Fi (00:09)
Welcome to Undubbed, the podcast that's unscripted, uncensored, and undeniably data. I'm Fi
Sarah (00:15)
And I'm Sarah. Now, Tableau Conference 2026 has wrapped, and there's so much coming for Tableau across cloud, server, and next.
today it's all about the good stuff on the horizon and what it means for the people who get to decide what their teams do next. Data leaders and decision makers.
Fi (00:39)
And we've got two of our favorite people to get excited about it with. Kirk and Candi are co-founders of Paint with Data, a Tableau consultancy and Salesforce partner in Canada. Kirk has 25 years in business analytics across Cognos. Whew, that's a name that's a blast from the past. IBM, Kinaxis and Salesforce plus Tableau. This year he's reached
the pinnacle of tableau recognition as a tableau visionary. He's also a Tableau ambassador, a Salesforce certified data cloud consultant, and he even wrote a book on data modeling and Tableau.
Sarah (01:18)
And Candice at Visual Analytics specialist with a stack of Tableau and Salesforce certifications. She heads up the Canada Tableau User Group and she's our fourth-time Tableau User Group Ambassador. Her real love is helping clients bring their data stories to life.
But before we dive in, please like and subscribe to UnDUBBED and most importantly, share this episode with a data leader who wanna know what's coming.
Fi (01:47)
Candi and Kirk, welcome to Undubbed. We're so glad to have you both here.
Candice Munroe (01:52)
Thanks for having us.
Kirk (01:54)
so pleased to be here,
Sarah (01:55)
Candi, there'll be plenty of people listening who haven't come across you yet. So tell us who you are, what you do, what you love doing in the data world.
Candice Munroe (02:05)
I didn't come to data analytics or Tableau naturally, like I was a pharmacist first. And I decided at a certain point that I liked the data more than I liked pharmacy, so I switched over. And Tableau as a product made that really easy for somebody who doesn't have a computer science degree.
But I do love I've always loved math and I've always loved analysis. so I was applying those in my other roles and you know, so just picked up β Tableau and and went with it.
Fi (02:38)
Hi Kirk, welcome back to Undubbed. For those people who haven't heard your story, we'd love to hear it too. Who are you? And what's kept you hooked on data all these years?
Kirk (02:48)
so like you mentioned I've been in data more or less for 25 years. And I was actually Candi started Paint with data, so I happy to join her. It's nice to actually you know, help clients day in and day out. And I guess I've just been a weirdly objective person for most of my life. So what motivates me, is when people say stuff and I'm like, that doesn't sound right. Like I have to go find out whether it's actually true or not.
It's just the way I'm kinda wired. So it just kinda naturally it's a really easy place for me to be in because I would be trying to find the answers to all these questions anyway.
Fi (03:20)
Yeah, awesome. I think that's a a thing with data people being naturally curious and really wanting to get to the bottom of things as well. I feel like I firmly sit in that camp.
Candi, we're going to switch now into question time Of everything that was announced at TC26, what's the one thing that you're most excited about?
Candice Munroe (03:45)
everything that they announced to Devs on stage was supposed to come out within a year. So that's really positive. there's definitely a different tone to the conference. there's definitely some good things, little fixes, but I think
Tableau's core has always been is that kind of analysis and the kind of curiosity and the flexibility to kind of do that. And I really feel that turning around now from the it's like all the little features that people keep asking for, they're still there and they're still adding those. But I feel that Tableau is getting back to the analysis, being the ones that uncover the
really deep inside and and so I found Matthew Miller's presentation very interesting that way, which is not a fully baked concept or it coming anytime soon, but I thought the tone of it was a good was a really good sign that they were thinking that way again, So that was very exciting to me.
Fi (04:40)
Yeah, so the what thing that you're referring to, I think, was called Tableau Solve, from your memory, like what were the things within Solve that excited you so much about the changes that will potentially come through?
Candice Munroe (04:57)
leveraging the current state of affairs and being able to get the answer, get the insight. So we've always used Tableau for finding the better insight or using analysis and that that quick like drag and drop to like quickly get to that answer. And that's part of what we like to say,
for our clients is we get to the offending record in, you know, four clicks or ten seconds or so like to try and build that sort of triage for them so that they can take like a a dashboard and dig into it and find the problem quickly and easily. And I think that that's what really spoke to me.
Sarah (05:40)
I really liked that part of Solve where felt like it had productionised the ability to forecast a little bit more
Candice Munroe (05:48)
and that's really in demand. you're always forecasting on your old data. So I mean, you know, anything can happen. But yeah.
Sarah (05:57)
Yeah. And then controversial, a little bit of write back as well.
Candice Munroe (06:01)
Which is asked for a lot. Tableau has always tried to be a little bit proper by like not allowing write back
there's an extension that we use for clients all the time that writes back, but yeah.
Fi (06:11)
Yeah.
I've implemented it at two large corporates before as well. it's one of these things that, whether it's tidying up dirty data, we can't go back to source system and edit things. So there needs to be a way to adjust who actually sold something. So that if we're doing sales performance reporting or being able to look at the data and
leverage sort of commentary that's in the moment and stuck then to that that particular record. Kirk, was there anything that stood out for you around Solve?
Kirk (06:45)
that was my big one. It's and and since you guys let a lot on the podcast, I'll tell you a quick story about it. like the night before I was talking to Matthew Miller and he goes, You in particular are going to like something I'm going to show tomorrow. And he wouldn't tell me a lot of it what it was, you know, the appreciation night when we got the hoodies or whatever. And the next morning and it and it's a quick story too, because the people know I've been talking to Francois a bit about golden analytics and stuff. I'm trying to name drop as much as I can here.
