Tableau 2020.2: The new data model and relationships

0:00In Tableau 2020.2 Tableau have added a new

0:02capability inside of Tableau called the

0:05data

0:05model. Now this data model has actually

0:08always existed, this is essentially an

0:10enhancement to

0:11an existing feature and that specific

0:13enhancement refers to relationships, being

0:16able to define the

0:17relationships between multiple tables in

0:19our data set. Now let me dive into this so

0:22I can show you

0:23the new interface and how this new feature

0:26works. I'm going to connect to a sample

0:28file that was

0:28actually used in the beta testing and alpha

0:31testing for this particular feature. It's

0:34called

0:34the beta bookshop and you actually see some

0:36of Tableau's video uses this data set. I'll

0:38try and

0:38put a link to it so you can download it

0:40from my dropbox if you want to follow along

0:43but otherwise

0:44I'd actually just recommend you just watch

0:45this video and try and understand the

0:47concepts on its

0:48own. Now the thing you'll notice

0:50immediately when you open this is that

0:52nothing is different. I've

0:54just connected to an excel file and apart

0:56from this icon you know asking you to drag

0:58the tables

0:59here which is slightly updated everything

1:01else is the same. The first time you'll

1:04notice the

1:04difference is when you drag in a table into

1:07your view. So as I drag it in and I drop it

1:10in the first

1:10thing you'll notice is that instead of a

1:13rectangle we get this square and this

1:15square is a little bit

1:16different. It doesn't have the typical sort

1:19of interface we're used to and we now have

1:22some new

1:22options here but I'm going to carry on and

1:24the first thing I'm going to do is I'm

1:26actually going

1:27to try and create a join as I typically

1:29would have done in my old data set and I'm

1:32going to join the

1:33author information to this book. Now when I

1:36go to do that you'll notice two things.

1:39There's a new

1:40link being created here. This is affection

1:43ately known as the noodle but it actually

1:46just signals

1:46to you that something is new and different

1:49here. It's not going to behave as we used

1:51to

1:51and as I drop that in and I zoom back out

1:54you'll see the new interface for

1:56relationships and so

1:58what's actually going on here? Well to

2:01summarize the data model allows us to work

2:04with data on

2:05two different levels. The main thing we've

2:08been used to up until now is called the

2:10physical layer

2:11where we can describe at a row level the

2:13relationships between data sets and then we

2:16can also do things like aggregated joins

2:18which are typically called blends in Table

2:20au and those

2:21have always worked at the physical layer.

2:23Whatever result that comes out of Tableau

2:26is essentially

2:27a table that actually has the information

2:30and query built into it. In this new

2:32version we've

2:33got something called a logical layer and

2:35that's what we're looking at here where we

2:37just define

2:38the relationships between the two tables

2:40and then Tableau does the hard work of

2:42figuring out how to

2:44aggregate the data set, how to avoid things

2:46like duplications and then it gives us the

2:49data as we

2:50should see it and so you'll notice that

2:52even though I've brought the author table

2:54in my summary

2:55doesn't have the bookshop data available.

2:58Okay you can see I have a little bug here

3:01where these

3:01disappeared for a second but you can see

3:03here that I only have data from my author

3:05that's because I

3:06have it selected here and that's why it's

3:08gray. If I select the book you'll see that

3:10that switches

3:11over and now this is gray and this is white

3:13and now you can see the information here.

3:16So some of

3:17you then might ask well how do I do a join?

3:19How do I do a blend? Have those been

3:21removed from the

3:22product? Well the answer is no because

3:24there's still a place for joins and blends

3:26in the new

3:27data model. Okay let me give you an example

3:29now. I have some information here about

3:32each and every

3:33book and I've sort of deliberately created

3:35a scenario where I'd want to do a join on

3:38this

3:38data set using a combination of these two

3:41fields to create the book ID. You can see

3:44an example here

3:45but I've actually just created these inside

3:48of excel. Now the thing here is I can't use

3:51join

3:51calculations in a relationship so I have to

3:54do a join in order to get these two fields

3:57to match on

3:58to my book table. Okay so to get the join

4:00interface up what you do is you go to the

4:03connection you'd

4:04want to join to and then click open and

4:06when you do that you then get the

4:08traditional view we're

4:10used to. You see this long rectangle and

4:12now what we've done is we've hopped into

4:14what's called the

4:15physical layer. At this point I can then

4:18drag the information into my view and you

4:21get the traditional

4:22sort of join interface that you're used to.

4:25You can then click on this interface and

4:27essentially

4:28edit the information. In my case what I'd

4:30like to do is I'd actually like to bring in

4:32all information

4:33related to the books. So I'm going to do a

4:36left join and I'm going to change this join

4:39to work

4:39slightly differently. I'm going to go ahead

4:41and create a join calculation and this is

4:43why I have

4:44to use a join in this particular case

4:46because I can't do this with relationships.

