Can you build an UpSet Plot with just one calculated field?

For this week’s #WOW2023 challenge , guest poster, Venkatesh Iyer asked us to create an UpSet Plot, with the added requirement of using just 1 calculated field.

To start with, I had to read up on what an UpSet plot was and looked through the blog post by Chris Love that was referenced in the challenge introduction. While this post gave me more clarity, it introduced more calculations than I was hoping for, so I started looking a bit wider for a bit of help. This YouTube video set me on my way.

Setting up the data

The requirements stated to limit to the Category of Furniture only, so after connecting to the Superstore data source, I added a Data Source Filter (right click data source > Edit Data Source filters) to restrict the information throughout the workbook just to the Furniture Category.

Doing this means I don’t have to keep adding the Category to the Filter shelf, and any FIXED LoDs I create will only be based on the subset of data that has been ‘pre-filtered’.

There are multiple charts in this challenge, and I used 5 sheets in total. Let’s start with the easy ones.

Building the Customer List

On a new sheet add Customer Name to Rows and Sales to Text. Format Sales to $ with 0dp, and widen each row. Remove row/column dividers and remove the ‘Customer Name’ column heading (right click and hide field labels for rows). Uncheck the Show Tooltips option from the Tooltip shelf. Name the sheet Customer List.

Building the Sub-Category Bar Chart

Add Sub-Category to Rows. Add Customer ID to Columns, then use the context menu to change the field to use the Count (Distinct) measure.

Note – I would typically create a specific calculated field containing the function COUNTD([Customer ID]), but as we only want a single calculated field in the solution, then this is the method to adopt.

Sort the resulting bar chart descending, and add Sub-Category to the Colour shelf and adjust to suit.

Widen each row and then click on the Label shelf and check the Show mark labels tick box. Align the labels middle left and format the font. Hide the axis and the Sub-Category column heading (uncheck show header). Remove all gridlines, row/column dividers/ zero lines etc. Uncheck the Show Tooltips option from the Tooltip shelf. Name the sheet Sub Cat Bar.

Identifying the groupings

Ok, now we’re at the point we need to identify the different ‘cohorts’ of customers based on what Sub-Categories they have purchased. Let’s build out a tabular ‘check’ sheet so we can see what we’re up to…

On a new sheet add Customer Name and Sub-Category to Rows. This simple table shows us that Aaron Bergman has at some point only ever bought Bookcases & Chairs, while Aaron Hawkins has purchased Chairs & Furnishings. These 2 customers are in different cohorts as they haven’t bought exactly the same combination of Sub-Categories. There are 15 different combinations in total.

Based on what I observed in the video, I can create a FIXED LoD calculated field to identify if a customer has bought Bookcases.

BC

{FIXED [Customer ID]: MAX(IIF([Sub-Category] = ‘Bookcases’,1,0))}

Pop this into the view, and we can see that there is a 1 reported against all the rows associated to each customer who bought Bookcases. So Aaron Bergman has a 1 against both rows, and Aaron Hawkins has 0 against both rows.

Creating similar calculated fields, specific for each of the 4 different Sub-Category values, and putting them into the table, we can see we have various combinations of 1s and 0s for each customer. Adam Shillingburg has bought all 4 types, so has 1’s across the board, while Adrian Sharni has only bought Furnishings, so has 3 0’s and a single 1.

Based on our understanding of what these fields are doing, we can combine what each one is doing into a single calculated string field.

Combo

STR({FIXED [Customer ID]: MAX(IIF([Sub-Category] = ‘Furnishings’,1,0))}) + ‘-‘ +
STR({FIXED [Customer ID]: MAX(IIF([Sub-Category] = ‘Chairs’,1,0))}) + ‘-‘ +
STR({FIXED [Customer ID]: MAX(IIF([Sub-Category] = ‘Tables’,1,0))}) + ‘-‘ +
STR({FIXED [Customer ID]: MAX(IIF([Sub-Category] = ‘Bookcases’,1,0))})

Note the order is based on the order of Sub-Category bar chart. Add this into the view on Rows (rejig the order of the measure values to match).

So with this one calculated field Combo, we now have a dimension we can use to count the customers against. The calculated fields I used to demonstrate the concept are now superfluous and can be deleted if you wish, if you remove the check sheet too. I chose to keep mine in for reference.

Building the Combo bar chart

ON a new sheet, add Combo to Columns and then add Customer ID to Rows, but as before, set it to use the COUNTD aggregation.

Sort the bar to be ascending. Check the Show mark labels option on the Label shelf and adjust the alignment to be bottom middle, and the font to be bold. Change the colour of the bar to suit. Hide the axis and the Combo heading (uncheck show header). Remove all gridlines, row/column dividers/ zero lines etc. Uncheck the Show Tooltips option from the Tooltip shelf. Name the sheet Combo Bar.

Building the dot plot

Add Sub-Category to Rows and Combo to Columns. Manually sort the Sub-Category rows so they are listed with Furnishings at the top then Chairs > Tables and Bookcases at the bottom. Sort the Combo field by COUNTD of Customer ID ascending

Double click into Columns and manually type MIN(0) to generate an axis. Change the mark type to circle. Add Sub-Category to Colour. Widen each row.

Add Sub-Category to Label. Then double click into the pill on the label shelf and manually change to add the LEFT function around the pill, so the pill becomes LEFT([Sub-Category],1) to get the initial. Again, this is typically something I would explicitly store in its own calculated field. Manually re-sort the rows again, as this seems to break the sorting.

Align the label middle centre and bold the font. Then add another instance of MIN(0) on Columns to create a 2nd marks card. Change this mark type to Line. Remove the field from the Text shelf, and move the Sub-Category pill from Colour to Path.

Make the chart dual axis and synchronise axis. Right click on the MIN(0) axis title at the top of the chart and move marks to back.

Then hide the axis and the Combo column and Sub-category row (uncheck show header). Remove all gridlines, row/column dividers/ zero lines etc. Adjust the size of the marks as required. Uncheck the Show Tooltips option from the Tooltip shelf. Name the sheet Dot Plot. .

Building the Legend

The quickest way to do this is to duplicate the dot plot sheet. Then remove Combo from the Columns. mark type of the line from line to circle and decrease the size to the smallest possible. Move Sub-Category from Detail to Label and align middle right.

Edit the label and add some spaces in front of the text to push the labels further to the right. Format the font to suit.

Adjust the MIN(0) fields to both be MIN(0.2) instead (just double click into the fields to edit). Then edit the axis to be fixed from 0 to 1. This forces the display to the left.

Hide the axis, and name the sheet Legend.

Adding the interactivity

Arrange the items on a dashboard, using layout containers and padding to organise the 4 main charts and ensure the dot plot aligns both vertically and horizontally with the other two bar charts.. The legend chart is a floating object.

Add a filter dashboard action to filter the customer list

Filter Customers

On select of the Combo Bar or Sub Cat Bar sheets, target the Customer List sheet, showing all values when the selections are cleared.

And that should be it! My published viz is here.

Happy vizzin’!

Donna

Can you create a dynamic small multiple chart?

For this week’s #WOW2023 challenge, Lorna asked us to recreate this small multiple (or trellis) chart which organises the time series charts per Sub-Category into a grid format, where the number of columns is determined by the user.

Whenever I need to build these types of charts, I often end up referencing this blog post by Chris Love from 2014, as this has the basis for the calculations required.

To get started, we need to capture the number of columns based on a parameter

pCols

integer parameter ranging from 1 to 5 with a step size of 1, that is defaulted to 5

On a new sheet, display the parameter, and add Sub-Category to Rows. Apply a sort to Sub-Category based on the field Sum of Sales descending.

Based on the pCols parameter, we need to determine which column and subsequently which row each Sub-Category should be positioned in. We will make use of the index of each entry in the list. Double click into the Rows shelf and manually type in INDEX(). Change the field to be discrete (blue). This will number every Sub-Category row from 1 upwards. To be explicit, edit the table calculation, to explicitly set it to compute using the Sub-Category dimension.

To determine the column for each sub-category

Column

(INDEX()-1)%[pCols]

the % symbol, is the modulo and returns the remainder when the INDEX()-1 is divided by pCols – ie if INDEX() = 12, then 12-1 = 11 and 11 divided by 5 is 2 with 1 left over, so the result is 1.

Add this to the sheet, set it to be discrete (blue) and also edit the table calculation to compute using Sub-Category. You can see that Chairs and Machines are in the same column. If you adjust pCols, the values will adjust too.

