P01 - Directed Dot Plot 12-26-2024 03:11 AM

[carlosbarboza]

Directed dot plot versus the known, familiar but, overly-encoded clustered bar chart. 

A directed dot plot has only one side encoded with a marker, preferably the latest or most current value along with a very thin bar or line to depict the variation versus the prior value of the dotted or circle-marked value. This subtle graphical encoding permits to show direction of growth. If pointing the right, growing; else, decreasing versus prior year...and finally, the length of the line showing by how much.  Many thanks to data graphics team at The Economist for the inspiration. Note: "directed dot plot" is a name I have come up with for this example, this is different than the know dumbbell plot with both sides are encoded with a circle marker. In the comments you'll find the original chart.

Now, let's enhance the directed dot plot by adding the possibility of sorting the graph from any of the 5 different variables: the axis items, the values from 2019 or 2020 and the delta in either absolute # or percentage % terms between the two years along with the option in either ascending or descending order on any of these variables.

Sorting the data can make it easier for the audience to decipher rank and order. Sorting grows more important as you add complexity. As a consultant, I've been asked too many times to sort or re-sort a simple bar chart or dot plot by the difference of growth between the values they represent. Seeing sorted information makes it easier and faster to see the differences of what grew more or less so you don't have to skip around visually to detect and decide what's bigger and what's smaller. Hence, cognitive load can be saved for our dear brains from a simple mouse click...

Now, what you sort and how you sort changes the information displayed but can facilitate in answering questions in mind. Alphabetical sorting is one, that could be taken as too banal and it shouldn't, such interactive feature could serve as a quick lookup reference so you can find faster the item of interest and re-sort the graphical display to see it's encoded value; within its graphical context in which is being displayed.

And finally, words by Nathan Yau on this second-edition masterpiece; "Visualize This", on page 180: "Visualization is usually about finding the order in things. Sort the data so that order is more obvious".

You can download the pbix file below. Feel free to poke it.

eyJrIjoiYWJiMzA0OTQtMmI2My00MGM0LWFiNTctMzg5NjJmNDQwYjZmIiwidCI6IjEwZWM3OTJjLTU4NzctNGU1ZS05OGE5LTFiMWQ3YTNjM2RlYiIsImMiOjR9

[3rdSun]

@carlosbarboza - great visuals mate! I've tasked myself to recreate these visuals to up my PBI game and I started with the 'directed dot plot'. Having gone through your PBIX file, I haven't been able to recreate the dot plot (thin lines) for some reason. I know the chart is based on the clustered bar chart but how have you transformed it to the dot thin lines ... I can't see anywhere in the visual format where you did that. What I'm I missing?

[carlosbarboza]

Hello @3rdSun , basically the trick lies in the Error bars option (number 3 in the image below)

Now in order to get the "line", you need to insert a measure for the values which are being compared to the current values (step 1 on the image above). I have renamed the measure to "units 2019" or step 2.  Then go to the Error bars option (step 3) and on the "Apply settings to" (step 4), select this renamed measure (units 2019). Then on the Options sections, activate or enable it by toggling the button to On. Now step 5, is about placing the measure of "04 dot_2020" on the upper bound slot (pointed by red arrow) which is a conditional formula to check if the current values are greater than the values being compared (previous year on this case), so if the current is greater than prior then plot the current values. With this logic, you get to plot the "thin line", as seen on the image above (step 6) after toggling on the option of "Bar", and you can also visualize the circled end markers by activating or switching the Markers option (step 7) on the toggle button to On.

Repeat this process for series: 2019 or DAX measure 02 sales_19 (renamed to 2019).  But remember for this series to be formatted you will need another conditional DAX formula, which for this example it is 05 dot_2019.

Hope these steps and image help.

[odeddror]

Carlos,
Very nice job, do you have a video of how did you make it sort I've tried to mimik your tables but no sucsess.

Thanks,

Oded Dror

[carlosbarboza]

Hi @odeddror , thanks for the comment. Please I have updated the pbix file with the corresponding credits to where I learned this sorting technnique by @v-kelly-msft 

You can also refer to it directly here: Solved: Re: dynamic sort by measure filter - Microsoft Fabric Community

Let me know how it goes.  Feel free to also reach me via LinkedIn: https://www.linkedin.com/in/merod/ 

[Oded-Dror]

Carlos Thank you for your response I finaly figure that out. thanks again Oded Dror 

[odeddror]

Carlos,

Thanks for your response I found it in sort visual by the measure you created.

Thanks,

Oded Dror

[carlosbarboza]

below on point # 1, its the source of inspiration for the directed dot plot example made in Power BI  - now it could also have been an arrow chart like in point # 2, but comparing both options is the encoding of triangles for the end markers: intuitive but also a bit excessive compared to a simple circle marker?

point # 3 is the classical dumbbell plot which work nicely for categorical comparisons, and I find the design decision of toning down the hue of the benchmark marker to help the other marker, "the main character" get more attention (red markers)

now point # 4 is again a dumbbell plot with 3 categorical items to compared and I still struggle to process it....do you as well ? I think another chart type would have served better for this representation.

Again many thanks to the data graphics of The Economist  for their weekly charts on their magazines. Yes, I do cut them and saved them on a notebook #ShamelessNerd