I promise that this will be my final post relating to the fifty-first week of #MakeoverMonday! When the earlier submissions rolled in, my eye was caught by three dashboards; those from Lindsey Poulter, David Krupp and @shivrajc. The final viz can be found here, and it is one that is the focus of this post.
As with the previous posts regarding visualisations by Lindsey, David (combined with Ashish Chauhan) and a bonus Michael Mixon article, I’ll aim to explore, articulate and take away tips relating to data preparation, data manipulation and presentation.
After a glance at the dashboard, there are a few things that I want to get to the bottom of:
- I want to confirm the calculations used to determine the Target met / Target not met shapes. They get done in different ways by everybody, so I want to see another interpretation
- I want to validate that the chart under the large metrics on the left is simply a bar-in-bar with a thin dark bar inside a larger, lighter bar
- I want to see the calculation which allows three steps of colouration to be applied to the Variance From Target Trend section
- Finally, I want to see how the three distinct sections in each “row” have been blended together so seamlessly. It looks great
After downloading the workbook and unhiding the hidden sheets, I get a sense of just how much effort has been put in to achieve the end product. Each metric has six sheets, so I’ll pick those apart one by one. Firstly, I can confirm that as far as data prep is concerned, the wide format of the default data source is used. When I looked at Ashish’s submission, I was interested to see the way he manipulated the date, and here is how it was handled this time round:
It’s kind of a hybrid of my “Excel-y” method and Ashish’s more technical approach. I like this approach and will try to remember this for the next time I need to format an unusual date string.
Onto those six sheets now. The first is simply titled “Label”, and manifests itself as the large numeric value on the left of each row. This is something I can do!
Really nice and easy. Nothing on Rows, nothing on Columns. Just the Actual metric dumped on Text. One down. Next up is a tab called “Variance”, and it’s the snazzy tri-tone series of bars:
Structurally simple, so the focus has to be on a couple of calculations. On Rows we have “Bus On-Time Variance”. It does what it says on the tin:
The values are expressed as percentages, as that is the way in which the metric is measured:
The next thing to check is therefore the calculation on colour, which will answer one of my four original bullet points. What does it look like?
OK, seems to make sense, but let’s bring that Latest Month field into view before jumping to conclusions:
A snappy Level of Detail boolean calc to validate if the date in question is the same as the latest date in the dataset. With that in mind, what is the “Bus On-Time Color” calc doing? It’s saying:
Irrespective of whether the variance is positive or negative, if it is the latest date, then call it “Recent”. If instead the variance is below zero, call it “True”, else call it “False”
So it’s the last date point, or it is a negative variance, or it is a positive variance. Nice and simple, but by my reckoning, it could be simpler:
With hindsight, it could be even more concise. After the “True” line, it could just be ELSE “False” END
The third sheet for the metric is “Trend”, and plots the line chart:
Pretty straight forward to interpret again. Dual-axis to enable the plotting of a dot on the end of the line, to draw attention to the latest date. The cleverness here is the Colour Card. Where the data point is not the latest month, the colour is set to match the background, thus hiding the other circles. A really neat way of doing things. Look what happens if I edit the colours:
I think it’s a smart little technique, as ordinarily I see the same achieved by writing a calculation to just plot the final date point. The approach used here is arguably easier, but just as effective. The fourth tab for the metric deals with another of my bullets. Is that “dark in white” chart just a bar-in-bar?
Yep! The primary axis is the Target, the secondary is the Actual. Size and Colour are used to create the distinction between the two. Next we look at a tab called “Variance text”, which simply shows the percentage variance to Target:
Simple enough to not warrant much further detail, other than to confirm that the field on Text is the same Bus On-Time Variance calculation revealed a few images above. Last, but not least, is the thumb up / down shape which addresses my first bulleted point of curiosity:
Another calc to take into consideration. Bus On-Time Shape is:
A boolean again. Is the Bus On-Time Variance less than zero? If it is, the Shape used is a downward thumb coloured red, otherwise the thumb points at the sky and becomes green. Cool. Three out of three bulleted questions answered. The final one asked how was this all pulled together so seamlessly?
Floating. Again. Seems like there is a correlation between aesthete and Floating on Tableau. With this dashboard, a series of Horizontal containers are Floated, bordered and shaded, and used as a mini-canvas on which the six sheet components of each metric are Floated, in addition to a couple of text boxes. Here’s a single metric:
It comprises the following. The main horizontal container:
A text box for the title:
The “Label” sheet is Floated:
And within that same space, four more separate Objects reside. First is the “thumb”:
Which partners the “Variance text” sheet:
Underneath both sits a Floating text box with the word “Variance”:
With the bar-in-bar chart representing the final component for this metric:
As was the case with the other vizzes from week 51, I’ve come to appreciate the art of design. A lot of effort goes into these submissions, but I reckon that straying from the “Tiled” path becomes pretty easy once you become accustomed to the “Floated” route.
This purposeful spell of deconstructing visualisations from the community has been quite emboldening. I don’t feel that any of these outputs are way beyond my technical capability, but they do highlight that perhaps my focus now should revert to two key areas:
I don’t know how much forward planning went into these submissions, but my planning is non-existent. I look at data, think in my head “that might work”, and bang it out. The quality of the dashboards I’ve taken apart this past week suggests to me that either each author has huge design credentials and experience, and / or they each take a breath and plan the end product before diving in.
I’m not a New Year Resolution kind of guy, but if I was, I’d be sorely tempted to inject a bit of patience and design into my forthcoming Tableau work.
Thanks for reading, and Happy Christmas everyone!