To determine which row each Sub-Category will be positioned in we need

Row

INT((INDEX()-1) / [pCols])

This divides INDEX()-1 by pCols and just returns the whole number. ie if INDEX() = 8, then 8-1 = 7, and 7 divided by 5 = 1.4. The integer part of 1.4 is 1.

Add this to Rows and set to be discrete, and adjust the table calculation as before. You can see Chairs and Phones are in the same row (but different columns), which Chairs and Machines are in the same column, but different rows.

Let’s rearrange – Move Column to Columns, Sub-Category to Text and remove INDEX() altogether, and you’ll get the basic grid layout we need.

Create a new field to store the date part we’re going to present

Month Order Date

DATE(DATETRUNC(‘month’,[Order Date]))

Add this to Columns, and set as exact date and add Sales to Rows and move Sub-Category to Detail. At first gland this may look ok, but if you look closely, you’ll notice that there are multiple lines on some of the charts.

This is because there are some states that didn’t sell some of the sub-categories on the month, and this affects the index() calculation when the Month Order Date is set to be a continuous (green) pill (the viz below highlights this better – Accessories is now indexed with 6 and 7…

So to resolve this, add Month Order Date as a discrete (blue) exact date to the Detail shelf underneath the Sub-Category field. Then change the Month Order Date field in the Columns shelf to be a Continuous (green) attribute. Then adjust the table calculation on both the Column and the Row fields, so they are computing over both Sub-Category and Month Order Date, but at the level of Sub-Category.

Format the ATTR(Month Order Date) field on Columns to be the custom format of yyyy, so the axis just display years

and then format the Month Order Date field on the Detail shelf, to be the custom format of mmmm yyyy, so the information in the Tooltip will display the date as March 2001 etc. Adjust the Tooltip to match.

The label for each Sub-Category needs to be positioned based on the y-axis at the maximum sales across the whole display, and on the x-axis at the last point in the date scale ie December 2023. For this we need

Max Sales in Table

WINDOW_MAX(MAX([Sales]))

Label Position

IF LAST()=0 THEN [Max Sales in Table] END

Add Label Position to Rows and adjust the table calculation so the Max Sales in Table nested calculation is computing by both Sub-Category and Month Order Date, and the Label Position nested calculation is computing by Month Order Date only. This should result in a single mark per Sub-Category displaying.

Make the chart dual axis and synchronise the axis. Remove Measure Names from the All marks card. On the Label Position marks card, change the Mark type to shape and select a transparent shape (see this post for details on how to get this set up). Move Sub-Category to Label and align top right.

Finalise the display by hiding the Column and Row fields (uncheck show header), hiding the right hand axis (uncheck show header). Format to remove all gridlines & zero lines and hide the null indicator. Remove the axis title.

You should then be able to just add this to a dashboard. My published viz is here.

Happy vizzin’!

Donna

Can you proportionally brush from multiple selections?

For this week’s #WOW2023 challenge, Kyle wanted us to build a viz that used selections on the viz rather than a set of filter controls to show how the sales for those selections were distributed.

This concept is referred to as proportional brushing and makes use of set actions to achieve the results. The complexity added here was the multiple selections being made.

6 sheets make up this dashboard – 1 for each bar chart, 1 for the KPI and 1 for the breadcrumb trail.

Building the basic bar charts

Create 4 sheets, one for each of the Region, Segment, Ship Mode and Sub-Category dimensions. The simplest way is to build one sheet, get all the formatting applied etc, then duplicate and replace the dimension on the duplicated sheet with the new one.

When building the first sheet, place the dimension (eg Region) on Rows and Sales on Columns, sorted descending. Adjust the Sales to be formatted to $ with 0dp. Hide the Sales axis, and format to remove all gridlines/axis lines/ zero lines and row/column dividers. Show mark labels and align centrally. Adjust the font label to 8pt. Widen each column if need be. Hide the dimension label from displaying (hide field label for columns). Adjust the tooltip to suit. Name the sheet based on the dimension.

Then duplicate this sheet, and drag the next dimension, eg Segment, and drop it directly on Region. If done properly, everything should seamlessly update. Re-name this sheet accordingly, then repeat the process until you have a sheet for each of the four dimensions.

Applying the proportional brushing

Create a set for each of the relevant dimensions.

Region Set – right click on the Region field in the data pane and select Create > Set. Select all the options to be in the set.

Repeat and do the same for each dimension, so you end up with Segment Set, Ship Mode Set and Sub-Category Set.

We need to determine the combination of all the values selected in each set. So we need

Is Selected Options

[Segment Set] AND [Ship Mode Set] AND [Region Set] AND [Sub-Category Set]

This returns true for all the records in the data which match the combined selections of the individual sets.

On the Region sheet, add Is Selected Options to the Colour shelf. The right click on each set in the data and and select Show Set, so the set of selections are listed on the canvas.

Change the options so only the Segment Consumer and rthe Ship Mode Standard Class are selected, along with all Region and Sub-Category values. Adjust the colours associated to the True and False values that are now presented

If need be, adjust the tooltip so the Is Selected Options is not displaying, then add the Is Selected Options field to the Colour shelf of the Segment, Ship Mode and Sub-Category sheets. Play with the set selections to see how the bars change. Once you’re familiar with the behaviour, reset all the sets so they all contain all the values.

Building the KPI sheet

On a new sheet add Sales to Text. Change the mark type to shape and select a transparent shape (see this blog to get this set up). Adjust the Label to include the text ‘Sales’ and format accordingly. Align middle centre. Add Is Selected Options to the Filter shelf and set to True.

Again, if you adjust the set selections, the value will adjust accordingly.

Building the Dashboard interactivity

Add the sheets onto a dashboard. I used both vertical and horizontal layout containers to get the objects positioned where I wanted. I also used blank objects set to height/width of 1px and with a black background colour to create the horizontal and vertical divider lines. You can see from the item hierarchy in the image below, how I laid out my dashboard (I like to rename my containers to help understanding)

Now add a dashboard change set values action for each of the 4 bar chart sheets.

Select Region

On select of the Region sheet only, target the Region Set. On running the action (ie clicking the bar), assign values to set, and when clearing the selection (clicking the bar again), add all values to the set.

Note – While not specified in the requirements, I noticed that the breadcrumbing functionality in Kyle’s solution didn’t behave if multiple selections of the same dimension were made – eg 2 regions were selected. I decided to add the requirement of only allowing a single dimension to be clicked (ie the single-select only box is checked).

Create a Select Segment, Select Ship Mode and Select Sub-Category set action using the same principals described above.

Creating the breadcrumb

I’ve added this last, so you understand how we can ensure each set only has either all the values in it, or just 1 value.

To create the breadcrumb, we’re going to build up some strings based on what the state of each set looks like. This involved several calculated fields…. I’m not sure if I’ve over complicated this though..

Anyway firstly, we want to capture the values that have been added to each set, so we need

Regions in Set

IF [Region Set] THEN [Region] END

Segments in Set

IF [Segment Set] THEN [Segment] END

Ship Modes in Set

IF [Ship Mode Set] THEN [Ship Mode] END

SubCats in Set

IF [Sub-Category Set] THEN [Sub-Category] END

The image below shows how each of these fields are behaving based on the set selections – if the value is not selected in the set, the Regions in Set field is Null.

Next we have fields to count how many different values exist in each of these fields.

Count Selected Regions

{FIXED: COUNTD([Regions in Set])}

Count Selected Segments

{FIXED: COUNTD([Segments in Set])}

Count Selected Ship Modes

{FIXED: COUNTD([Ship Modes in Set])}

Count Selected SubCats

{FIXED: COUNTD([SubCats in Set])}

Again you can see from the sheet below, this is counting the number of selections, which is ‘fixed’ (ie the same) for every row.

Now, while this is showing 2, as we’ve manually clicked on the set options, in practice when driven from the dashboard, we’re either going to have all values in the set, or just 1. So based on this assumption, we now just want to get the name of the single selection

Selected Region

IF SUM([Count Selected Regions]) = 1 THEN MAX([Regions in Set]) ELSE ” END

If there’s only 1 item in the set, then get it’s value, otherwise return ‘blank’.

Just testing this behaviour, we can see below that with all the Regions selected, the Selected Region field is empty, but with 1 value selected, we show that value.

