gonna put the retention agent also here. I'm gonna put this right here. Get rid of this. Again, this is also an insight that's given us an idea to see which agents have an ally in disconnect. Customer called to disconnect. Are we doing everything we can to retain them? Are we giving them incentive to retain them? What are we doing? Why is a particular agent agents have been an outlier in disconnect. So we can again figure out what the issue is and mitigate that. So the next graph that I want to show over here, I suppose we can put— yeah, I think Richard did show this as well, the region disconnect. Get rid of this and then do that. All right, so we have this. We're going to see that we have these little things poking out here. We're going to go to Layout, identify what that is. I think it's one of these. It's not that. Let me just go back down here. Okay. All right. Well, this helps you in the interest of time. I'm just gonna skip through it. What this helps you do is identify which particular element I wanna say, or object you wanna get rid of. Just click through it and identify it. And remove that. So what I'm going to do here is to pause on this one and see if we have any questions, insights, analysis, so we can talk and be able to get a better, you know, to be able to get a better insight as to how we put this together. So I will, I guess, Richie, I guess we will just circle back and see if folks have any questions or any further insight that we want to get provided.

Thanks a lot. So no questions from the audience yet, but for anyone in the audience, if you do have questions, then please do ask them now. In the meantime, that was a very cool dashboard. So I have a few questions for you while the audience is thinking.

Yes, sir.

So in terms of the target audience for this kind of dashboard, Uh, so a lot of this is, um, it's a fairly technical dashboard. How would you adapt the dashboard depending on like whether it's another analyst looking at it or whether it's, for example, like, um, a, a customer lifecycle retention manager looking at it? Um, so if you've got someone from the business side of things, how would you go about tweaking the dashboard for different audiences?

Very well. So you're talking about the visuals, how to change the visual to suit a particular need?

Yeah, or I mean even like the whole thing. Would you change the layout? Would you just change like the, the format of individual plots? Like how would you approach customizing for different audiences?

Very good question. So I think me personally, when I design that, but I look at the audience, especially when I'm doing the dashboard for executive, that's because they just, they don't want granular detail, they just want high-level detail. So I guess to answer your question is all about figuring out what the audience type is. Even color scheme matters, right? And layout matters. So I try to put that into perspective, take that into consideration to look to see what the audience is. So to answer your question, if I want to, let's do this. I could make a duplicate of this. And then go here, delete everything, delete all of that, delete this, and put this right in the middle. So this way what happens now is this, I have everything put in one gigantic view because oftentimes, sometimes this is all the individual wants to see. Now depending on the audience also, you probably wanna set up a filter. You put up a filter here that can just go through each one of these and be able to zero in on what they want to see. So I don't know if that answers your question or not, but I think, think about the audience first, what they're looking to do, how they want to have it laid out and build it out. But that's, that suits your needs. Sometimes if you're doing for multiple people, for example, multiple departments, for example, you want to build in a filter that allows each department leader to select their folks. To look at.

Okay, so in general, you want to, I guess, display only the information that people need to see in order to make the decision, and you don't want any extra information that's going to distract them from that, correct?

Yes, yes, in that sense, exactly. That's what I'm saying. Yes.

Okay, nice. Uh, and you mentioned color scheme. Uh, I mean, since we've got some time, uh, do you want to talk us through how you go about like changing the color scheme for, uh, like a particular order? It's like, like I guess first of all, like, what colors do you pick? Are you going corporate color scheme, or is there another principle behind how you choose the colors?

Very good question. So because we did this, you know, I just used the obvious default. So when we do it on our side, we usually like to mirror our corporate colors, you know, this is, that's, you know, so to answer your question, yes, I pick colors that match the corporate theme. If I'm building it for my industry or my company, that's what we go for. But in a nutshell, I think that's what I'm talking about color schemes. Sometimes color scheme can be very distracting. So if you don't pick the right one, it can be so that instead of folks identifying what the problem is with the data, they're talking color scheme.

Okay. All right.

I mean, if you want to demonstrate quickly, just like how you change the colors, that'd be quite useful.

Okay, so we're gonna go here. So let me just go into— go to this particular sheet. We pick this. Let me just bring this. Okay. Which one is— okay, disconnect. Okay, let me just move this out of here. We're looking at a change of these colors, you mean, right?

Yeah.

Okay. All right, so let me just move this a bit. Move this out of the way. I'm gonna make a copy of this, uh, duplicate. Remove. Something like this works, or?