I remember talking to Francois in 2018. I'm like, dude, you gotta let go. We'd like write back products, right? Like I'm like, people want to plan, right? Like we like so many of our use cases of clients come to us, it's like, we want you for sales forecasting. I'm like, we can't really do sales forecast. Like, and we people don't want to do like full on budgeting and stuff. There's lots of programs for that, but like
little things like rough head count planning or how many people or could what if we put people in this territory instead of that territory. And the way he did it was pretty slick. You know, it was very tableauly with the like just grab a cell and drag it down and it will automatically fill in the forecast and stuff. So as long as the models are kind of robust there, I was really excited.
so I thought it was the best part. I thought it was way, way, way overdue. It's too bad that Sigma had to come along and have that feature for Tableau to finally wake up.
But there but it's there, if it's coming. Do you know what I mean? I'm excited. I'm equally excited about solve and and and I know Fi and I talked about this offline, but
The quality of life stuff I get really excited by I know other people don't, but I know like Andy Kriebel I continue to name drop, has been doing like map layers now in NLT for a long time. And I'm like, the whole reason I love Tableau is it gives me XY coordinates. I don't want to create my own XY coordinates. So things like, you know, multiple or like
chart layers and set a map like I mean it was so obvious. Like you already generate the XY axes, why do I have XY coordinates? Why do I have to do it? Right? Like I I won't do it at a belligerent, so I'm really excited for it to come.
Fi (08:42)
Yeah, so for people that don't understand what a map layer is or what potentially a chart layer is, can you just walk through your explanation of what that might look like?
Kirk (08:52)
yeah, if anyone's used certainly anything like Photoshop, it like you know, you you do is you drop layers on top of each other, right? Or even in PowerPoint, you can drop like an object on top of and you make it transparent and drop it on top of another one. And I guess to set back for people who don't know what Tableau did from day one, which is a little different, is the reason you do so many funky chart types is Tableau, when you drag things to rows and column shelf, they automatically create
Fi (09:00)
Mm-hmm.
Kirk (09:18)
An XY axis on where to place things. So with maps, all that they are is just x, y coordinates. That's all they've ever been, right? so it's just an XY. And then so what map layers were that came out a long time ago now.
maybe in 2019, is the ability to just layer maps on top of each other, right? So why can't we take other pills and layer those on top of each other like we can with map layers? So that I that's what it looked like is what's coming. So the other cool thing about that is they can every layer could be from a different data source because the data doesn't need to know about each other. They're just getting mapped on top of each other. Right. So sometimes it's really hard to do things like scaffold dates together. But
You could effectively scaffold dates together by dropping different dates and then different measures on top and they're just going to sit on top of each other, which will give you a kind of like a cool time series stuff without needing to do some crazy data modeling in the background.
Candice Munroe (10:14)
we met up with Roxanne and and and like complimented her on the devs on stage and in the data village afterwards. And it wasn't obvious, I guess in the flow of the but it the Viz layers is going to be four different it's like she used four different data sources.
so she's like, Yeah, and it's going to be if we're and we're really? And she went, that wasn't clear. Like, darn. I thought that was clear. And we're like, right, sure. But that but that has a lot of implications too, right? Because it like she's obviously not joining She's just using the XY coordinates to like to layer those on top of each other. So
Kirk (10:34)
Ha ha ha.
Fi (10:40)
They did so well on the
Kirk (10:41)
Mm.
Candice Munroe (10:52)
I mean in that her in her case I it was a date axis and she was planning her trip. So and then she was layering on four different data sources which had comparable dates So that could be cool.
Fi (11:03)
Yeah.
Sarah (11:06)
Yeah. It's some other things I see use for and I've seen it used in the map players is even just some basic formatting that Tableau doesn't allow for. So you know when things are kind of just not where you want them and I'm hoping with with this Viz layers that you'll be able to kind of, you know, get them get them a little bit nicer and you know, just little things I think like that that I've seen used in Map Players. So looking forward.
Fi (11:31)
I do have one of them.
Candice Munroe (11:31)
Yeah, definitely a little
bit of a cleanup too, right? Because there's the the conversion from different dynamic field formatting, she called it, we're using three different sheets using the hidden containers to like layer on different sheets is clunky, right?
Seems like a simple thing, but
Fi (11:47)
Yeah, I mean that's going to save so much time and performance So the performance is going to be awesome coming in, and Sarah's laughing on the inside because I like I'm kind of the performance one in our in our team and so coming through, so I I'm genuinely excited about that. But I did want to come back, actually pull back to solve on just one thing before we move on to the the next exciting topic.
Candice Munroe (11:52)
Yeah.
Fi (12:12)
So I heard word on the street, I'm not going to name names, but I heard that that it may be out in beta this year sometime. Now that would really surprise me. β I hope that that's something that comes through. but yeah, I wouldn't mention it unless it was somebody that would be in the know.
Candice Munroe (12:22)
Nice.
Kirk's going to go call Matthew Miller and make sure he's on the beta.
Kirk (12:34)
No,
well so the two things he was on record are is that everything had to be at least be code, not Figma. and the I think that it had to at least have the hope of shipping within the year. So that could line up.
Sarah (12:46)
Okay, so Candi, there's been a bit of a worry doing the rounds that Tableau Cloud has been racing ahead and kind of left Tableau Server a little bit behind. What do you feel that TC26 signaled about where server's heading? And is there anything you're excited about there?
Candice Munroe (13:03)
That Tableau Server seems to be getting a lot of love in that they're moving even the LLM stuff, they're moving to Tableau Server, I did not think that they would do that.
I think that really means that they've really accepted that it's going to be around.