Create equivalent fields for each dimension

Selected Segment

IF SUM([Count Selected Segments]) = 1 THEN MAX([Segments in Set]) ELSE ” END

Selected Ship Mode

IF SUM([Count Selected Ship Modes]) = 1 THEN MAX([Ship Modes in Set]) ELSE ” END

Selected SubCat

IF SUM([Count Selected SubCats]) = 1 THEN MAX([SubCats in Set]) ELSE ” END

The order of the dimensions displayed in the breadcrumb is fixed, regardless of the order in which you click the options. That is, if you click a Segment then a Region, the breadcrumb will display the <segment> followed by the <region>. But if you click the Region first and then the Segment, the breadcrumb will still display the<segment> followed by the <region>. Based on this, we can create string values for each dimension that differ depending on whether we know there is a selection made against a subsequent dimension (ie should we include the ‘>’ character or not).

Let’s go through in order. Firstly, no selections made

All Segmentations BC

IF [Selected Segment]=” AND [Selected Ship Mode]=” AND [Selected Region]=” AND [Selected SubCat]=” THEN ‘All Segmentations’ END

If all the ‘selected’ values are empty, then all the sets contain all the values, so display ‘All Segmentations’.

If there are selections made, then the dimensions are ordered as Segment > Ship Mode > Region > Sub-Category

Segment BC

IF [Selected Segment]<>” AND ([Selected Ship Mode]<>” OR [Selected Region]<>” OR [Selected SubCat]<>”) THEN [Selected Segment] + ‘ > ‘
ELSE [Selected Segment]
END

If there is only 1 Segment selected and at least 1 of the other dimensions has been selected too, then add the ‘>’ character after the Segment name, otherwise just show the Segment.

Ship Mode BC

IF [Selected Ship Mode]<>” AND ([Selected Region]<>” OR [Selected SubCat]<>”) THEN [Selected Ship Mode] + ‘ > ‘
ELSE [Selected Ship Mode]
END

Similar to above, but this time, we only need to compare with the dimensions that are below Ship Mode in the display hierarchy.

Region BC

IF [Selected Region]<>” AND [Selected SubCat]<>” THEN [Selected Region] + ‘ > ‘
ELSE [Selected Region]
END

There is only one dimension below Region. As Sub-Category is at the bottom of the ordering, we don’t need anything special – the value of the Selected SubCat field will do.

On a new sheet, add All Segmentations BC, Segment BC, Ship Mode BC, Region BC and Selected SubCat to the Text shelf. Change the mark type to shape and change to use a transparent shape.

Adjust the label, so all the fields are ordered correctly and positioned exactly next to each otherwith no spacing/carriage returns between. Align the label middle left.

Show the set controls, and then test the functionality by altering the selections, ensuring either only 1 value or all values are selected

Once you’ve finished testing, ensure all values are selected in all sets.

The add this sheet to the dashboard – I had the title and the breadcrumb in a vertical container, which was the left hand side of a horizontal container

And hopefully that should be it. My published viz is here.

Happy vizzin’!

Donna

Can you build a flexible tabular display?

The challenge this week was to provide a balance between the user requesting ‘all the numbers’ and the visualisation designer trying to encourage ‘best practice’. It was inspired by Sam Parsons‘ viz here, which he discusses in this video.

We’re going to use 2 sheets to build this, one to display the text and one to display the bars. But first we need to define some parameters for the user to drive the display

pDisplayType

string parameter containing a list of two options, and defaulted to ‘Bars’

pShowExpanded

boolean parameter, defaulted to False

The measures we’ll be displaying are the Sales in 1000s and the Profit Ratio percentage, so we’ll need

Sales (k)

[Sales]/1000

format to 1 dp

and

Profit Ratio %

(SUM([Profit])/SUM([Sales]))*100

custom format this to ▲ #,##0.0;▼ #,##0.0 (you can use this site to copy & paste the symbols required).

Let’s start to build out the tabular view.

Add Category and Sub-Category to Rows. Add Region and Order Date (as a blue, discrete pill, at the YEAR level) to Columns. Add Measure Names to Filter and filter to Sales (k) and Profit Ratio %. Add Measure Values to Text. Add Measure Names to Columns in front of the Order Date field. Swap the order of the pills in the Measure Values box.

Add subtotals (Analysis > Add all Subtotals) and set Row Totals to Left and Column Totals to Top. Add Column Grand Totals (Analysis > Show Column Grand Totals).

Set the totals, so they are displaying the average values (Analysis > Totals > Total All Using > Average)

Right click on the label ‘Grand Total’ in the table, and format. In the left hand pane, change the label from ‘Grand Total’ to ‘All Categories’

Right click on the ‘Total’ label in the columns and format and change the Totals label to ‘Avg per Yr’

Now we have the content of the table, we need to add the functionality to expand/collapse (hide or show) the Year columns based on the pShowExpanded parameter. For this we need

Show / Hide Columns

IF [pShowExpanded] THEN ‘Show’ ELSE ‘Hide’ END

Add pShowExpanded to the view, and add Show/Hide Columns to Rows.

You should see that the value for Show/Hide Columns is only visible against the Year columns and not against the sub totals. With pShowExpanded set to False the value displayed is ‘Hide’

Now right click on the text ‘Hide’ and select the Hide option from the context menu

The Year columns will disappear. Change pShowExpanded to True, and voila! the columns are displayed. Uncheck Show Header against the Show/Hide Columns so the text ‘Show’ is not visible when expanded.

The ‘hide’ option is a really sneaky trick, but so very effective at not displaying specific rows/columns without filtering them out. It isn’t obvious it’s been applied though (there’s no clues in the canvas), so as a top tip, I recommend always adding a note to the worksheet caption to remind your future self (or a colleague who needs to maintain the workbook), that this has been applied.

Formatting the table

Now the functionality has been nailed, we need to apply the formatting & styling to the table.

I set the fonts to Tableau Medium, colour #7d4e47 throughout. The ‘All Categories’ Grand Total label was 11pt, the Category labels 9pt and the Sub-Category labels 8pt. The Region labels were 11pt, the Measure Names labels were 9pt and the ‘Avg per Yr’ and YEAR(Order Date) labels all 8pt. The Text of the numbers in the main body of the table were all set to 8pt too.

The background of the workbook was set to cream: #fefaf1

and then the total and grand total rows were coloured #faf4e7 and #f5eddc accordingly

To colour the font, add Measure Values to Colour, and right click and select Use Separate Legends

This will then give you two colour legends.

Adjust the Sales (k) colour legend – select any diverging colour palette, set Stepped Colour to 2 steps, and check Include Totals, and then click on the colours at each end and set both to the same brown (#7d4e47).

Repeat similar steps for the Profit Ratio % colour legend, but this time also set the range to be centred at 0, and apply different colours to the ends (#ff6f6a and #7d4e47).

Remove the Category & Sub-Category and Region/Order Date labels from displaying (hide field labels for columns/rows), and adjust the width and height of the cells to give a bit more spacing. Hide the tooltips from showing.

Create the bar table

Duplicate the sheet. Add Measure Values to Columns, and the click on the Label shelf and uncheck Show Mark Labels. Click on the Colour shelf and change the opacity to 35%. Reduce the Size of the bar to suit, then hide the Measure Values axis (uncheck Show Header).

Click on the Tooltip shelf and enable tooltips, then adjust accordingly.

Re-apply any of the formatting to the text that may have been lost – for some reason I lost all the Tableau Medium style fonts…???

Swapping the sheets on the dashboard

Create a dashboard, add a vertical layout container and add both sheets.

We will make use of dynamic zone visibility to only display the sheet we want based on the value of pDisplayType. To do this, we need some more calculated fields

Show Bar

[pDisplayType] = ‘Bars’

Show Numbers

[pDisplayType] = ‘Numbers’

Go back to the dashboard. Add the pDisplayType parameter. Click on the Bar sheet so it is selected (grey box) and on the Layout tab, check the Control visibility using.. checkbox and choose the Show Bar field.

Repeat the same steps for the Text table, but select the Show Numbers field instead. Depending what value is displayed in the pDisplayType parameter, one of the sheets should disappear, and changing the pDisplayType value should swap between them.

Now, style the dashboard as required.

My published viz is here.

Happy vizzin’!

Donna

Let’s Analyse New Product Sales

For the challenge this week, Sean wanted us to identify the sales of ‘new’ products compared to previous years. This challenge was born out of a requirement he had at work, where he discovered the definition of a ‘new product’ was different to what he’d assumed.

In this instance, the ‘client’ only cared about comparing sales in the same month across the years. Any sales related to the products outside of the specific month were irrelevant and were to be ignored. A product was then counted as ‘new’ in the first year (of the specified month) that there was a sale.