4:48So let's just

4:49type this in book id one and plus book id

4:52two and this basically has the effect of

4:56concatenating

4:58these two ids into one new id and when I

5:01hit apply and click ok you can now see that

5:05that physical

5:06table is actually manifested here. I can

5:08see the book and the info in one table and

5:10you've got the

5:11join information here and so this is pretty

5:14much ready to go. Now in order to exit this

5:18physical

5:18layer I just need to hit the x here on the

5:20top right hand side. You can see I'm

5:22circling it with

5:23my mouse and so when I hit x out of that I

5:26'm now back onto the logical layer and now

5:28you see

5:29something new. Tableau is letting you know

5:32that you've got a join in this particular

5:34table

5:36and it's made of two tables and if you

5:37double click it you can see the joins and

5:40again we can

5:40exit that to go back out to the logical

5:43layer and now you can see these two items.

5:46Now the key thing

5:47to understand here is that when you then go

5:50and start building visualizations using

5:52this new data

5:53model Tableau is going to be doing some

5:55work behind the scenes to figure out what's

5:57the best

5:58way of connecting these two data sets. So

6:00let's go ahead into sheet one and

6:02immediately you're going

6:04to see the next new thing and that is the

6:06interface here on the left hand side. If I

6:08just

6:08sort of scroll in and zoom into this you'll

6:11notice two things. Firstly the different

6:14iator between

6:15dimensions and measures has sort of been

6:18removed and been replaced by this line here

6:20. So you just

6:21see the subtle very subtle line next to my

6:23mouse here where the arrow is you can just

6:25see me

6:26hovering to the right of it. That line

6:28shows you the distinction between your

6:30measures and

6:31dimensions on a per table basis. The next

6:34thing to be clear about is that the tables

6:37each have

6:37their own sections and I can actually

6:39collapse these tables like so and then you

6:42have some

6:42general calculations below here. Now the

6:45way this works is if you create a

6:47calculation that spans

6:49metrics across multiple data sets then they

6:52'll actually appear here at the bottom of

6:56this view.

6:57So if I get a metric from books and authors

6:59the calculation will appear here. Now if I

7:03create a calculation that is solely using

7:06items from the book table then that

7:08calculation will

7:10appear here at the bottom of the book table

7:12and it will appear in the usual sort of you

7:16know

7:16expectation where you'll have you know

7:18dimensions up here and measures just below

7:20here. So that's

7:22how that works. Now the next thing to be

7:24aware of is that you can still see the

7:26difference between

7:27a dimension and measure if you just click

7:29and drag one of these two you'll see here

7:31you get

7:32this differentiator that shows you where

7:34the separator is between dimensions and

7:36measures.

7:37Okay so that hasn't gone away it's still

7:39there and you can see I've changed this to

7:41a dimension

7:42and I can drag it right back and put it

7:44back as a measure. Okay we've introduced

7:47you to the

7:48interface we're now in a sheet and we're

7:50already connected to our data. Well how

7:52does this change

7:53the behavior of our data source? Well the

7:55thing to be aware of here is that when we

7:58start bringing in

7:59items let's say I bring in a book in this

8:01particular case the book id I can bring in

8:03a title as well

8:04just to make this a little bit more person

8:06able. Okay we can just go ahead and interact

8:09with our

8:09other data set so I can bring in first name

8:12and last name of the other authors and in

8:15the

8:15background Tableau is essentially doing the

8:18queries required to go and bring that data.

