Who doesn’t love a map? For week 22 we looked at global internet users across almost 200 countries. Andy and I both had a fair bit of critique for the original viz and I am very pleased with the numerous great submissions from the community.
WELL DONE!
What was great to see were the high quality visualisations which managed to either do completely without a map or used a map very effectively to enhance the data story around internet user numbers.
There is also definitely a trend towards more simplicity and vizzes across the Makeover Monday community are becoming simpler and clearer. It’s great to see that the feedback we give as well as the comments many of our community members are contributing on Twitter on a daily basis, have helped everyone to improve their work and tell better stories with data.
And a non-data related development I have noticed that I’m excited about is that the regular members of the community seem to really have gotten to know each other. It’s almost like we’re working together several days each week and people are starting to recognise each other’s style and preferences and can tell by the image of a viz who the author is. The tone in which everyone is talking to each other, giving tips, answering questions, making suggestions and encouraging newbies is really supportive and friendly and a very positive reflection of the greater Tableau and dataviz community. I wanted to say thanks for being really cool people and for making this project a positive learning experience for everyone who gets involved.
We’re really looking forward to meeting many of you in person at next week’s live Makeover Monday event during TCOT London.
Now let me share a couple of lessons from this week’s challenge.
LESSON 1: AVERAGE OF AVERAGES
Yup it happened again and I’m just as guilty of using it during my analysis of the data. The dataset contained internet users per 100 people, which means it was essentially a percentage. An average number per year per country.
The data is at the country level of detail. We could go up to a regional level by grouping the data, or even a global level by taking the average across the entire dataset. But that’s a bad idea. Why? Well, we are not considering the weighting that went into the underlying data, e.g. the population of each country in a region.
During my viz process for this week’s makeover, I initially did this wrong, but Andy gave me some feedback, so I fixed it before submitting. Reminder: get feedback from others, and feel free to ask them before publishing.
There has also been a discussion about it on Twitter this week and Andy pointed out that he went about it the wrong way, too.
If you averaged your averages, check your viz to see what the numbers are and ideally fix it to show the right results :-). Check out Charlie’s blog for a great explanation.
My recommendation: Always check at what level of detail your data is captured. Ask yourself whether your aggregations are correct and a sensible way to represent the data. Ask others for their feedback on the way you treated the data.
It’s not a crime to do stuff wrong, so don’t be discouraged if you made a mistake. Just make sure you learn from the collective work of this community and if you don’t understand why what you did was wrong, then ask.
LESSON 2: 2X2 DASHBOARDS
This isn’t something wrong as such, it’s more a matter of preferences and my attempt to encourage you to try something new.
We tend to see a few submissions every week which are very much traditional BI dashboards. There’s nothing wrong with building those dashboards but often they are not as engaging as they could be. Partly that’s because of the layout and people being so used to the ‘KPI dashboard’ look and feel that they don’t really engage with the content.
But also, these BI type dashboards we see being submitted often appear to be a collection of charts that are not strongly tied together through a story and don’t have a natural flow to them.
My recommendation: Be bold and try a different layout. If you want to have multiple charts in your dashboard, why not go for a tall portrait layout and have the charts follow each other to tell a story? They don’t need to be in a 2×2 matrix.
Add some text boxes in between to break up the view and provide content as well as connect your visualisations within a bigger overall story.
Now let’s look at the favourites for this week…
FAVORITES
Author: Neil Richards
Link: Tableau Public
What I like:
- Neil’s design is clever, thoughtful and unique. You can tell that he considered a number of factors when putting this together, including context, colours and layout. It’s great to see that effort translate into a beautiful and well executed viz
- Despite the ‘behind the scenes’ work that went into the tiled map, the viz looks simple and effortless and the data gets to speak for itself
- The data tells a story through design and colours. No additional text needed.
- The legend contains enough detail in each map to help me look at the years in isolation as well as the bigger picture
- The colours are well chosen and the font is nice and simple to balance out the structure of the design
Author: Alicia F. Bembenek
Link: Blog
What I like:
- Alicia’s viz is simple and compelling.
- The black background works well for the topic of ‘Internet’ and the colours chosen for the circles look good on black
- The title provides good context and tells me what I should look for, i.e. a pattern that is consistent for the countries. So I am expecting some sort of visual grouping of the data points
- And alas, the dots that are grouped vertically by year (colour) shift together to the right as they change colour and size. Clever!