In this case we are always looking at the month of September (the previous month to ‘today), so if Product X was sold in Sept 2021 and Sept 2023 only, then Product X would count as new in Sept 2021, and only the sales for Product X in Sept 2021 would be included in the subsequent values displayed.

Now I thought I’d nailed my understanding of the requirement, and went ahead building out a tabular view of the data using table calculations, but my final output did not match Sean’s result – in fact it was quite way off. I stepped through my logic, even sharing some samples with Sean to sense check, and we seemed to concur on the products identified as new, so I was incredibly puzzled as to what was causing the value discrepancy.

I messaged my friend, Rosario Gauna, to share my numbers and see what she came up with…. she matched Sean! Doh! I then had to spend time dissecting my solution and Rosario’s to see which products I was/wasn’t counting. Eventually I found the problem….

In using table calculations, I had included all the fields the viz required which included the Segment (Home Office, Corporate, Consumer). This meant that in my logic to identify a ‘new product’, I was counting the product as new in the first year that the product was sold in each Segment – ie if Product X sold in Sept 2022 against Segment Home Office and then sold in Sept 2023 against Segment Corporate, I counted both sales as being ‘new’ as the segment differed. However I only wanted the Sept 2022 sale to count as ‘new’.

With this understood, I tried to revise my initial table calc build, but to no avail. So I opened up a new workbook and started again, this time taking a different track. This is what I’ll blog, but I felt it important to explain how even a requirement I thought I’d understood, could still get mis-interpreted.

Building out the calculations

The first requirement is that we’re only looking at comparing the data for the previous month to ‘today’. If this was being developed for a live business application we’d reference the TODAY() function, but since I want this viz to continue to display on Tableau Public after we hit 2024, then I’m going to use a parameter to hardcode ‘today’.

pToday

date parameter defaulted to 4th October 2023

From this, I can then determine the month we need to compare

Month to Compare

DATEPART(‘month’, DATEADD(‘month’, -1,DATETRUNC(‘month’, [pToday])))

working from right to left (inside to out), this takes ‘today’ and finds the 1st of the month (ie 1st October), then goes back 1 month (ie 1st September), then gets the month number ie 9.

As this returns a number, it automatically is listed in the ‘measures’ section of the data pane (under the line), but I dragged it into the top half as I want the number to be discrete.

I then created

Filter Month

DATEPART(‘month’, [Order Date]) = [Month to Compare]

which returns true for all the orders where the month of the Order Date is in September.

Let’s build up a table so we can start to see what we’re working with

Add Product Name and Filter Month to Rows, Order Date (at the Year level) to Columns and Sales to Text.

In the example above, 3-ring staple pack has sales in September 2022 and in other months in 2020 and 2021, but we don’t care about those months, so this product is counted as new in 2022, with a value of $11.

3M Hangers With… have no sales in September in any year, so we won’t be counting that product at all.

6″ Cubicle Wall Clock has sales in both September 2022 and 2023, so this product will be counted as new in 2022 with a value of $83.

So we can filter the date just to Filter Month = True (move Filter Month from Rows to the Filter shelf).

Now we want to identify the earliest year for each Product Name.

Year First Purchased in Month

{FIXED [Product Name],[Filter Month]: MIN(YEAR([Order Date]))}

Add this onto Rows, and you can see below we’re picking up the right year.

Now we can identify the value of the sales we want to count.

New Product Sales

IF [Year First Purchased in Month] = YEAR([Order Date]) THEN [Sales] END

Only get Sales for the year that matches the minimum year. Format this field to $ with 0 dp.

Adding this in to the table, we can see we only have a value against the first year.

So we’ve identified the sales values of the products we care about. We can now aggregate this at a higher level, and incorporate the other dimensions.

Remove Sales from Text, add Category to Rows and Segment to Columns. Remove Product Name and Year First Purchased in Month.

The viz displays the % of total new product sales by Category per Segment. Right click on the New Product Sales pill and add Quick Table Calculation, selecting Percent of Total. Right click on the pill again and select Compute Using -> Category. Right click on the pill and Format and format to % with 0dp. Add another instance of New Product Sales back into the table.

Now we want to be able to compare the sales for the current year/category/segment against that of the previous year.

Previous Year Sales

LOOKUP(SUM([New Product Sales]),-1)

This looks up the New Product Sales value of the previous (-1) field. Add this field into the table, and right click and Edit Table Calculation. Adjust so the calculation is computing by Year of Order Date only.

You should be able to see that the values from the previous year for each Segment & Category are now displayed against each cell. As there is no data for 2019, there are no values listed for this field against the 2020 fields.

Now we have the values we want to compare in the same ‘row’ of data, we can compute

% Diff From Previous Year

(SUM([New Product Sales]) – [Previous Year Sales]) / [Previous Year Sales]

custom format this to ▲0%;▼0% (use this site to get the images to copy & paste), and add this into the table ensuring the table calculation is set to compute using Year of Order Date only, as you did above.

And now we have all the components needed to build the viz.

Building the bar chart

Duplicate the table sheet, then apply the following steps

  • Move % of Total New Product Sales to Rows
  • Move Category to Colour and adjust accordingly
  • Move New Product Sales and % Diff from Previous Year to Label
  • Remove Previous Year Sales
  • Add another instance of % of Total New Product Sales to Label ( I tend to do this by pressing ctrl, and then clicking on the pill in the Rows and dragging onto Text – this creates a duplicate of the pill and preserves the table calculation settings. Otherwise add an instance of New Product Sales to Label, then add the Percent of Total Quick Table Calculation and ensure the pill is set to compute by Category.)
  • Move Segment to be in front of the Year(Order Date) field on Columns – if this changes the viz display, ensure the mark type is set to bar.

Tidy up the display by

  • removing all gridlines/zero lines etc
  • remove row dividers
  • Adjust title of the y-axis
  • Hide the Segment/Order Date column heading
  • Hide the tooltips

The text and formatting of the labels also needs to be adjusted. We don’t want the ‘vs PY’ text displaying for the 2020 years when there is no difference from previous year.

Label PY Text

IF NOT ISNULL([Previous Year Sales]) THEN ‘vs PY’ END

Add this to the Label shelf and ensure the table calculation is set to compute using YEAR of Order Date only. Then adjust the layout of the label text and the format the size of the text.

Now you can just add to a dashboard!

Happy vizzin’!

Donna

Target Achievement Report

The #WOW2023 challenge this week was set by guest challenger Valerija Kirjackaja, who asked us to recreate a tabular view of data.

Model the data

The requirement involved both the use of the Superstore data set and a custom Target data set, so these need to be combined. In the data pane, relate the two data sets together applying the fields below in the screen shots as the relationship fields (not you’ll need to create a relationship calculation to build the 3rd relationship.

Building the table

The table displays Category & Sub-Category along with a 3rd dimension which will differ depending on user selection. So we need to enable this choice.

pBreakdown

string parameter containing a list of options, defaulted to Ship Mode

Then create a calculated field to determine the actual field to show based on the parameter selection

Breakdown Dimension

CASE [pBreakdown]
WHEN ‘Segment’ THEN [Segment]
WHEN ‘Region’ THEN [Region]
WHEN ‘Ship Mode’ THEN [Ship Mode]
END

The user will also need to select a month. I chose to use a calculated field and parameter to drive this.

Month

DATE(DATETRUNC(‘month’, [Order Date]))

and this then feeds the parameter

pMonth

date parameter, formatted to custom date format of mmmm yyyy (to display September 2023). The list is populated when the workbook opens from the Month calculated field created above. Default date is 01 Sept 2023.

The report will need to be filtered based on the date selected in the parameter, so create

Filter Date

[Month] = [pMonth]

On a new sheet, add Category & Sub-Category (from the Orders data set) and Breakdown Dimension to Rows. Show the parameters. Add Filter Date to the Filter shelf and set to True.

Now, the table shows some conditional formatting, which is a clue that we’re not going to building this table in the basic way. Instead we’ll be using what I refer to as ‘fake axis’ to define placeholder columns, which we can then use to have more flexibility on how the data is displayed.

Double click into the Columns shelf and type MIN(0.0).

Change the mark type to Shape and use a transparent shape (see this blog for more information on doing this). Add Sales to the Label shelf and align left. Add all subtotals (Analysis > Totals > Add All Subtotals) and then set the all to display at the top (Analysis > Totals > Column Totals to Top).

Adjust the formatting, so the row dividers are at the highest level

and the row banding is also at the total pane/header level

Type in another instance of MIN(0.0). This will create a second marks card. Adjust the Sales pill and apply a quick table calculation of percent of total. Set to compute using the Breakdown Dimension field. Format the field to % with 0dp.