Fi (13:19)
Interestingly enough, Paint with Data is also partnering with Biztory to use TabMove like Us as well, which helps people to move from server to cloud in an automated way. That's really exciting. What did you see in the product tabmove that made you so excited to jump on board?
Kirk (13:39)
well I saw you guys interview Timothy and assumed you knew what you were doing, so that was the only reason no, the fact that like they can move everything mostly. two things. One that like it's predictable enough that it can be a fixed price contract, which we normally hate doing because we get burnt every time. but this one's a lot more predictable, and that they can move, you know, everything.
And and Timothy, like only he can with his level of humbleness for how smart he is, is like, it doesn't always move virtual connections. I'm like, people have to be out of their mind if they're using virtual connections anyway, it's okay. So I would get them off it as part of this process.
Fi (14:07)
Yeah.
Kirk (14:13)
you have virtual connections, no. So yeah, the fact that it can move things that in the combination the prep doesn't need published data sources anymore, can use bridge directly, like makes that a lot better. So yeah, no, no, so we're super excited about it.
Candice Munroe (14:15)
No.
Fi (14:27)
diff
Candice Munroe (14:27)
guys
really interesting Timothy and Tristan the creativity to come up with the ideas they have I I attended Tristan's session as well and he basically walked through his hackathon project and it was really interesting thoughtful too like because I think the accessibility stuff like Tableau I don't know
what is in some of the other products for accessibility. But we have a client who is low vision and the way he describes how he uses it and and those things, like I think has not gotten a lot of love in recent memory. And they've put a lot of those features in now, which was also in Devs on Stage, there were some accessibility features that were mentioned as well.
And and some of the stuff is really cool, 'cause you don't necessarily think that somebody with accessibility problems is necessarily going to be using Tableau, but you know, in a workforce force like everybody has to have access to it. So yeah.
Fi (15:18)
Kirk, I know that composable data sources have been a real pet project for you on the lead up into TC.
Kirk (15:20)
T
Fi (15:28)
And they were talked about a lot and I'm super excited about them. Can you tell us what they actually unlock for an organisation?
Kirk (15:38)
yeah, they they unlock a lot. And I think we'll we'll even get more once we get into it. It feels like one of those you know capabilities that's going to allow us to do that, right? So first, I guess the problem without composable data sources. So what happens today is imagine you have a published data source that's available for you, and someone asks you a question about that data source and and you don't own the data source, right? So
What you effectively have to go back is go back and try to find all the tables that compose that data source. What were the relationships between them? find another table, the table of missing information, and then you have to republish that, right? So not surprisingly, people end up with like a bazillion published data sources. Now, in the age of AI, if it
catches on more. Like now when you get agents asking questions, they also talk about being chatty and using tokens. Now to figure out which of these bazillion data sources, but so do people even without the agents, right? So that's the first problem. The second problem is there are a lot of data prep problems that you can't solve in Tableau desktop or like in a published data source that you might want to use prep for. So
About three years ago I wrote a blog on Ken and Kevin's site on when to use prep and when to use desktop and when you can use the middle. and the reason why is because when you took a prep flow, unless you had access to a database, you have to publish it as a published data source and then it's a dead end. And it's also one big table, it's just a flat table. so now you'll be able to use
the output of prep inside other data sources. So that's at a high level. in the workflow though, where it's where they've really killed it with the UX, If you're in the flow of work, which was the whole point of Tableau in the first place, back to my example of you got this published data source, say, and someone says, we're talking about mapping, right? So someone says, I'd really love to map where our customers are
And you're like, wow. I can't believe geography is not in here. But say you have a geography table that sits there as a published data source, you can just go to your data source pane without touching the published data source, just go add another published data source, create a relationship to it, and then continue on with your analysis. And because it's always relationship-based and not join-based,
It's not going to break anything that you've already built. Because that's one of the other great advantages of relationships over joins is that your workbooks, nothing in your workbooks going to get messed up because it's just a table hanging off the end and it's not in the query tree of anything that you've that you've already built. So yeah, the use cases are going to be really, really big. So I'll wrap by another one. Like just like you guys came to the presentation, Candi and I did at TC. I think a lot of there's some really complex LODs.
Or even things you'd want to do in LOD that you couldn't do. So what we want is everyone wants this trade-off of I want atomic level detail, but then so I want every single transaction, but I also want to be able to do these like complex aggregations on top of them. Like now you could take that transactional table. you could create an aggregate table out of it, do your crazy LODs in it in prep so they only run once a day, take that table.
And just bring it back into your existing data source with the atomic level detail. and then you don't you know, Tableau's going to get a lot faster in that case. Like a lot faster and be able to do the kind of calculations that you can't do in Tableau today. So, yeah, you see a lot of this. I have a new blog out called Paint with Data Insights where I'm trying to write about this as well. I'll be doing a whole bunch of tug talks. I'm with Will Perkins, but we'll we'll like we'll be out there talking about this more 'cause I think it's
it's going to be really big the last thing is the semantic thing because you asked so almost no one puts comments on their fields in Tableau. I find almost weekly I go to people, you know, you can put a comment on that, and then other authors can see what the field's used for, and AI can read that field. And they're like, What? Like they don't even know it's there, right? So it's a lot easier now because
Take my other example where I had to go find all those tables, like pre composable data sources. Am I going to put all those comments in again? Right? And they might be different, right? But from a governance and scale, if we publish more tables as published data sources as opposed to full on data sources, then people can compose, which is where the name comes from, their own data source. And it's going to have all that kind of rich metadata, which is going to be a lot more like a semantic layer than you know these.
joined up published data sources that you know we see so typically today in Tableau.