8:20Now you might also get a scenario where you

8:24want to do a different type of question you

8:27see a lot

8:27of analysis when you do joins is typically

8:30predicated on a primary data set so you

8:32typically

8:33ask questions from the perspective of that

8:36primary data set but what if you wanted to

8:39ask questions

8:40in a slightly different order? What if you

8:42wanted to analyze all the authors who are

8:44currently

8:45writing books but haven't published a book

8:47yet? Well how would you do that? Well here

8:50in Tableau

8:51because we have this relationship we don't

8:54have to create another connection from the

8:57perspective

8:58of the author you see back in the previous

9:00versions of Tableau that's what we would

9:02have

9:02to have done and then potentially blended

9:05the data in to see those two perspectives

9:07but here

9:08we can actually answer that question

9:09ourselves just by building a different view

9:12and I'll go

9:12ahead and do that in another sheet. Okay so

9:15we're in the sheet I'm going to go ahead

9:17and start

9:17building the view first thing I'm going to

9:20do is actually bring the author ID into the

9:22view okay

9:22and then what I want to do is bring the

9:25count of books written by that author and

9:28so if I just

9:29drop that in you'll notice that you see I

9:32have some authors who haven't got any books

9:35to their

9:35name okay and so what I can do is I can

9:37actually bring that into the filters just

9:39by dragging that

9:40in and filter this to show the values that

9:43show zero and here are my authors who haven

9:47't according

9:48to my data set published a book. Now it

9:51might just be that this relationship needs

9:53a little bit more

9:54analysis or you might have gaps in your

9:56data but the key thing here is I'm not

9:58having to go back

9:59and change my connection I'm doing

10:02everything from one connection okay so this

10:05is a pretty powerful

10:06new feature it fundamentally changes the

10:08way you think about data sources it's going

10:10to change the

10:11way you work with data sources as well but

10:14it doesn't materially change the type of

10:16charts or

10:17visualizations that you can build it just

10:19makes it much much easier to get to the

10:21data that you

10:21need much much faster and also in terms of

10:23performance it might improve the way things

10:26work because whereas in the past you might

10:28have had for example one connection and

10:31then three

10:31other connections to blend to at different

10:34levels of aggregation while with

10:35relationships you can

10:36do all of those different levels of

10:39aggregation directly from one connection

10:41because each view

10:42has its own unique query to the data source

10:46each view has a different substantiation of

10:49the same

10:49data and you define that by basically

10:52building your visualization and also doing

10:54calculations

10:55in tableau and so that's a really really

10:57powerful powerful concept to be aware of.

11:01Now I can't possibly cover all the

11:03important information that's required to

11:06understand for

11:06this feature the best way to learn is to

11:08actually start playing with it yourself and

11:11start

11:11understanding how it works open existing

11:14workbooks maybe take on some challenges

11:16that you typically

11:17needed blending or some really complex LODs

11:21for but tableau have got some great

11:23resources that

11:24I'm just going to highlight right now I'm

11:26going to open up my powerpoint here because

11:28I've actually

11:29got a slide with some really really

11:30fundamental concepts that they've got on

11:32their documentation

11:33you can find links to these images and

11:36tables in the description below but one

11:39thing to be aware

11:40of here is that all of this information has

11:43been sort of synthesized in a couple of

11:46very simple

11:47articles that highlight the difference

11:49between the two concepts and the key thing

11:52to also remember is

11:52that we're not losing the old behavior of

11:55joins and unions and blends we still have

11:58that capability

11:59we're just adding a new concept and so as I

12:01go through these images you can see that

12:04tableau

12:04have very clearly articulated how these

12:07things work and also differentiated the

12:09difference

12:10between the two this table in particular is

12:13a really powerful explainer of what is

12:15going to

12:15be different from the two and also what to

12:18expect because there are a lot of quirks

12:20and habits that

12:21we've gotten used to where actually now we

12:23can be a little bit more brave with how our

12:26data sets work

12:27and it might meaningfully change the amount

12:29of data preparation that you're requiring

12:30to do

12:31up front there's also some comparisons

12:34between relationships and blends and when

12:37they can be

12:38used and then finally there's some guides

12:41of when to use each of these options so

12:43relationships

12:45become the new default in fact every time

12:47you open up a visualization you start

12:49building and

12:49connecting to data it will by default start

12:52with relationships and in most cases this

12:55will normally

12:56just work fine but if you start noticing

12:58some odd behaviors or you think those

13:00relationships aren't

13:01performing correctly you need to bear in

13:03mind that it's not going to solve every

13:05single scenario for

13:06example here we have joins which are still

13:08applicable I showed you an example of that

13:11before

13:11and also unions and blends will still have

13:13their individual places when you have data

13:16sets coming

13:16from lots of different places and you need

13:19to be able to sort of do some basic

13:21aggregation to

13:22those before you then use them in the

13:24relationship so this is a really really

13:26powerful feature

13:27the last thing to show you is if I go back

13:29to tableau and I go back and edit my data

13:32source

13:32for the advanced users out there this

13:35relationship is actually quite a powerful

13:37thing because not

13:38only do I have the ability to you know use

13:40one field for the relationship I can add

13:42more fields

13:43as part of the relationship and that's a

13:45really powerful way of sort of working with

13:47this but the

13:48other thing is if you actually understand

13:49some of the advanced concepts behind this

13:51data model

13:52what you can do is you can actually give

13:54tableau some information about the nature

13:56of your data sets

13:57but the key thing here is these are

13:59actually quite sensible defaults so unless

14:02you understand

14:03these concepts of cardinality referential

14:05integrity don't change these they can do

14:09some really funky

14:10things with your data sets but if you do

14:12understand them then you've got the sort of

14:14granular controls

14:15here in order to sort of play around with

14:17how these work and so what output you get

14:20from your

14:20data source okay that's been a really short

14:23introduction into the data model I highly

14:26encourage you to have a play with the

14:29feature and get in touch with comments

14:30below I'm probably

14:31going to find a way of doing a series of

14:34videos on this topic with an example that

14:36we can follow

14:37through lots of different scenarios but

14:39that will probably take a bit of time to

14:41find a good example

14:42data set that people can follow along with

14:44but notwithstanding that what I will do is

14:47I'll start

14:47posting blog posts to this particular topic

14:49that cover this in more detail as they come

14:52out from

14:52the community so be sure to check back on

14:54this video and see what what blog blogs

14:56have been posted

14:57in the description below otherwise that's

14:59pretty much the end of the video if you

15:01like the video

15:01hit hit like hit subscribe if not if you've

15:03got some feedback drop it in the comments

15:05below and

15:06I'll try and get to those very soon

15:09otherwise I'll catch you in the next video

15:11take it easy