To stop the % displaying on the Total rows, click the table calc field and select Total using > Hide

Add another instance of MIN(0.0) and this time add Sales Target value from the Sheet1 data set to the Label shelf.

You can see the target values fill down against the Breakdown Display rows, but we don’t want this.

So to work out whether we’re on a Total row or not, we’re going to make use of the SIZE() table calculation.

We want to count the number of Breakdown Dimension rows being shown.

Count Breakdown Dimension Rows

SIZE()

change this to be discrete, add this to Rows and adjust the table calculation so it is compute using the Breakdown Dimension field. You can see the value against each row represents the number of (in this case) different Ship Mode values.

What isn’t visible is the fact that the SIZE() values against the Total rows are actually 1. Move the Count Breakdown Dimension Rows to the Tooltip shelf on the All marks card. If you now hover over the rows of data, the tooltip should display 1 against this field when it’s a total row and the count otherwise. We’ll use this to identify the Total rows.

This is a common technique that is used. Note though, that if the ‘variable’ you’re counting over only has 1 value, then any logic based on this will apply to that row too (see the Copiers row for Ship Mode & Sept 2023).

So to display the Target Sales as we want, we need

Sales Target to Display

IF [Count Breakdown Dimension Rows] = 1 THEN SUM([Sales Target]) END

ie only show the value of the target on the total rows (or when there’s only 1 value in the dimension being counted).

Replace the Sales Target pill on the 3rd MIN(0.0) marks card with this Sales Target to Display field. Make sure the table calculation is set to compute using Breakdown Dimension

Now we need the variance

Variance

(SUM(Sales)-[Sales Target to Display])/SUM([Sales])

Format this to ▲0.0%;▼0.0%

and we need to identify if it’s +ve or -ve

Variance is +ve

[Variance] >=0

Create another instance of MIN(0.0). Add the Variance field to the Label shelf and verify it is computing by Breakdown Dimension

Add Variance is +ve to the Colour shelf. Adjust the colours, then on the Label shelf, set the font to match mark colour.

To change the ‘Total’ labels to ‘All’, right click on one of the labels, and change the label in the left hand pane.

Finally, adjust the formatting – I set the font throughout to be a darker colour, set the All labels to be bold, removed all gridlines, zero lines and column dividers, hid the MIN(0) axes and also hide field labels for rows. Turn off tooltips for the whole table.

Building the dashboard

I then added the sheet to a dashboard, and placed it within a vertical container. Above the vertical container, I used a horizontal container in which I added text boxes for the column headers. The text box for the variable dimension, referenced the pBreakdown parameter value. I also then used a floating blank set to a height of 2px with a background colour of grey to give the appearance of a line above the column headings.

In the dashboard heading, refer to the pMonth parameter to display the date.

Once published, I did have to tweak the width of the heading text boxes and the position of the floating line within the web edit view.

My published viz is here.

I essentially ‘faked’ the table header row but I still only used 1 sheet to build the table and I was able to dynamically change the title of the dimension column, so hey I hit the brief 🙂 If you want to understand how to build the header row into the actual table itself, then check out Rosario Gauna’s blog.

Happy vizzin’!

Donna

Can you build a drunken tree chart?

This week, Luke set a challenge he’s had in his back pocket since he first joined as a WOW coach. He did provide several good clues within the requirements to help build the chart.

The challenge involves some data modelling (unioning 2 instances of the Superstore data set together) and what I refer to as ‘normalising’ of dates – to get data spread across multiple years to set to the same year.

I have to admit, I’ve had a busy week so far, attending a conferences and dealing with some personal matters, that I feel I’m going to struggle to get a thorough solution guide documented in a timely manner – it won’t be long before we’re on to week 39….

So for this week, I’m going to direct you to get help from my fellow #WOW participator and Visionary, the most excellent Rosario Gauna, who has already published her solution guide : English | Spanish

Our approaches were very similar – completed in a single sheet, though, as if often the case, Rosario’s solution is far more elegant than mine!

My published viz is here.

Happy vizzin’!

Donna

Can you show placeholders for a maximum of 6 goals?

It was Lorna’s turn to set the challenge this week, based on a real world scenario she’s encountered with a colleague. The premise was to be able to show the progress, by country, against a maximum of 6 goals selected by the user. Placeholders for the 6 options had to remain visible at all times (unless the user selected more than 6, in which case a message should appear).

Building the core viz

As Lorna alludes to in the requirements, we will use sets to capture the goals the user can select.

SDG Name Set

Right click on SDG Name > Create > Set. Select up to 4 options.

On a new sheet, add SDG Name Set and SDG Name to Columns, Country to Rows and add Country to the filter shelf, and limit to Australia, Finland, UK & USA. Right click on the SDG Name Set field in the data pane and select Show Set

We need the SDG Name value to only display for the values selected

SDG Name Header Label

IF [SDG Name Set] THEN [SDG Name] ELSE ” END

Add this to Columns before the SDG Name field.

Now we need to always display 6 columns of data, ie in this case, all the SDG Name values In the set and the first SDG Name values not in the set. We will use the INDEX() function to help us label the columns position.

INDEX

INDEX()

Right click this field and Convert to discrete, then add it to the Columns after SDG Name.

Edit the table calculation (click on the triangle symbol on the blue INDEX pill), and adjust so it is computing by Specific Dimensions. This should be all the fields except Country. The columns should be labelled sequentially from 1 to 17.

Add INDEX to the Filter shelf as well. Initially just select 1. Then adjust the table calculation of this field to match above, and once done, edit filter and select values 1 through to 6. This should leave you with 6 columns

Now we have the structure, we can start building the core contents of the table. For this we’ll be using what I refer to as ‘fake axis’.

Double click in the Rows shelf and manually type in MIN(0.2), then double click again and manually type in MIN(0.6). This results in the creation of a MIN(0.2) and a MIN(0.6) marks card on the left hand side.

The MIN(0.2) marks card is going to be used to show the information about how the goal is trending (ie the arrow symbol), while the MIN(0.6) marks card will be used to show the status of the goal (the text displayed). We need new fields for this, so the values only display for the selected goals.

Trend

IF [SDG Name Set] THEN [SDG Trend] END

Goal

IF [SDG Name Set] THEN [SDG Value] END

Click on the MIN(0.2) marks card. Change the mark type to shape. Add Trend to the Shape card. Right click on the Trend pill and change the field from a Dimension to an Attribute.

Doing this stops the field from impacting the table calculation, and you should get back to having 6 columns displayed.

Adjust the shapes using the Arrows shape palette. For the Null value, I set it to use a transparent shape, a custom shape added to my shape palette. See this blog for more information on doing this.

Also add Trend to the Colour shelf. Once again, adjust the pill to be Attribute and then adjust the colours.

Now click on the MIN(0.6) marks card. Adjust the Mark Type to be Text. Add Goal to the Text shelf and change to be and Attribute. Add Goal to the Colour shelf, change to be Attribute and adjust colours to suit. I also adjusted the font to be bold & size 10pt.

Set the chart to be dual axis and synchronise the axis. Edit the axis (Right click on the left hand axis) and fix the axis from 0 to 1.

Hide the axis, the IN/OUT SDG Name Set pill, the SDG Name and INDEX pills (right click the pills and uncheck Show header).

Right click on the Country row label and the SDG Name Header Label column label in the viz and hide field labels for rows/columns.

Right click within the viz to format. Set the background colour of the pane to light grey.

Remove all gridlines, axis rules, zero lines. Set the column and row dividers to thick white lines

Click the Tooltip button on the All marks card, and uncheck show tooltips. Set the viz to Fit width.

Building the Goal legend

On a new sheet, add SDG Value to Columns. Change the mark type to Circle then add SDG Value to Label and Colour. Manually re-order the columns and adjust the colours as required.

Double click in the Columns shelf and type in MIN(0.0). Edit the axis and fix from -0.1 to 0.5 – this will shift the symbols and text to the left.

Adjust the size of the circle shape to suit, and set the font of the label to match mark colour. I also set it to be bold.

Remove all gridlines, axis ruler, zero lines. Set the background of the pane to be grey and the row/column dividers to be thick white lines. Hide the axis and the column headers, and uncheck show tooltips.

Double click into the Rows shelf, and type the text ‘Goal’ (including the quotation marks). Hide the ‘Goal’ label that then displays.

Building the Trend legend

Repeat similar steps for above but add the SDG Trend field to the Colour, Label and Shape shelf. Adjust the shape & colours to those used before.