Fi (20:14)
So you've mentioned there that we'll gain some performance improvements because we'll be able to push out the fixed LODs, for instance, into prep. β but we've also got the benefits of relationships as well, which often kick out the LODs as well. how is the performance that you've seen
Kirk (20:24)
Mm-hmm.
Fi (20:34)
on composable data sources when you're pulling t them together and creating the relationships, have you seen performance improvements holistically? Or have you seen that sometimes these relationships can also slow things down?
Kirk (20:47)
yeah, it's early days 'cause I only have it on the beta site and I don't want to put real data on the beta site. but in talking to the product management and development teams, it should perform as well as say a typical relationship or multi fact relationship model does now. Because it effectively creates the same query trees that they would.
what I'm hopeful for, I know they want to get a whole bunch of more features in. personally, what I'd like to see, they're working on things like creating aggregate metric tables and stuff in Tableau as a plan, sorry. I don't think they're working on. I'd rather them do a better prep integration directly into Tableau and and then work on optimizing these queries so that they get even faster. Cause I think I think that's where they have to go. Cause you know, they've been building these things like multi-fact and this.
I hope that they spend some time optimizing that a little bit. if you can imagine what I'm picturing as a UX almost is, when you drag a logical table in today and you right click and go open and then you get that old school Tableau join interface. Imagine if when you clicked open, you got all the options you had in prep and it brought you into prep to like create something. And then when you're done, you don't have even if a prep outflow. What you have is you have a logical table, which is what the prep
Sarah (21:48)
Mm-hmm.
Kirk (22:04)
output would have been without having to jump out to prep, do that work and come back in from a UX. But it'd be a lot easier than building something new. Do you know what I mean? Because they could effectively just call an existing component and just make the UX a bit better.
Fi (22:10)
Mm.
What
Candice Munroe (22:17)
you kind of have to take a step back how we envision this going because currently people have these massive data sources that they have kind of patched together to answer a particular question or for a particular project and then
it's attached to this dashboard, and then time moves on, and then they get a new dashboard, and somebody asks a new question, and they create a whole new big data source that answers that question that has components of this, and then maybe something else,
and then and then the point is is that in order to do that everything has to be a lot leaner it has to be clean it has to be certified it has to you know like in order to be able to do that and not spend a million dollars like in tokens
Like these things have to be very organised.
Like if we take the components of it, like we take the source datas and we build a good source data through Tableau Prep or whatever, we certify that and make that a component. And then we make another one a component. And it seems like then we're having all kinds of components, but in actuality, like having five components that keep getting matched together to answer different questions is less.
Than building these big data sources that sit in Tableau Cloud or Tableau Server and only get used once and then shelved and then forgotten about, and then we don't know what's in them. And but but by having these separate components and then putting them together at the time, like in different formations, makes our systems much leaner and easier to use and easier to leverage AI against, and therefore easier not to chew up a whole bunch of tokens.
Kirk (24:02)
Yeah.
Candice Munroe (24:06)
that's the whole like coming together of all of this stuff, like composable data sources and AI and having certified data sources a lot of the things that we see happening going forward, like are kind of going to come out of these pieces
Kirk (24:22)
Yeah, just quickly, the other that flows out of them is now you're not using service accounts for databases flowing around because people can just use the published data sources. if your published data sources that are your components of this composable data sources are all extracts, then you don't have to worry about great big snowflake bills or whatever. Like there's so many downstream benefits to this as well.
Fi (24:43)
Right. I I think just one thing that you touched on there is something that I've heard maybe four or five times in the last two weeks, which is people can save a lot of money through extracts because once you've extracted it up there, there's no ongoing cost for you to be running every like all of your queries against it. Whereas if you're running queries directly off Snowflake or Databricks or whatever,
Kirk (24:54)
Mm-hmm.
Fi (25:08)
your queries can be really chatty. I mean, we've got a dashboard at the moment, the front page is twenty five sheets. That's twenty five individual queries that are firing off. Like if that's going against a cloud database instead of going against a extracted data source, you know, that that will soon add up if people are using it regularly.
Candice Munroe (25:27)
people spend so much time on this data source development but it does help to have
a scaffold of what's going to happen at the top level we have these data components and then we can be flexible with how we put them together having these extracts and published data sources
gives us the flexibility, to not incur all these extra costs, which people don't think about.
Sarah (25:50)
I feel Tableau's been on a journey. When I started using it just over 10 years ago, I remember being a little bit shocked that there wasn't a semantic layer that I could access for this reusability. And we kind of had to think in big flat tables for a long time. And it's gone on that journey. Like you said, it's it's got certified ones now and relationships came along. But this feels like that kind of final step where it's going to allow us to think more.
like data analysts and data architects again. And we're going to have to stop making these like shortfalls that we've had to make in the past. So I'm excited to read your blogs that are coming up and we'll be sure to put those in the show notes.
Kirk (26:27)
Mm-hmm.
Fi (26:31)
Candi, AI showed up in nearly every demo this year. Can you explain to us in plain terms what does agentic analytics actually change about how a data team works day to day?
Candice Munroe (26:44)
Agentic Analytics opens up a natural language corridor into your data. So instead of having a static dashboard, it allows you to ask your question as opposed to having to spend a lot of time making visualisation.
Back to Kirk's comment about having a well curated data source that has descriptions and
things that the AI can use. that has to be all in place before somebody can just start typing a natural language question in against a data source to get something truly out of it. It doesn't mean that there's not work involved in it. it just means that if you've put in the time and you've put in the work and you and you've really governed your your data sources well and you've cleaned them up and made them AI ready
Kirk (27:23)
No.
