Tableau Pulse - Metrics Unleashed | Announced at Tableau Conference 2023

0:00Hey, it's Tim here. In today's video, we're

0:02going to cover Tableau Pulse, which was

0:03announced at the Tableau conference keynote

0:05. Let's get started.

0:07So the easiest way for me to explain this

0:09feature is to sort of walk through it chron

0:11ologically, and I'm going to try and edit in

0:13footage from the keynote in the order so

0:16you can kind of get a sense of how it works

0:18.

0:18As it is today, if you're using Tableau, in

0:20fact, let me hop out of full screen. If we

0:22go over to Tableau Cloud, you can see I'm

0:24just in a normal dashboard, and if I click

0:27on a data point, go up to watch, select

0:29metrics, you can see that we get this

0:31ability to create a metric on the right

0:33hand side.

0:34The problem is, is these have been limited,

0:36they've relied on the dashboard to be

0:38created. And one of the things you got a

0:40sense of at this keynote is that Tableau is

0:42, to borrow a word, decoupling so much of

0:44its analytical stack, so you can choose

0:46which parts of it you want to use and where

0:50.

0:50Embedding is one capability, but metrics is

0:52another. And so what they announced is

0:54Tableau Pulse. And if I can sort of try and

0:57describe to you what Tableau Pulse is, it's

0:59a range of technologies sort of brought

1:01together. It's a bit of metrics, a bit of

1:05Tableau GPT, which I'll cover in a separate

1:06video, and they've given it a landing page.

1:09And think of this landing page as a

1:10destination for people who typically don't

1:13start their analytical work in a data

1:15source, or these people might typically go

1:17straight to a dashboard and instead of

1:19going to a dashboard, they'll have this

1:21ability to go to a landing page, which

1:23covers the metrics that they're most

1:26interested in.

1:27How do you build a metric? Well, there's

1:29two ways. There's the most simplest way,

1:31which is an author, someone who knows a bit

1:33about the data source, can essentially use

1:35an interface to describe the metrics they

1:37want to build what kind of context, what

1:39kind of time periods.

1:41Once they've done that, that metric or that

1:43whatever we're going to call it, let's call

1:46it Pulse for now, becomes available in the

1:48Tableau Pulse interface. Once there, Table

1:51au GPT technology is able to look at that

1:54particular metric and add sort of small

1:57flares, additional questions, things that

1:59things you should know about.

2:01And it can augment that initial metric with

2:03some more information. That's sort of a

2:05rough guide of how I think it works. I have

2:07to say, I think because fundamentally it's

2:09not been released and what we probably saw

2:12was like a conceptual sort of working setup

2:14of it.

2:15So the really powerful thing here is that

2:17the way the interface is set up for users

2:19to build these metrics is actually quite

2:22accessible, a lot more accessible than

2:24something like Tableau desktop.

2:27And they're also quite powerful because the

2:29thing that I think absolutely everyone

2:32slept on during the keynote was the fact

2:34that these metrics can be embedded in a

2:37dashboard, in emails, in Slack, and also of

2:40course on Tableau service.

2:42So these are truly portable. And the

2:44biggest one is in a dashboard because when

2:46they're in a dashboard, some of the chart

2:48design and some of the way those things are

2:50set up, it's incredibly hard to get

2:52anywhere close to that even neat design.

2:55And if they're bringing that interface to

2:56an author, then this is going to really

2:58incentivize authors to start to use this as

3:00a way of building simple metrics that

3:02people can follow up on.

3:04Now, the other thing is that this

3:05technology is sort of infused with

3:07something called Tableau GPT. Tableau GPT

3:10was also announced at conference. I don't

3:13want to dive into that too much here.

3:15But nonetheless, Tableau GPT is playing the

3:17role of helping people ask smarter

3:19questions. It can do that in a couple of

3:21ways. The first one is by suggesting

3:23questions to you. So you can load up Table

3:26au Pulse and you might see three suggestions

3:28of things you can ask.

3:30The other way you could use this is just

3:31through the search function. So in the

3:33search function, you could type something

3:35in and it might create some sort of a chart

3:36and a story, which then might lead you to

3:38Tableau Pulse. That's sort of another route

3:39.

3:40And then the final route is you can almost

3:41ask an open-ended question. In the keynote,

3:44they showed a demo of someone asking, "Hey,

3:46what else should I know about AirFries?"

3:49And the great thing about that question is

3:50it obviously has context of the data source

3:52. It obviously has context of the open-ended

3:55ness of this question.

3:57And it's going away and finding some

3:58insight that could be pertinent to that

4:00particular thing you've asked, so AirFries

4:03in this case. And it then generated a chart

4:04.

4:05Now, that is a demo. That's sort of its

4:07conceptual way of working. But if it

4:08actually works like that, that is

4:11incredibly powerful because what you can

4:13start to do, I think, in the future is if

4:15you're an analyst, if you're an admin, the

4:18metadata of what's being asked is now an

4:21absolute goldmine.

4:23Because what people ask sometimes is

4:25different to what they tell you they need.

4:28And so by seeing the metadata, you can

4:30start to build a much richer picture of

4:32what data sources you need to go out and

4:35work with, what your data engineers need to

4:37be prioritizing in terms of workstreams and

4:39workflows.

4:40But also, more importantly, are people

4:42asking questions about data sources, fields

4:44, or assets that you don't yet collect? And

4:47therefore, you can sort of stand up the

4:48associated projects to make sure that data

4:50is collected and stored.

4:52So all in all, I thought Tableau Pulse was

4:54a really interesting feature set.

4:57Ultimately, I said in my keynote roundup

4:59that this was Tableau Metrics Unleashed,

5:02and I stand by that. I think it looks

5:05absolutely fantastic.

5:07The closest thing I've used to this is

5:08something called Altrix Auto Insight. And

5:11Altrix Auto Insight, to me, was always the

5:14example I would sort of showcase to people

5:17when people talked about our stated because

5:19the problem with our data in the past

5:21has been it's required you to ask questions

5:23in a very prescriptive way. And it's kind

5:26of coached you. The interface was designed

5:29around coaching you to ask the right

5:30question. Whereas this seems like a truly

5:32open way of asking questions.

5:35And this is sort of the power of GPT

5:36technology, large language models being

5:39trained on the way people talk and have

5:41conversational discussions around analytics

5:43. And now it's available inside of something

5:46like Tableau.

5:47So I think Tableau Pulse is super exciting.

5:49A few key things that I sort of wonder

5:51about, all of this AI technology has to run

5:54somewhere, the training, the learning, all

5:57of that has to run somewhere. And that

5:58would typically suggest that this might be

6:01a Tableau cloud product.

6:03If it's also going to be available on Table

6:04au Server, then I think the genuine question

6:06that Tableau Server admins are going to

6:08start asking is, what are the resource

6:09implications for me? Because AI model

6:12training is not an easy or straightforward

6:14thing.

6:15And if it's the kind of thing that can run

6:16in the background on very small resources

6:19or can run in a sort of quiet time of the

6:20day, then fine. But as these features start

6:23to haul out, you're going to see computing

6:25requirements and minimum requirements start

6:28to step up.

6:29And it might just be that Tableau cloud is

6:30the only way to keep up with those

6:32requirements. So let's wait and see. Anyway

6:34, that's it for this video. Hopefully catch

6:37the next video where I'll talk about Table

6:39au GPT and what was announced at keynote.

6:42Thanks for watching, and I'll catch you in

6:42the next one.

6:43Transcribed by https://otter.ai

6:55[ Silence ]