Okay, yeah, so I guess the idea is that in that marks thing, there's like the, I guess, the color thing. You map that to a different variable. Is that the main process?

Okay, lucky. I think I know what you're talking about. I think I misunderstood what you were asking for. So let me just go back to which one is this one reason I think is this one. So we go here.

Yeah, I can't remember exactly how it works. I know there's a color option there and oh, well, just move it along.

Okay.

So somewhere in Tableau there is a way of changing the color scheme. All right, maybe we move on while he's thinking of that. So I guess one of the big points of this is around trying to figure out how do you prevent your customers from churning?

Yes, essentially, yes.

Do you have a sense of like what the most important factors in that tend to be? Like, because I know you've all worked on this, maybe you can all talk us through this. Like, what tend to be the most important features there in that model of like why a customer churns or not?

All right, so we have, we have this built where we, you know, it's like, so basically from my experience, it comes down more with, I think Rishabh touched base on it, service issues, and Liz also touched on the billing and, and bad experiences. So One of the models I feel, you know, we kind of put together is customer loyalty, where we can map out which customers are likely to disconnect because they have service issues. So we map out customers with service issues, customers with technician issues, customers with billing issues, and customers with OSAT, for example, NPS. Net Promoter Score. If we can identify a particular area that has a high churn rate, especially they have competition there, we have different service providers there, there's a very good chance these customers will churn if things are not going well. So the identification is trying to figure out which customers fall into these categories and what we can do with them. Do we Provide them incentives like promotions? Do we make sure that when they have service issues, we respond to them? Outages, for example, is a big part of this too, where a particular area that experiences a lot of outages, especially with the competition in the area, these are all catalysts or things or variables that contribute to a customer having a higher desire to disconnect the services. So lots of things, but I think the most important that I want to touch on is outages, service issues, which is obviously connected to outages, and the experience that they have when they call into call center.

Okay, uh, that makes a lot of sense that, uh, if you're having service outages, then customers are more likely to churn. I guess that's true in many industries. If your product doesn't work, then customers are going to not continue to buy your services. I like the way that this is going to show up in the data. All right, so getting back to Tableau, I guess, I'm curious as to whether there are any like Tableau features you think everyone should be aware of. Like, are there any things that, like, what's your favorite feature that you want to promote for people to use? Rishabh, do you want to take this one? Sure.

So specifically for Tableau, like, real-world examples, we couldn't show that in the this specific demo, like I would say like anything to do with tabPy is one of the good, nice, nice to have things on Tableau. Like as you can see the data, we, what we have created at a large scale, we have some kind of data quality issues. We need to handle them. So we can have our custom Tableau, custom scripts for Tableau in Python and then leverage both of things together for our analysis. There are some other tools like if you don't want to do ad hoc on Tableau, we can always pre-cleanse our data at the source and then load. But that sometimes we miss to do the pre-cleansing and it's like too much a hassle to go back into your data storage platform and do the cleansing and reload all the data sources. It's better to like if it's just one-off thing, we can always connect to a tab by server and then run our custom scripts on that. Also, you can do custom modeling if you want, like some kind of statistical modeling or something like that.

No, Tableau is very cool. I do love the idea that like if you want to do some like complicated data cleaning, or I guess if you want to like do some machine learning in there, like you don't want to do that necessarily in Tableau itself, you just want to get all that crunched in Python and then leave Tableau for focusing on the visuals and creating the dashboard. Okay, Liz, how about you? What's your favorite Tableau feature?

Honestly, more recently I went to a demo where it showed that you can connect like Salesforce to Tableau and like all the data gets to kind of upload there and it regenerates itself there. So I would say that that's probably one of my more favorite newer features.

Okay. Yeah. And of course Tableau is owned by Salesforce, so it's about time they had some proper good integration between the two systems going on. So yeah, certainly if you're building dashboards based on sales data or the customer data that's directly in Salesforce. You don't have to export it to a database and then re-import it into Tableau.

Okay.

All right. Actually, are there any AI features in Tableau these days? I don't know whether they've been introducing those features yet. No idea. Okay.

I'm not sure about that. No, I don't really think so, but maybe Rashad would know, honestly.

I know they were hyping it for a long time. I don't know what got launched.

We can't hear you, Rashabh.

Can you hear me now?

There we can. Yeah.