Candice Munroe (27:36)
then when you really need the answer, then you can start to just ask it as opposed to have to dig into Tableau and start making visualisations and narrowing things down,
Sarah (27:45)
And I know you're talking about descriptions and comments Fi, you were working with one of our clients and you were getting AI within the tool to help with the descriptions and they were pretty bang on. So not the huge job that it maybe it needs to be, but definitely one that's required.
Fi (27:56)
Yeah it's
Candice Munroe (27:57)
Absolutely.
Fi (28:03)
Yeah, it's it's a skill, I think, and a discipline is probably the best way of describing it. No analyst or developer wants to document everything to the nth degree and provide that context. But the amazing thing about the new Tableau environment in Tableau Next, and that's primarily for Salesforce core customers.
Is that you have to go through the semantic layer in order to get to the visualisation. So it creates that barrier or that discipline. And because the Tableau agent is in there and you know the LLM requires that context of that query, it's really important to get it right. But I did find that when I hit the little like auto-generate, it it got me about
90, 95% of the way there and then I would just update it. It was the occasion that it was completely off track. But you know that's the same as any AI result.
Kirk (29:06)
my reaction when I use that is very when I look at an empty box and have to think about typing something, it's very daunting and I don't want to do it. When the AI gets it, even say it got 70% right, whether it's 70 or 90, it's a very human reaction, which might be a crappy one, but it is to go, that's not quite right. And like you want to fix it anyway. Like it like appeals to your I want to write a better, like it's so much easier to tweak an existing one than it is to write into a blank.
text box.
Candice Munroe (29:33)
Yeah,
to have to come up with like something entirely on your own it's easier to edit
And I agree, cleaning up a data source is monotonous work. So to have the AI kind of assist you and make it
more accessible is much better,
Sarah (29:48)
Now, Kirk, Tableau MCP has been out in the wild for a while now, Connecting to tools like Claude and Slack. What's the one thing a leader should be doing about it right now?
Kirk (29:52)
Hmm.
if you're going to embed it in a port so the first thing is you definitely have to not use the equivalent of a service token, right? So I'm not even sure
if you could do that, you'd be might be violating your license agreement anyway, right? So if you're going to use it inside a portal, you should be using without getting overly technical, you should be using the JSON web token and user attributes to do it. But let's just say a more simple case where you're going to let people ask questions in the pod. You should make everyone use their own personal access token, which is easy to grab. And then they should be using that when they register the MCP, because then if you have road level security, and I tested this recently at
like it it keeps the row level security. I even tried to get it to break the row level security and I got chastised by Claude for asking us to do this. In a good way. the other thing though is there's been a lot of people in the Tableau community, especially we were talking about semantic models already. Who've been very I don't want to hear about semantic models or whatever. I'll give you a good example of this is we're working with transportation company that does it.
today that's that we're just starting this Claude journey or whatever. And they have cost per mile. And the field's called CPM. and so you you guys do marketing. So we get they get the CPM field. You ask Claude, you know, I want to know the cost per mile, whatever. The field before the fields commented, right? what it does is it doesn't find that when I ask for cost per mile. I'm sure what it did was it went to the web and it looked at CPM and it went CPM is cost per 1000
impressions because that's the more common CPM definition. And then what it had to do is it went and it found a bunch of data sources and it created its own data source on the fly and then it created its own calculation. it's pretty crazy how many tokens it consumes to do that. I think someone I think it was Igor but I don't want to put him on the spot said you know they were doing
Candice Munroe (31:52)
Yeah.
Kirk (31:53)
They were testing their queries against MCP, and I think they're averaging 58 cents a query or something. You wouldn't use AI for very long if it was costing you 58 cents a query, right? So they're very chatty and less. So it's very easy to set up. If people haven't done it before, I recommend to do it. It's particularly easy in Cloud Desktop and free. You can have the MCP set up and running and asking questions in I bet you end to end two to three minutes.
What I recommend people do if they have a chatty tableau cloud site is make a new cloud site just for your AI questions and like strip down all the data sources and make it easy for the agent not to have to look all over the place to get answers to stuff and and risk confusing people. But it's pretty cool. and the new use case, it's not that far off because the desktop authoring API is in its third release now without getting too much into this.
Fi (32:29)
Sure.
Kirk (32:44)
Specifics under NDA, but people are going to come up with amazing use cases. Like Tristan had that one last year at DataFam Europe where two people could update a workbook at the same time. I don't know if you've seen that presentation. Amazing, right? But but imagine even little things like round of corners just came out, right? Imagine you've got 10 dashboards in a workbook and you say, I want round of corners on them. Imagine the pain in the butt, but you could just say,
Could you please round all my corners at, you know, at twenty five and do inner padding at 10 to make sure there's space and boom, your workbook would be updated, right? Like, so that is like very close. that spec's very close. So the AI can already do it. It's just the spec's not that there, right? And then the last thing I'd say is it's very easy to write skills if people haven't played with them in things like Claude. So a skill is
Fi (33:17)
Yeah.
Candice Munroe (33:19)
Ha ha.
Kirk (33:35)
you could write a skill to say, you know, we don't want people ever to have pie charts. even people are going to ask for pie charts when they do it, give them a tree map. And it like it'll force a tree map. So it and these skill files are written completely in natural language or whatever for it to interpret.
I think it's exciting. I think people are going to jump in, realise they get wrong answers a lot, and they're going to realise that they really have to curate their content properly to A not spend a ton of money and B make sure that people get the right answers if they make that available.
Sarah (34:09)
I'm just thinking about, if you could put your client's style guide in there with your own that says, I know, Fi's whole face lit up when you talked about like just padding consistency, like those things that are so tedious for us.
Kirk (34:24)
Right.