Handling more than 6 selections

For this requirement, we need to determine the number of items in the set.

Count Set Members

{COUNTD(IF [SDG Name Set] THEN [SDG Name] END)}

and then use this to create some boolean fields

More than 6 Selected

[Count Set Members] > 6

Less than 6 Selected

[Count Set Members] <= 6

Ensuring that less than 6 items are selected, add the Less than 6 Selected field to the Filter shelf of the main table viz and set to True.

If you select more than 6 goals, the viz should disappear.

On a new sheet, double click into the space beneath the marks card where the pills usually sit, and type ‘Dummy’ (with quotes). Change the mark type to shape and set to use the transparent custom shape. Move the Dummy pill to the label shelf, then edit the label and change the text to the error message.

Align the text middle centre and fit to entire view. Uncheck show tooltips. Add More than 6 Selected to the filter shelf and select true (if true isn’t an option, go back to the main viz, and select more options sp the viz disappears, then come back to this sheet and try again).

All the sheets can now be added to the dashboard. Ensure the core table viz and the error sheets are added to a vertical container without the title showings – the charts should expand and collapse as the selections are made – this can be a bit tricky to get right.

My published viz is here.

Happy vizzin’!

Donna

Can you build a dashboard for pre-aggregated metrics?

For the challenge this week, Kyle asked us to recreate the visualisation above using an adapted version of Superstore which had a customer count metric for 3 dimensions (Category, Segment and Region) along with ‘no’ dimension (null) pre-aggregated at a Yearly or Monthly Level.

By this I mean that, at a Yearly level, when the date was 1st Jan 2019 say, a row of data existed for the (distinct) customer count of all the combinations of the 3 dimensions and null. In total 80 rows for the one date.

As the data was pre-aggregated, it made no sense to say the customer count for Technology is the sum of all the rows where Category = Technology and this would mean data was being double counted.

Pivoting the data also wouldn’t yield the desired result. So the aim of this challenge was to be able to identify the relevant rows of data that needed to be displayed based on the options selected by the user.

Building the calculations

Parameters will be driving the user selections, so these need to be set up

pDateGrain

string parameter with a list of 2 options: Monthly and Yearly. Defaulted to Monthly.

pColour

string parameter with a list of 4 options : Category, Region, Segment, None. Defaulted to Segment

Similarly, create pXAxis and pYAxis parameters similar to above, but default both to None.

On a new sheet build a tabular view with

  • Table Names, Category, Segment and Region on Rows
  • Order Date set to discrete (blue pill) exact date on Columns
  • Customer Count on Text
  • Show all 4 parameters created

The rows of data need to be filtered by Table Name (as defined by the pDateGrain parameter) and a combination of Category, Segment and Region based on the options selected in the other 3 parameters.

To filter by the Table Name we need

Filter – Date Grain

[pDateGrain] = [Table Names]

Add this to the Filter shelf and set to True.

Change the pDateGrain parameter to Yearly as there is less data to see/check.

Based on the options selected in any of the other 3 parameters, we need to find matching rows.

For example, if pColour is Segment and the other parameters are None, we are looking for the rows where the Segment column is not null, but the Region and Category columns are (we would be after the same rows if pXAxis was set to Segment, and the other parameters were None or, if pYAxis is Segment and the other parameters were None).

In this case, we’re looking for 3 rows of data – those highlighted below

If instead any two of the parameters were set to Segment and Category and the other None, then we’d be looking for rows where Segment is not null, Category is not null and Region is null. This would be 9 rows in total (a snippet of which is shown below).

We also need to deal with scenarios where all three parameters were set to something different, or all set to None as well as handle if multiple parameters are set to the same thing.

Now to do this, I ended up building a single field to use as filter that contains all the scenarios. As I was building it up, I figured there should be a slicker way, and there is (check out Kyle’s solution), but if your brain is wired the same way as mine, then you’ll end up with this

Filter Rows to Include

IF [pColour] = ‘None’ AND [pXAxis] = ‘None’ AND [pYAxis] = ‘None’ THEN
//no options selected
IF ISNULL([Region]) AND ISNULL([Category]) AND ISNULL([Segment]) THEN TRUE END
ELSEIF (([pColour] <> ‘None’ AND [pXAxis] = ‘None’ AND [pYAxis] = ‘None’) OR
([pColour] = ‘None’ AND [pXAxis] <> ‘None’ AND [pYAxis] = ‘None’) OR
([pColour] = ‘None’ AND [pXAxis] = ‘None’ AND [pYAxis] <> ‘None’)) THEN
// one of the 3 options selected, so work out which dimension
IF [pColour] = ‘Category’ OR [pXAxis] = ‘Category’ OR [pYAxis] = ‘Category’ THEN
IF ISNULL([Region]) AND ISNULL([Segment]) AND NOT ISNULL([Category]) THEN TRUE END
ELSEIF [pColour] = ‘Segment’ OR [pXAxis] = ‘Segment’ OR [pYAxis] = ‘Segment’ THEN
IF ISNULL([Region]) AND NOT ISNULL([Segment]) AND ISNULL([Category]) THEN TRUE END
ELSEIF [pColour] = ‘Region’ OR [pXAxis] = ‘Region’ OR [pYAxis] = ‘Region’ THEN
IF NOT ISNULL([Region]) AND ISNULL([Segment]) AND ISNULL([Category]) THEN TRUE END
END
ELSEIF (([pColour] <> ‘None’ AND [pXAxis] <> ‘None’ AND [pYAxis] = ‘None’) OR
([pColour] <> ‘None’ AND [pXAxis] = ‘None’ AND [pYAxis] <> ‘None’) OR
([pColour] = ‘None’ AND [pXAxis] <> ‘None’ AND [pYAxis] <> ‘None’)) THEN
// two options selected, so work out which dimensions we need
IF ([pColour] = ‘Category’ OR [pXAxis] = ‘Category’ OR [pYAxis] = ‘Category’) AND ([pColour] = ‘Segment’ OR [pXAxis] = ‘Segment’ OR [pYAxis] = ‘Segment’) THEN
IF ISNULL([Region]) AND NOT ISNULL([Segment]) AND NOT ISNULL([Category]) THEN TRUE END
ELSEIF ([pColour] = ‘Category’ OR [pXAxis] = ‘Category’ OR [pYAxis] = ‘Category’) AND ([pColour] = ‘Region’ OR [pXAxis] = ‘Region’ OR [pYAxis] = ‘Region’) THEN
IF NOT ISNULL([Region]) AND ISNULL([Segment]) AND NOT ISNULL([Category]) THEN TRUE END
ELSEIF ([pColour] = ‘Segment’ OR [pXAxis] = ‘Segment’ OR [pYAxis] = ‘Segment’) AND ([pColour] = ‘Region’ OR [pXAxis] = ‘Region’ OR [pYAxis] = ‘Region’) THEN
IF NOT ISNULL([Region]) AND NOT ISNULL([Segment]) AND ISNULL([Category]) THEN TRUE END
//or the two options selected are the same dimension
ELSEIF ([pColour] = ‘Category’ OR [pXAxis] = ‘Category’ OR [pYAxis] = ‘Category’) AND ([pColour] = ‘Category’ OR [pXAxis] = ‘Category’ OR [pYAxis] = ‘Category’) THEN
IF ISNULL([Region]) AND ISNULL([Segment]) AND NOT ISNULL([Category]) THEN TRUE END
ELSEIF ([pColour] = ‘Segment’ OR [pXAxis] = ‘Segment’ OR [pYAxis] = ‘Segment’) AND ([pColour] = ‘Segment’ OR [pXAxis] = ‘Segment’ OR [pYAxis] = ‘Segment’) THEN
IF ISNULL([Region]) AND NOT ISNULL([Segment]) AND ISNULL([Category]) THEN TRUE END
ELSEIF ([pColour] = ‘Region’ OR [pXAxis] = ‘Region’ OR [pYAxis] = ‘Region’) AND ([pColour] = ‘Region’ OR [pXAxis] = ‘Region’ OR [pYAxis] = ‘Region’) THEN
IF NOT ISNULL([Region]) AND ISNULL([Segment]) AND ISNULL([Category]) THEN TRUE END
END
ELSE //all three selected, but they could be all the same dimension or 2 of the three the same
IF ([pColour] = ‘Category’ OR [pXAxis] = ‘Category’ OR [pYAxis] = ‘Category’) AND ([pColour] = ‘Segment’ OR [pXAxis] = ‘Segment’ OR [pYAxis] = ‘Segment’) AND ([pColour] = ‘Region’ OR [pXAxis] = ‘Region’ OR [pYAxis] = ‘Region’) THEN
//all three different
IF NOT ISNULL([Region]) AND NOT ISNULL([Segment]) AND NOT ISNULL([Category]) THEN TRUE END
ELSEIF ([pColour] = ‘Category’ OR [pXAxis] = ‘Category’ OR [pYAxis] = ‘Category’) AND ([pColour] = ‘Region’ OR [pXAxis] = ‘Region’ OR [pYAxis] = ‘Region’) THEN
IF NOT ISNULL([Region]) AND ISNULL([Segment]) AND NOT ISNULL([Category]) THEN TRUE END
ELSEIF ([pColour] = ‘Category’ OR [pXAxis] = ‘Category’ OR [pYAxis] = ‘Category’) AND ([pColour] = ‘Segment’ OR [pXAxis] = ‘Segment’ OR [pYAxis] = ‘Segment’) THEN
IF ISNULL([Region]) AND NOT ISNULL([Segment]) AND NOT ISNULL([Category]) THEN TRUE END
ELSEIF ([pColour] = ‘Region’OR [pXAxis] = ‘Region’ OR [pYAxis] = ‘Region’) AND ([pColour] = ‘Segment’ OR [pXAxis] = ‘Segment’ OR [pYAxis] = ‘Segment’) THEN
IF NOT ISNULL([Region]) AND NOT ISNULL([Segment]) AND ISNULL([Category]) THEN TRUE END
ELSEIF [pColour] = ‘Category’ OR [pXAxis] = ‘Category’ OR [pYAxis] = ‘Category’ THEN
IF ISNULL([Region]) AND ISNULL([Segment]) AND NOT ISNULL([Category]) THEN TRUE END
ELSEIF [pColour] = ‘Segment’ OR [pXAxis] = ‘Segment’ OR [pYAxis] = ‘Segment’ THEN
IF ISNULL([Region]) AND NOT ISNULL([Segment]) AND ISNULL([Category]) THEN TRUE END
ELSEIF [pColour] = ‘Region’ OR [pXAxis] = ‘Region’ OR [pYAxis] = ‘Region’ THEN
IF NOT ISNULL([Region]) AND ISNULL([Segment]) AND ISNULL([Category]) THEN TRUE END
END