- I really like the font Alicia used and the uncluttered design of her chart, no unnecessary lines or labels
Author: Steve Wood
Link: Tableau Public
What I like:
- Steve’s viz is simple and neat and he went through a couple of iterations to get to his final viz after receiving questions about it
- The maps are a nice way to provide context for his overall message
- Focusing on the top 5 countries by highest proportion of internet users gives the viz a good story
- The colours work well and in combination with grey stand out very nicely
- Steve identified the issue with filled maps and instead of avoiding it, tackled it and provides a lesson for all of us that he included in his viz
Author: Michael Mixon
Link: Tableau Public
What I like:
- Michael’s design is outstanding. He very effectively combines small multiples of sunburst charts with a great colour scheme, a regional structure and a view across time. In one chart.
- The viz is engaging and makes me want to explore. And instead of giving me the overwhelming task of picking a single line, I’m just encouraged to select a region to get more detail. So I do and there’s a whole lot of additional information provided
- The detail of the additional information is great and the way it is presented makes it accessible for the viewer
- At any point I have the flexibility to return to the starting point which makes me more likely to interact more, rather than getting stuck and just leaving the page
- The context provided helps me understand the data from different angles
Author: Ulrik Willemoes
Link: Tableau Public
What I like:
- Ulrik’s viz has a simple design, uses colour effectively and tells a story
- Nice big call outs on each panel support the story and give the viewer the necessary context
- Colours are meaningful and really show the changes over time
- Going in increments of 5 years is a clever way to get around the gaps in the dataset
- The bar chart accentuates the context provided in the text
- The grey and purple combination works really well
Author: Miguel Cisneros
Link: Tableau Public
What I like:
- I cannot end the week without talking about Mike’s viz. This is stunning work.
- This kind of design draws me in and makes me want to interact with the viz, so I went to Tableau Public and clicked around
- It’s not just visually compelling but it also tells a story, the story of progress for each of the countries included in the dataset
- While there are other and simpler ways of conveying the information, it is a great example of what’s possible in Tableau and it will hopefully serve as inspiration to the community to try a few different things.
What a great collection of vizzes and nice lesson on averages!
Hi Eva,
Thanks for the lessons!
I’m trying to replicate the one of Ulrik Willemoes… Could you tell me what are the Bins doing, exactly? What are they counting?
And… is the LabelPosition parameter only there so we can adjust the label size on the dashboard?
Very appreciated!
Esther
The x axis represents the % of population with internet access, from low to high. They do not appear to be discretely binned.
This actually brings up a point related to my comment below.
If it’s valid to plot the proportions in a histogram to compare them, it must also follow that comparing the mean or median is also valid.
Regarding averages:
I brought up Andy’s use of the overall average to him this week, confused that it seemed to contradict the advice that has been given (though not convinced that it was actually an invalid comparison).
After a lot more thought and reading, I remain unconvinced that comparing to the average proportion is invalid.
There are two main things that can be calculated here, and two main methods of doing so:
Method 1: Use the Weighted Average, by factoring in the population of each country (this is what we’ve embraced).
Method 2: Average the Proportions (this is what we’ve shied away from).
But the issue is not that one thing is valid and the other is not; the issue is that those two things answer two different questions.
Method 1 answers the question “What proportion of people worldwide are connected to the internet?”
Method 2 answers the question “What is the average proportion of people connected to the internet per country?”
Asking and answering Question 2, and further asking “What is the proportion of people connected to the internet in my selected country, and how does that compare to the average proportion?” is not invalid.
What makes Method 2 invalid is trying to use it to answer Question 1. As in, you cannot calculate the average of the proportions, and claim that said average represents the worldwide proportion of people connected to the internet.
I think that it is important to fully understand what each number can and can’t tell you, what the implications of using each in comparisons or summaries of your data.
It is important to properly communicate in your data visualization what number you’ve used, what comparisons you are making, and the meaning and significance of those comparisons are.
It is important to provide the necessary context within which the user can understand the numbers and the comparisons.
But having done that, comparing a proportion to an average of proportions seems a perfectly valid thing to do.
Thoughts?
Further point/clarification/question:
If it’s valid to compare the proportions of multiple groups meeting a given criteria (e.g. the proportion of people in each country connected to the internet) to each other, despite the differences in sample size between those groups, does it not follow that comparing measures of distribution and center for those proportions is also valid?
Excellent points Jamie and great explanations. Thanks for taking the time to write this up. It’s a very valuable contribution to the conversation.