And it's as simple in the skill file of literally saying, use this color palette. Right. We always want our corners slightly rounded. We want inner padding to be this. We want if there's two objects side by side, give them five pixels, but give them ten on the outside, so there's ten pixels everywhere. And then you could just say update to our style guidelines and update the thing to your style guidelines. Like that's close. It's not there today, but it's like that's within
Candice Munroe (34:25)
Yeah, well you know
Fi (34:40)
Yes.
Sarah (34:41)
Yeah.
Kirk (34:51)
a small number of months, I would guess.
Candice Munroe (34:54)
Well, for a hot second there we had like templates, right? But you know, any good PowerPoint designer knows that if you're going to do a PowerPoint presentation and you could like copy the slides so that everything is in the same exact spot so that when you go from dashboard to dashboard and it doesn't bounce around. Like, that drives me crazy.
Fi (35:13)
Mm.
Mm, me too.
Candice Munroe (35:15)
Right? Because you can't
do that in Tableau. Like you have to reinvent the wheel every time because if you're doing a a dashboard or you a series of dashboards for a company, there's a side panel and like then you have to like you have to write down what all the parameters are and sometimes it's one and sometimes it's three. It's kills me and it's silly things. it doesn't really
Fi (35:20)
Yes.
Candice Munroe (35:36)
add a lot of value to the data and the analytics and the and the insights you're getting from it, but it's like very irritating when you're working with it.
Fi (35:46)
It's the trust,
it's the trust factor as well. So all of those little things that you are doing to make it look pixel perfect that, are boring to do, let's face it, they're boring to do, but it actually improves the trust when people look at it because they're like, things aren't moving around, they're not in a different space. I mean, I've been working on a client's risk and compliance.
Candice Munroe (35:50)
Yeah. Yeah.
Fi (36:11)
It's more of a presentation than a dashboard is what I would call it. so they they need to review what they're doing, you know, in terms of each quarter in their compliance. And I've been stuck with 350 sheets in this dashboard because it's quite comprehensive and each tile has somewhere between three and six sheets.
Kirk (36:14)
Mm.
Fi (36:35)
bringing it together and there's rounded corners. So there's everything that you're saying, Candi. There's different padding, there's different requirements in the in those corners. There's everything to get it looking absolutely perfect. But now they can run their risk reporting at the click of a prep button. You know, they run the workflow, it brings everything together,
it's really easy for them to input the data into, but that's not great for tableau. So we reshape the data and then it pushes it out and then they can just download it as a PDF and everything's done and it's perfect. But it takes a lot of patience. And it actually got me thinking, is this really what I want to be doing with my life? You know, like like and and so like, yes, the analyst role is
Candice Munroe (37:11)
Yeah Yeah.
Kirk (37:12)
Right.
Fi (37:17)
Evolving. Yes, there's more governance practices, which I don't think people are talking about enough or writing about enough at the moment. Around, you know, well, what does that look like? What does ethics look like? What does governance look like in what I'm pulling together? What do I need to be doing? And we need to evolve the way that we are delivering work. And hopefully some of this extra grind will be taken out, by.
Kirk (37:42)
Yeah.
Fi (37:45)
These great LLMs.
Candice Munroe (37:47)
Well, that's ultimately the thing, right? we had some poor little thing in a company that was like spending a month curating this Excel dashboard like that looked beautiful in the end. But like, do you know what I mean? Like she would get the download from the the company, then she would have to do all these
modifications to it and it would take a long time. And then we gave them Tableau. We gave a dashboard and Tableau, and now we seem to be getting bogged down with these little minutiae items, right? That, you know, that, you know, hopefully AI will clean up.
Sarah (38:20)
Yeah. So what I'm hearing is faster
to develop, get rid of all the finicky things or as much as we can, but watch out for those tokens. That's a that's a big call for the leaders. So maybe faster to develop, faster to perfect, but do look out for those hidden costs.
Kirk (38:31)
Yeah.
Fi (38:43)
So moving into the last part of our PODDY today, I thought I'd throw up β the final presentation slide that was at Devs On Stage, which shows you all the things that they talked about, not in labs, but what's actually coming as well. I thought it'd be good to put this up here and get you both.
to call out the things that mean something to you that we haven't spoken about and explain what they will mean to people out in the wild.
Candice Munroe (39:17)
there's a lot of interesting things there. β from the Tableau MCP is going to be huge because I think that's going to be β our vector into easily connecting Tableau up to an AI. It's it's pretty slick, as Kirk has alluded to. the rest API connector, I'm really looking forward to seeing.
the what the data devs do with that that's going to be a big community thing to see what those guys come up with. β
Like navigate and chart when β during the presentation was definitely my favorite section. β I liked Roxanne's story and I thought β that she had a lot of really cool features, that the chart layers all had their own data source. β so that was that's pretty interesting. I'll look forward to seeing
how that can be implemented β with β lots of different use cases. β I forgot about the drive time.
Yeah, that's going to be cool
Sarah (40:19)
And that was
I think I saw that in this release, 2026.2
Candice Munroe (40:24)
is that
I mean that could have a lot of implications, for the transportation companies. β I we
Fi (40:30)
And I think that people even
said to you, Sarah, that it was better than Google Maps. Was that right?
Sarah (40:35)
Yeah, that's what Roxanne was saying when we caught up with her afterwards. Yeah.
Candice Munroe (40:36)
β yeah, yeah.
Fi (40:40)
Yeah, so that's really exciting. the traveling salesman is always like one of the big things that people are trying to analyse and understand, how long it will take for people β to drive and so instead of having just an entire circular radius, the fact that there were ragged edges to the radius was super interesting. So it would be really cool for
clients that have area managers or deliveries that are going on.