END

Blimey! A bit monolithic I know, but it just grew organically as I tried out the different scenarios step by step. Unfortunately the above doesn’t copy over the formatting nicely, as there are nested (tabbed) IF statements which makes it (a bit) easier to read.

Suffice to say, I’m not going to walk through step by step, but it’s checking for all the different permutations are discussed above, and marking the relevant rows as True. This field can then be added to the Filter shelf and set to True.

Kyle’s solution, essentially replaces this one calculated field, with 3 calculated fields – 1 per parameter – which are all then added to the filter shelf. It’s much neater 🙂

So now we’ve identified the rows we want based on parameters, but there is also the ability to filter the rows further based on the values of the Category, Segment or Region.

Add each of the 3 fields to the Filter shelf and select the All option, then show the filters on the view. For each of the Category, Segment and Region filters, set the option to show Only Relevant Values. This will prevent the NULLs from showing as an option when the relevant dimension is listed as one of the parameter selections

As you can see from the above image though, Region is only showing Null, and this is because in the example above, Region isn’t selected as an option for the pColour, pXAxis or pYAxis parameters. When it comes to the dashboard, we don’t want the Region filter to be visible in this case. To help with this, we need 3 further calculated fields.

Show Filter – Region

[pColour] = ‘Region’ OR [pXAxis] = ‘Region’ OR [pYAxis] = ‘Region’

This returns True if one of the 3 parameters contains the value ‘Region’. Similarly, create Show Filter – Category and Show Filter – Segment fields.

The final calculated fields we need are to help build the ‘cross tab’ view.

X-Axis

CASE [pXAxis]
WHEN ‘Category’ THEN [Category]
WHEN ‘Region’ THEN [Region]
WHEN ‘Segment’ THEN [Segment]
ELSE ”
END

Y-Axis

CASE [pYAxis]
WHEN ‘Category’ THEN [Category]
WHEN ‘Region’ THEN [Region]
WHEN ‘Segment’ THEN [Segment]
ELSE ”
END

Colour

CASE [pColour]
WHEN ‘Category’ THEN [Category]
WHEN ‘Region’ THEN [Region]
WHEN ‘Segment’ THEN [Segment]
ELSE ”
END

Now we’ve got all the fields needed to build the viz.

Building the viz

The quickest way is to duplicate the sheet we’ve built, as all the filters need to apply, so

  • Duplicate the sheet
  • Remove all the fields from Rows
  • Change the Order Date field on Columns to be continuous (green pill)
  • Add X-Axis to Columns
  • Add Y-Axis to Rows
  • Move Customer Count to Rows
  • Add Colour to the Colour shelf.
    • Adjust the colours to suit.
    • Change the value of the option in the pColour parameter, and readjust the colours. Repeat so that colours are set for Category, Segment and Region.
  • Add Colour to the Label shelf

Remove all gridlines, axis and zero lines. Remove the Y-Axis and X-Axis row/column labels by right clicking the text and selecting Hide field labels for rows/columns. Edit the Order Date axis (right click axis -> Edit) and remove the axis title.

Add Order Date to Tooltip and format it to the ‘March 2001’ date format. Adjust the tooltip as below

Hiding the filters

Add the viz to a dashboard and arrange the parameters and filter controls in the relevant location. I used layout containers to help with the organisation.

Select the Category filter and on the Layout tab, select the Control visibility using value checkbox and select the Show Filter – Category field.

Repat the same steps for the Region and Segment filters, selecting the equivalent calculated fields.

And that should be it. My published viz is here.

Happy vizzin’!

Donna

Can you build an interactive Viz in Tooltip?

This week’s #WOW2023 challenge was inspired by Sam Parson’s TC presentation where he demonstrated the concept of an interactive Viz in Tooltip (the workbook he presented is here).

I was aware when I set this challenge, that this was likely to be on the higher end of the difficulty scale, but WOW challenges to me have have always provided a source of inspiration and ideas to take forward into my day job. And by blogging the solutions, I provide myself with a guide to refer to when the need arises. When Sam presented the concept, I immediately wanted to understand how he’d done it, and by setting it as a challenge it provided me with the opportunity to dig into it, and get that documented ‘how to’ guide 🙂

As mentioned in the requirements, I built this using multiple sheets, so we’ll start by just building out most of those sheets.

Building the Scatter Plot

We’re only concerned with data over the last 2 years, so we need to define some measures relevant to these years

Current Year

ZN(IF YEAR([Order Date]) = YEAR({MAX([Order Date])}) THEN [Sales] END)

{MAX([Order Date]} is a Fixed Level of Detail calculation which returns the maximum date in the data set. This calculation is then comparing the Year associated to that date with the Year of each Order Date, and if they match, return the Sales. Wrapping in a ZN ensures a value of 0 in the event there are no Sales.

Prior Year

ZN(IF YEAR([Order Date]) = YEAR({Max([Order Date])}) -1 THEN [Sales] END)

YEAR({Max([Order Date])}) -1 returns the year associated to the latest date then decrements by 1 to get the value of the previous year.

We also need to calculate the difference between the sales across the 2 years and categorise based on the difference

Sales Performance

IF SUM([Current Year]) / SUM([Prior Year]) > 1.1 THEN ‘Increasing’
ELSEIF SUM([Current Year]) / SUM([Prior Year]) < 0.9 THEN ‘Decreasing’
ELSE ‘Static’ END

If the current year sales > 10% of the previous year sales then flag as ‘increasing’, else if current month sales < 90% of the previous year sales then flag as ‘decreasing’ else flag as ‘static’.

Add Prior Year to Columns and Current Year to Rows. Add Manufacturer, Sub-Category and Category to Detail. Change the mark type to Circle. Add Sales Performance to Colour, adjust colours to match and reduce opacity to around 80%. Re-order the colour legend to display Increasing at the top and Decreasing at the bottom. Name the sheet Scatter.

Building the bar chart

On a new sheet add Order Date at the discrete (blue) Month level to Columns. Add Current Year to Rows. Change the mark type to bar and add Sales Performance to Colour. Reduce the opacity of the colour to 80%.

Add Prior Year to Rows. Change the mark type on the Prior Year marks card to gantt bar. Remove the Sales Performance pill from the colour shelf on this marks card. Adjust the colour to black.

