The Clearly Podcast

Vizualization Pitfalls

Summary

This week, we are mostly being rude about pie charts.  To be clear – we like pies, just not pie charts.  At the heart of the issue, is that bar charts give a better view of overall magnitude as well as the relationships between categories.

We then move on to 3D charts – which are not in the standard visualisation set in Power BI – which always end up much harder to use than their 2D counterparts.  

Be careful of overcrowding the page.  This not only looks cluttered and confusing but also creates a performance issue.  Multiple pages with drillthroughs can create a much clearer story with the data.

Think about how you are going to maintain the report into the future.  Around 6 visuals on a page would be a good rule of thumb.

Be careful of custom visuals since they don’t work with Power BI Report Server (ie hosting on premises).

When it comes to maps, think about how you can use the data to make real-world decisions.  Shailan’s fly tipping report can be found here.  ArcGIS is generally the better mapping tool in Power BI, but one to use mainly if you really need mapping and can make use of its additional layers of data.

The Decomposition Tree is a great way of managing the interactive exploration of data that our older listeners may remember from such tools as Pro-clarity and Performance Point Services.

For the developers amongst us, you can create your own custom visuals which may allow you to work with libraries like D3JS.  Once again, do you really need to do it?

You can download Power BI Desktop from here.

If you already use Power BI, or are considering it, we strongly recommend you join your local Power BI user group here.

Transcript

Andy:
Hey everyone, welcome to the Family Podcast episode 9: Visualization Pitfalls, recorded on December 2, 2020. I'm Andy Clark, joined by Shailan Chudasama and Tom Gough. Today, we discuss common pitfalls in building Power BI reports, focusing on why we dislike pie and doughnut charts, speed optimization with visuals, the pros and cons of custom visuals, and effective mapping techniques. Enjoy the show! Hey guys, how are we doing?

Tom:
Good, thank you.

Shailan:
Good.

Andy:
Great news to start: we have a new subscriber, Jason Smith! Now we have more than just our moms listening. So, let's dive into visualization pitfalls. Shailan, this topic was your idea. Can you explain what you mean by visualization pitfalls?

Shailan:
Sure. A common pitfall is using pie charts when bar charts would better represent the data. Pie charts don't show proportions as clearly, especially with more than three or four categories.

Tom:
Exactly. Pie charts make it hard to compare data side by side, and they get messy with many categories. Bar charts provide a clearer visual representation.

Andy:
So, neither of you likes pie charts. Can you explain why to our listener?

Shailan:
Pie charts are not as effective for showing proportions. A bar chart is usually better.

Tom:
Yes, bar charts let you compare multiple categories more easily. Pie charts are limited, especially with small differences between categories.

Andy:
What about doughnut charts?

Tom:
Doughnut charts are slightly better because you can put a number in the center, but they still have similar limitations to pie charts.

Shailan:
And remember, if you put a label in the doughnut, send it to the back so it doesn’t block interactions.

Andy:
Users seem to love pie and doughnut charts. Why do you think that is?

Tom:
It might go back to school days where pie charts seemed complex and interesting. They just stick with people.

Shailan:
Infographics often use pie charts to show percentages, but a bar chart can do the same job.

Andy:
When looking at a report, what are the pitfalls to avoid with visualizations?

Shailan:
One pitfall is filling space with redundant visuals. Less is more. Don’t overload your dashboard with too many visuals.

Tom:
Too many visuals slow down performance because each requires a data query. Keep it simple to ensure faster performance.

Andy:
Is there a rule of thumb for how many visuals to use?

Shailan:
No strict rule, but focus on telling a story with the visuals. Use enough to convey your message without redundancy.

Tom:
Exactly. Limit the number of visuals to what’s necessary to tell the story effectively.

Andy:
What visualizations work well together?

Shailan:
Bar charts work well with line charts. Bars are great for categories, and lines are good for trends over time. Tree maps are also useful for classifications.

Andy:
What about custom visuals?

Shailan:
Standard visuals are tested and secure. Custom visuals can have security issues and might not be supported in all environments, like Power BI Report Server.

Tom:
Yes, and they can slow down your reports. Only use custom visuals if necessary and be aware of their limitations.

Andy:
Let’s talk about mapping. Users get excited about maps, but are they always useful?

Tom:
Maps need to add value. For instance, mapping sales often just shows population density. Use maps to show meaningful data, like sales per capita.

Andy:
I’ve seen effective use of maps showing changes over time, like mapping incidents for city funding. Any other good use cases?

Shailan:
Yes, like mapping military training exercises or showing flight paths for casualty evacuations. These uses provide real insights beyond just looking cool.

Andy:
What new visualizations are people excited about?

Tom:
The decomposition tree is a new and useful visual. It lets you break down data dynamically, which is great for exploring and explaining data.

Shailan:
Yes, it’s more flexible and visually appealing for drilling down into data compared to traditional hierarchies.

Andy:
What are your top tips and pitfalls for visualizations?

Shailan:
Top Tip: Use standard visuals and tell a clear story with your dashboard.
Pitfall: Avoid redundant visuals that provide the same information.

Tom:
Top Tip: Keep it simple for better performance.
Pitfall: Avoid using pie charts.

Andy:
Top Tip: Be consistent with your visuals across reports.
Pitfall: Don’t forget to include labels or notes to explain your visuals.

Thanks for listening! To get in touch, visit clearlycloudy.co.uk for the UK and Europe, and clearlysolutions.net for North America. Email us at ourpodcast@clearlysolutions.net. We’ll be back on December 28th with an interview with Hugo Soto, author of The Analyst Mindset. Have a Merry Christmas and stay safe!