Kirk (41:10)
Yeah, when I look at the list, I think β imagine data Tableau data sources on public could be really cool for things like, you know, they do IronViz and I always think I'm going to sign up and I'll get that data set and I'll play with and you never do, but if it would if you could just connect to it on public would be β would be kind of neat. β the bring your own connector to Tableau Cloud. A lot of people who probably aren't in the enterprise space don't get how big of a deal that is, because β
bridge is kind of a pain in the butt and that gets you around from having to use bridge, which could be β a pretty big thing for sure. The other one is I I hate β the idea of people using Teams 'cause it's such a crap product. But β but I do think that Tableau add in for PowerPoint could be big and help Tableau out like from losing maybe some things there to Power BI, etcetera,
it's becomes a pretty big deal. We talked about chart layers. Transparent tool tips would be cool for sure.
Sarah (42:05)
my β accessibility alarm bells went off when they said that one, β it's it's done.
Kirk (42:10)
It's fun it's funny that
it's β three after accessibility viz authoring too. SVG exports actually a big thing because some of the tableau image look pretty bad and SVGs scale obviously perfectly and are a lot smaller for sure. we talked about data driven formatting. The other thing that
I want to be able to do conversational data modeling. Like I don't expect it to build a data model once, but it would be nice to say.
You know, we use this table at as both a child and parent relationship. So sometimes it's got to join back to itself. but you know, worded in business speak and do it. And then β the last one I think they meant bring like the viewer a conversational analytics experience inside Tableau so they didn't have to go to Claude and use MCP to do it kind of thing. So as long as they make that last one.
β as accessible to viewers without some crazy license level, that could be really good. you guys left before I asked this question of Mark and Team, but I don't understand why they're so hyped up on this headless BI. I get why they have to do headless BI. I don't know why they would push people to headless BI though, it's great if you want to ask questions from
Claude or Cursor or whatever. But you know what? It's even better. You can ask them from within Tableau, right? Like why would they drive people out of the platform and they didn't need to drive people out of the platform? So β i if that's what that last one is, which I think it is, do you know what I mean? Could be pretty big as well.
Fi (43:38)
I kind of have a a different point of view on driving people out of the platform. I really love the direction that they're heading with with Headless. And the reason being it feels disjointed to force people to log in somewhere else to view their data and analytics. And if they can be doing it in the flow of where they are, now whether that flow is slack or
your favorite teams or β it's within Salesforce because they're an AE and they need to be doing their sales. That to me feels a more natural progression for people to use data than asking them, hey, can you log on here and then find the report and which is a clunky experience β when people aren't doing that naturally every day like us data people.
Kirk (44:30)
Yeah.
Candice Munroe (44:30)
Yeah.
what I think what is not obvious from the mobilise and unlock section that was more obvious live is there's a lot of interaction with Slack. So like Slack β played a big part in her β
showcase of the new features and how she dug into Slack and she looked for answers of like what could my next project be? Like what are people looking for? Tableau kind of combing through that thing and then producing a visualisation for her. That interaction I thought was really cool. Like for I mean we use Slack a lot. So I mean that could be very interesting
on a enterprise scale to be able to leverage those day-to-day conversations and determine like what people need answers to and like where they're going for those answers to sort of assimilate some of that information and bring it all in one β under one roof and and get everybody on the same page, I think could be key, like in the headless BI kind of conversation for sure.
Kirk (45:45)
so in the Tableau Plus SKU, or they had traditionally included Tableau Next. So in this way, I'm definitely agreeing with you, Fi which which is I wish they would have bundled Slack with it. Do you know what I mean? So like I like the idea of a
Fi (45:55)
β
Kirk (45:59)
Slack first kind of UI for a lot of this. Like, cause it's a conversational place anyway. Right. So my last name drop of the night would be like I talked to Mark Recher about this a couple of weeks ago. I like it because whether I'm talking into it when I'm speaking to a group to get an answer or I'm DMing someone or like a small channel, right? Or I'm asking an agent, that's a very natural feeling
Like they're all the same, kind of, if you know what I mean. So I do like the Slack first is like that UX is very natural. β and it's also because it can listen, it's also a great place β to get β your data driven alerts and all that stuff, which you know, we kind of get, but they're not, they're not that smart now. They could be a lot smarter, right? Like the the agent could be listening for things specific to you to be able to give you. I do like Slack first.
Fi (46:30)
Mm.
Kirk (46:52)
β for that stuff for sure.
Fi (46:55)
There's a couple of other things that you guys didn't talk about on this page that just make me really happy, β which is the Tableau Next extensions. So that's being able to use Tableau Cloud visualisations within Tableau Next, which I think is super cool given currently there's not a lot of flexibility in the different types of visualisations. So
Candice Munroe (46:55)
I think
Fi (47:23)
I'm really looking forward to that. β will be interesting because I'm assuming that many people won't end up with both the the the plus β licensing on that, but we'll see how it goes. And the other thing was just SVG export because I like to have a higher resolution on things that come out and sometimes the PNG's not good enough.
Sarah (47:44)
Hmm. one other thing the radial charts. Esther did present it in Tableau Next, and I asked her later on if that was coming to Tableau Desktop as well, which it is. She said she just chose to present it in Tableau Next. But one piece that was in there that I thought may have been overlooked was the ability to turn a ban on.
like just a quick hit of a button and a band pops up. So I'm hoping to see that in more charts as well. That would be really nice. most of our time we put a chart up there and we're always doing a BAN call out on it as well.
Kirk (48:18)
I think it feels like Tableau is the only BI product left that doesn't have a ban as a chart type too. Like everybody has a ban as a chart type.
Fi (48:26)
Interesting.
Sarah (48:27)
Good thinking.