So for sure there are some kind of Tableau agents and all those things, but I would say we don't use that much in our day-to-day as a day-to-day analyst kind of stuff because sometimes what happens, like, you have a so large schema while you're working with that, it's a little difficult to explain all the stuff to agents. You need some kind of a, I would say middleman to explain that stuff. And it's always faster to write that script on writing manually compared to going to agent and ask him to write. So yes, there are some innovation which Tableau is doing in the agent side, like all those stuff, but still, it still will take some time to ingest into day-to-day life.

Okay, so it's a sort of coming soon, probably useful in the near future thing, but for now it's like just sort of watch it and see.

A nice-to-have feature.

Okay, all right. And I'm just curious, like, since you're building a lot of these dashboards, I guess, like, who do you— who is your sort of main audience for these? Like, how— like, who do you build most of the dashboards for then?

So that I would say that can be categorized as two type of audience. Once you get where the audience one can be like they want just a summarized view of everything what's happening. There might be a second audience which we need a lot more time to develop those type of dashboards, like how everything dynamic. So they want to view what is their performance, their sales and everything, like a kind of a personal dashboard sheets. So there is like 2 ways of dashboard we create, like kind of an executive dashboard and a day-to-day dashboards, and both of them have their, like, their own approach of creating those type of dashboards.

Yeah, kind of like to Abe's point, like sometimes, like, it would be for, you know, VPs who have no interest in, like, looking at things that are more granular and they want to see everything that's very high level. So then that would be a different type of view, and then if we're you know, have a different type of ask, and it's like for a sales leader, for example, then, you know, they want to get down to the nitty-gritty of everything.

But predominantly, we build it for leaders, leadership, right? Like, they— I don't want to use the word ground-level leadership, but immediate leadership that look at agents or look at reps, you know, predominantly, that's what we go for. But like, it's like, and depending on who the audience is, you build it If it's a person like in the C-suite, they don't want the clutter. They just want the meat and potatoes. Yeah.

Okay. Very, very simple stuff. So actually, that leads me to one of the big problems with analytics is you build all these dashboards and then people use them for a week and then they abandon them and you feel like you've wasted your time. So I'm curious, how do you make sure that you're building dashboards that will get used? Do you have any techniques for ensuring you're building the right thing?

Me personally is having constant update modification, sometimes changing the schema around, sometimes moving this around, especially if it's something ongoing, then people are going to have to come back to it and visit because it's ongoing, they have to keep track of it. But if it's something like a one-off where it is more or less, very high level, and especially when you do KPI dashboard, for example, those are pretty much consumable because leaders have to constantly go and look at it to track performance and coach it as necessary. But when it's more like the executive style dashboard, people just go in once a month to look to see what the what the trend lines are. So I tend to sometimes once in a while, you know, change the layout just because I want people coming, that something is different. They see movement, right? They see something has changed, so they want to come back. If they want to go and they look the same all the time, it feels like nothing has changed, so they stop coming back.

Okay. Yeah. So I can certainly see how there's a big difference between something where it's a very operational dashboard. It's like, well, you want to know is something breaking right now in real time versus it's just something about, well, what's changed this last month, this last quarter or something, you're only going to expect occasional views. All right, we've got a question from the audience. So The Enthusiast asks, how is churn being defined in the state set? Is it with inactivity, downgrade, etc.? I don't know. Yeah, do you want to talk through this? Yes, maybe explain differently, like net retention rate, gross retention rate, all these sorts of things.

Yeah, these are total disconnects, right? So yeah, I'm glad you brought it up because a customer could have multiple services and they disconnect one, but this is account disconnect, flat-out disconnect.

Okay, um, and do you have to build like, like how do you take account of different metrics then, uh, or endpoints in this? Like what Would you ever— would you want different dashboards for different types of churn, or would you try and build them into the same dashboard? How do you divide up what you're going to do within a single dashboard?

Very good question. So we can bring it all in one, one view and then have one dashboard that just gives us the entire disconnect. So we know the customer disconnect, and then also look at another dashboard that explains downgrades. So we can see what product people are downgrading or removing from their services and trace it back. So I think it's going to be important to have a high level in the grand scheme of things, how many customers we have that are disconnected. But while we're worried about that, we also want to worry about within the internal of those active customers, which one of them is removing services so we can try to figure out why that is. Is it a particular product? That has been downgraded, why we're not doing well with our product, are we failing on our promises with our product, so we can look to see what the insights are and how to mitigate them. So I think a lot of times people are focusing just on the disconnect, but then they fail to realize that within the active customers, people are disconnecting the particular services. We want to track that and figure out why that is and how to bring it back full circle.