Make the chart dual axis and synchronise axis.

Hide the axis, remove all gridlines & row/column dividers. Format the months to be abbreviated to the first letter. Right click on the Order Date label at the top and hide field labels for columns. Set the sheet to Entire View. Name the sheet Bar.

Building the KPI

On a new sheet add Current Year to Text. Adjust the format (size & colour of font) and align middle centre. Set the sheet to Entire View. Name the sheet KPI.

Building the % Change Indicator

Firstly we need to capture the value of the % change in sales

Sales Performance % Change

(SUM([Current Year]) – SUM([Prior Year]))/ SUM([Prior Year])

Use custom formatting to format as ▲ 0.0%;▼ 0.0%

On a new sheet, add Sales Performance % Change and Prior Year to Text. Change the mark type to square and increase the size to as large as possible. Set to Entire View. Adjust the font size of the text to match the display and align middle centre. Add Sales Performance to Colour. Name the sheet % Change

Identifying the data to filter

The final sheet we need is one to serve the title of the Viz in Tooltip (ViT). It includes references to information related to the mark selected on the scatter plot, namely the Category, the Sub-Category and the Maufacturer. Before building this title sheet though, we need to understand how we’re going to identify the mark selected so we can filter other sheets based on it.

Typically, for most use cases, when you add a worksheet to the tooltip of another sheet, you want to ‘filter’ what’s displayed in the ViT based on the mark you’re hovering/clicking on.

So by default, when you add a worksheet via the Viz in Tooltip functionality (see Tableau KB here for more info), the markup that’s automatically added looks like below

<Sheet name=”My ViT Sheet” maxwidth=”300″ maxheight=”300″ filter=”<All Fields>”>

where the filter property is set to <All Fields> which is the instruction to pass information from the ‘parent’ sheet through to the ViT sheet.

For this challenge however, when a user first hovers on a mark on the scatter plot (the parent sheet), the information displayed in the Viz in Tooltip (ViT) sheets is for the whole unfiltered data set (ie display information related to the Current Year and Prior Year Sales across all manufacturers, sub-categories & categories). But once selected (clicked on) the information displayed in the ViT should be filtered just to that mark.

For this to work, we can’t use the filter property of the ViT markup. That property needs to be set to “” to ensure the resulting sheets aren’t filtered ‘on hover’. So we need another way to drive the filtering behaviour.

We need to use sets.

We’re going to use sets to capture the Manufacturer, Sub-Category and Category of the mark that is clicked on. So to start, right click on Manufacturer > Create > Set and select all values.

Manufacturer Set

Repeat the same steps to create a Category Set and a Sub-Category Set.

Create a dashboard and add the Scatter sheet to the dashboard. Then create dashboard set actions (Dashboard Menu > Actions > Add Action > Change Set Values)

Select Manufacturer

On select of the Scatter sheet on the dashboard, target the Manufacturer Set, assigning values to the set when the action is initiated, and adding all values to the set when the action is cleared.

Note – ‘assign value to set’ will replace any values already in the set (ie all values) with the relevant value based on the selection made, whereas ‘add values to set’ just appends the selected value to the values already in the set.

Repeat the above steps to create set actions for the Category Set and the Sub-Category Set

Navigate to the Bar sheet. Add Category Set, Sub-Category Set and Manufacturer Set to the Filter shelf, and click on each pill and Show Set to list the sets and their selected values on the left hand side (this is just so you can see what’s going on).

Initially you can see all values in all the sets are selected. Now navigate back to the dashboard and click on a single mark. Then come back to the bar sheet and check the results…

You should see a change to the bars as they are now being filtered by only the values which have been assigned to the set via the ‘on click’ action.

Building the title sheet

Now we have the sets established, we can build on these to generate the information needed for the title sheet.

To start with, we need to understand how many values have been captured in each set.

Count Categories

COUNTD([Category])

Count Sub-Categories

COUNTD([Sub-Category])

Count Manufacturers

COUNTD([Manufacturer])

Then we need to build up a title based on what’s been selected

Title

IF [Count Manufacturers] = 1 THEN MIN([Manufacturer])
ELSEIF [Count Sub-Categories] = 1 THEN ‘All ‘ + MIN([Sub-Category])
ELSEIF [Count Categories] = 1 THEN ‘All ‘ + MIN([Category])
ELSE ‘All Manufacturers’
END

and a sub title

Sub Title

IF [Count Manufacturers] = 1 AND [Count Sub-Categories] = 1 AND [Count Categories] = 1 THEN ‘(‘ + MIN([Category]) + ‘ > ‘ + MIN([Sub-Category]) + ‘)’
ELSE ‘* ‘ + STR([Count Manufacturers]) + ‘ Manufacturers across ‘ + STR([Count Sub-Categories]) + ‘ Sub-Categories’
END

On a new sheet, add Title and Sub Title to Text. Format the background of the worksheet to be orange, set to Entire View, then adjust the text and font format and align top left. Name the sheet Title.

Now navigate back to the Bar sheet, and for each of the fields in the Filter shelf (the set ones), make them apply to selected worksheets, KPI, % Change, and Title

As a result, across all 4 sheets (not the Scatter one), you should have the 3 set fields as filters with the ‘multiple worksheet’ symbol indicating a shared filter.

If you go back to the dashboard and click on a mark then check the Title sheet, the information displayed should update.

Building the Viz In Tooltip

Now we have (most of) the components we need, let’s start to put together the actual ViT.

On the Scatter sheet, click on the Tooltip button to open the Edit Tooltip dialog.

Start by deleting all the text.

Then from the toolbar, click Insert > Sheets > Title to add the Title sheet to the tooltip. You should have something like

The key to getting the tooltip to display ‘nicely’ is to consider the height and widths, and align the markup text. Sometimes this does take a bit of trial & error and can also look differently when published to Tableau Public.

Adjust the above, so the maxwidth =500 and maxheight = 100, filter = “” and the whole line of text is centred.

Then add the other 3 sheets, using carriage returns to add space between the sheets as required, and adjusting the heights and widths.

If you go back to the dashboard and hover on a mark, you should see the display below for All Manufacturers

and if you then click on a mark, the display should adjust to filter

Making the ViT interactive

Edit the Tooltip on the Scatter sheet, and add a section at the bottom that references the Category , Sub-Category and Manufacturer fields (add via the Insert menu again). Style the font as you wish

Now if you go back to the dashboard and click on a mark, you can then also click on one of the links added at the bottom. In this instance I clicked on the Chairs link and all the marks in the scatter plot related to chairs were highlighted and the ViT data all updated to show the values associated to the Chairs Sub-Category

This is happening ‘automatically’ due to the fact the Allow selection by category option on the Tooltip is checked. This is a feature (along with Include command buttons) I personally often switch off.

Now ideally, we’d be finished at this point, but we just need to add a final feature, due to the fact that some Manufacturers exist across multiple Sub-Categories. For example, below while I have clicked a mark that is related to the Global Manufacturer in Chairs, clicking Global in the links at the bottom highlight all the Global Manufacturers across all Sub-Categories, so we can’t get back to seeing the information just about the selected mark.

Adding the ‘Current Mark’ selection

We need to capture the ‘product hierarchy’ for each mark into a single field

Category | Sub Cat | Manu

[Category] + ‘|’ + [Sub-Category] + ‘|’ + [Manufacturer]

Add this to the Detail shelf of the Scatter sheet.

We will need a parameter to then capture the ‘product hierarchy’ for the selected mark

pSelectedMarkIdentifier

string parameter set to ”” ie empty string

On the dashboard, add a parameter action

Identify Current Mark

On hover of the scatter sheet on the dashboard, set the pSelectedMarkIdentifier parameter to the value stored in the Category | Sub Cat | Manu field. Keep the current value when the selection is cleared.

Finally, we need to have a link to select in the tooltip, so we need

Current Mark Identifier

IF [pSelectedMarkIdentifier] = [Category | Sub Cat | Manu] THEN ‘Current Mark’
ELSE ”
END

Add this to the Detail shelf of the Scatter sheet, and then update the Tooltip and add a reference to the Current Mark Identifier field

If you now go back to the dashboard and test by clicking on a mark associated to the Global Manufacturer, you should be able to click on Current Mark using the link in tooltip after clicking other links, and get back to what you would have seen in the tooltip when you first clicked on the mark.

It should just now be a case of sorting out the layout on the dashboard.

Congratulations on getting this far! My published viz is here.

Happy vizzin’!

Donna