Fi (48:28)
Alright, folks, Candi and Kirk, this has been a fun look at what's on the horizon for Tableau. Candi, we'd love to hear from you first. What's the one thing you really want a data leader walking away from and itching to go and explore?
Candice Munroe (48:44)
I really like it when the data leaders are β really driven I come from pharmacy and and healthcare and so often in my world before that it was like the data was so hard to get and you can imagine clinical trials and it's so complex and people don't understand the metrics. They don't understand the statistics behind it. So they just make these off the cuff kind of decisions.
but I'd like I'd like to see the data leaders build a data driven, data based decision making organisation and to be proud we're talking about having all of this in sort of in a flow of work. So
Is the answer there where they need it, as opposed to having to leave Tableau and go to a different product to like even like we were talking earlier about how like being able to have write back. So if they have to leave Tableau and go to a different product and update that file and then come back to Tableau and go, okay, is that right now? It's like that like totally kills their workflow. But as a data leader, if you want to build an organisation that's
found let's foundation is the data and the information and the insight, then you're going to want to have something that flows with people's work and that they can have it at their fingertips and they can have it where they're having those conversations as opposed to having to go, I got to go look that up. we're getting closer to that spot and we're getting as opposed to if you want to have a conversation about data
We're all going to have to move to Tableau. Instead of that, now we're moving from a situation where it's like, okay, well, where are you going to work? where are your conversations going on? Like let's move Tableau to that. And then you can just, work as you usually would and work in your workflow to make sure that those insights are surfaced when you need them.
Sarah (50:41)
Mm-mm. And I think we've been talking about keeping it in the flow for ages and it's nice to see it's it's getting there more and more. And Kirk, over to you. What would you like to leave our listeners with today?
Kirk (50:52)
first thanks for having us on. We had so much fun together in San Diego. It's nice to continue it here this morning tonight. I would say more than ever, the important thing is for data leaders to realise what you really are now more than ever business consultants. So it's not your technical skills that are the most important thing. And certainly
Don't get hung up on the purity of whether it be a visualisation or the way you use fourth normal forms in a database or whatever. Like you the key is like take the time to actually understand the kind of questions that the lines of business need answered and become an expert in being able to answer those with data. I think is what's really big. And then your plus one is
To be a connector Tableau makes this a lot easier now than it used to. When you're talking to sales, go, are you also talking to service? Because maybe when service calls go up, that's a leading indicator of your sales going down, or what it might be. So you can be the kind of connective glue that helps them work through that stuff, like by either disproving
things that they think are true that are causing friction, right? Or creating connections to make the business run better because there are correlations that matter. your consultant's first technology's second, right? I think is what you should get because the technology's changing quick so any so quick anyway. you know your technical skills obviously matter, but they don't matter like what they used to matter.
Sarah (52:25)
I think we forget how much departments work in silos
Fi (52:30)
Absolutely.
Candice Munroe (52:30)
Yeah, I think that connector role
Like to the to be able to have a bird's eye view of the organisation and their data needs is β really important more than ever to have is to set up the right backbone.
Fi (52:46)
So where can people find you both and learn more about your work?
Kirk (52:50)
certainly at paintwithdata.com, although we don't spend enough time keeping that up to date. We should it's always a good reminder. and then we're both, maybe me more than Candi very active on LinkedIn. But but especially if you're in North America, feel free to reach out. we do a lot of advisory style consulting services, which are our favorite gigs. So we let people
contract with us for very small amounts of hours, like even the smallest, say twenty hours. we do building out stuff, but sometimes it's just as simple as empowering them so that they don't make the mistakes that we've made over a long period of time, or because they don't live with this stuff day in and day out. So they can find us on LinkedIn or Paint with Data, I guess. Or now Paint with Data Insights as well. So
Fi (53:34)
Yeah, I mean we can both vouch for your fractional services since we've both used them as well. And it's been absolutely wonderful to know that we're not having to commit a massive amount of money, but then having that ability to have a conversation and go back and forth on things that aren't necessarily within our wheelhouse and just getting that support. So definitely worth that as well. And I think that
what we're seeing more broadly in market is that there's a lot more fractional services that are popping up globally.
Sarah (54:06)
Yeah. And of course if you are looking at upgrading from Tableau Server to Tableau Cloud and are in based in Canada, reach out to Kirk and Candi also. And of course if you're in New Zealand, Australia or Singapore, reach out to us.
Kirk (54:07)
I can
Fi (54:19)
Ha ha ha
Sarah (54:21)
So thank you so much for coming on and sharing everything you have today with us. We've genuinely learned so much.
Fi (54:29)
Your insights have been invaluable and we really appreciate you taking the time out in your evenings to spend it with us and with our community.
Sarah (54:37)
If you've loved this conversation as much as we did, hit follow, leave us a review, and share this episode with your fellow datafam
Kirk (54:38)
Yeah.
Fi (54:48)
What a conversation, Sarah. I mean, if you take things away to today, I think people should make it these: one, Tableau's core or the standardised products on desktop, cloud, and server are genuinely getting some love this year. two, the foundation that you're building with your data is definitely worth investing in. The composable, the governed, the well-modelled, everything that
the data and the visualisations get built on. And three listening to other people talk about what happened at TC reminds me there's so much coming and I'm so excited.
Sarah (55:29)
Couldn't agree more. Huge thanks to Candi and Kirk. You'll find all the links to their work and everything we've mentioned in the show notes below. And if you got a lot out of this, don't forget to share it with a leader or a data enthusiast you think would get just as much out of it as we did.
Fi (55:46)
And don't forget to subscribe. This has been Undubbed where we're unscripted, uncensored, and undeniably data.