Okay, so there's often a disconnect between I've built a dashboard and then I'm going to take some action to solve a business problem. And it sounds like you want some level of diagnosis in between that. But talk us through how do you go from like that? I've got a churn dashboard, I'm going to have an impact on those churn metrics.

Very good question. I think, you know, When we bring everything in, remember we were talking about connecting it to the outages people have, connecting it to NPS, how the customer is rating the service that we provided them via the customer interaction with an agent. If a customer has a tech visit them and we— there's something that I think most people forget is when you do a dashboard and you look at the surface-level details as to this person had disconnected, you know what's happening, you don't know why. The why often comes via looking at the pattern the customer has had. So we have a model that when a customer calls in, we track the interaction. We can review that to look to see these customers that have disconnected, here are all the interactions they've had with our agents. What is the common denominator? How are our agents responding to them? How are they responding back to us? Figure out the common denominator. I believe we cannot be 100% at anything, But if we could come close to it as possible, we can mitigate it. Hence the reason why we figure out— and one more thing that I think most people miss is when we're doing well, we just assume we're doing well, so we just keep doing it. But we don't know why we're doing well. We got to figure out why we're doing well, the common denominator, and amplify that. Or exploit, exploit that, you know, sorry to use the word exploit, but manage that and make it better. If we're doing bad based on the common denominator, we can mitigate it as well.

Yeah. And I think like to Abe's point, like looking at the patterns is the most important part. Like we're never going to get everything 100% right. And like, you know, completely eliminate churn, you know, like people are always going to leave for whatever reasons. And some reasons we may not even be able to track, but it can fall into a bunch of different buckets. Like it could be a geographical issue or to Abe's point, an outage, or it could be, you know, a sales rep in a particular space or just a specific sales rep where we would have to go in, maybe do extra training or or have a conversation. And it's not even sometimes about pointing fingers. It's just having a conversation and seeing maybe what we're doing wrong and how we can fix it.

Absolutely.

Yeah.

Sometimes you get a bad salesperson. I used to work in a call center and it's like, okay, yeah, there are some people who are useless, but also sometimes just like, I just had 3 customers in a row who's like yelling at me for like half an hour.

Okay.

Yeah, it's, it's tricky, like trying to pin down exactly what's going on. But I guess the data ought to be able to tease out some of these things.

All right. And that's why we don't focus on the one-offs. We focus on like a, like a 4 or 3, 4 or 5 for the entire year. You know, you map it out that way so the one-off doesn't become noise that you chase.

Absolutely. Okay. We've got another question from the audience. Runa is saying, as an analyst, Should dashboards be standalone or do you want a report to accompany them? Good question.

I would say like for those type of stuff, like two type of dashboards. One is executive dashboards where it's better to have some kind of report or summary, but where there's a dashboard which is for like day-to-day monitoring and tracking, that type of dashboards won't require any kind of summary or reports. That's kind of a day-to-day user dashboard, but that dashboard needs to be modular and we need to create in such a way that can be used in different devices like either in iPad, phones, or laptops. But whereas an executive summary, you need a proper snapshot and a dedicated summary which goes with that explaining what are the positives or what are the negatives of that specific things and report on those, I would say, KPIs.

Yes, tricky one. Like, I have to say, I've seen quite a lot of reports where it's like, oh, there's like a screenshot of a dashboard in there. So it's like, well, you know, you, you kind of get the data from one place to another. But yeah, if you need a narrative, then you got to have a report because dashboard's more like designed to be explored or interpreted by a human. So yeah, if you're sharing a story with lots of people about what's going on, then you probably want report format, right?

So the dashboard is more or less what's going on. The report is why it's happening.

Oh, I do like that distinction. That's good stuff. All right, we are coming up to time now. Before everyone dashes off, I'm going to say next week we've got a couple more sessions on a similar theme. So on Tuesday, we have a session on how to create a marketing funnel using Microsoft Excel and Copilot together. On Wednesday, we've got a session on data visualization for data storytelling. So we've got Brent Dykes, who's like a very big name in data storytelling, coming along for that. So please do come back for those sessions. Then on April 1st, we've got the Radar Virtual Conference. It's one of our biggest events of the year. Please do sign up for that if you've not signed up already. All right, with that, thank you, Abe. Thank you, Liz. Thank you, Rishabh, even though you're invisible at the moment. But yeah, That was a really great session, very much enjoyed that. All right, thank you to everyone in the audience who asked a question, thank you to everyone who showed up today. See you all next time.