Is it possible to create a “data visualization hierarchy of needs” like Maslow’s hierarchy of human needs? I’ve tried that in the pyramid above. Here are the details:
- Is it the right Medium? A chart can really help processing data and minimize information overload, but what about “chart overload”? Don’t make a chart if you don’t need one, and make sure that a chart is the best vehicle for your message.
- Answer a Question: Let’s assume that you do need a chart. A chart is (should be…) a visual answer to a very specific question. What is the question? Try to use is as the chart title: does the chart still make sense? Make sure you chose the right data to answer the question.
- Present the Data accurately: A chart is a visual translation of an underlying data table. This translation shouldn’t distort proportions and relationships between data points.
- Emphasize the Patterns: The whole point of presenting the data visually is to take advantage of our built-in pattern detectors, but you must help them. Choose the right chart type. Don’t display the data randomly. For example, don’t sort a categorical variable alphabetically (that’s almost random). The sorting key should be the data itself.
- Establish Priorities: Create layers of reading. Grid lines are less important than the data, so they should be grayed out. Some series are leading actors, other series provide context.
- Multiply Perspectives: Add detail, create multiple views of the same dataset, add interaction.
- Grab Attention: Add color, make your chart aesthetically pleasing.
What do you think? Am I missing something? What would you change? Add your comments below.
Interesting idea! But is it about a process, or is it about a chart? How does it, for instance, compare to the iterative Computational Information Design process described by Ben Fry? Is this somehting iterative as well?
But why isn’t data on the bottom? I mean, without data there is nothing to visualize and so the question whether the medium is right becomes irrelevant. Especially if it’s a hierarchy of ‘needs’, then I would suspect the most important need to be the data itself, at least if your definition of data is ‘the raw data’.
But perhaps question should be even lower than data, because a question may be the driving force of even starting the entire process: without a question there is no need for data.
Isn’t patterns conceptually something different? I think patterns are created by the data itself, while choosing the right chart is a deliberate choice of the creator of the chart. Choosing a pie-chart because that shows a pattern is not really the right choice I guess.
For your last point ‘make your chart aesthetically pleasing’, how would you do that? Just randomly adding color might reduce the effectiveness of your chart, but effectiveness is not necessarily the same as aesthetically pleasing. What’s your opinion on that?
And just a few questions: Why is attention colored orange? What about visually encoding the data, is that a concept that should be in there? Is priorities in your definition the same as Tufte’s data-ink ratio?
Anyway, I think it’s an interesting and creative idea to convert the concept of data visualization to the Maslov pyramid of needs, but I am not sure if I grasp your idea yet…
Jorge,
I like the idea of a pyramid to visualize the needs for data visualization. I think Jan Willem is right, when he is asking, if we are talking of needs or a process.
If I understand Maslow’s hierarchy of human need correctly, every layer of the pyramid has to be achieved one after the other – starting with the one at the bottom. More like a process …
I would start with “question” at the bottom. Without a question any visualization is nothing but a nice colorful image.
Second layer is “data”, because it’s the base for answering the question.
After analyzing the data, you find your “message”, the answer for the question.
Now one has to choose the right “medium”: which chart-type to underline the message best.
I agree to Jan Willem, that “pattern” is inherent to the data. The creator can only choose the right chart-type to visualize the patterns best – you are right! And no pie-chart, please 😉 … but is pattern a meaningful element in the pyramid? I doubt this.
Is “priorities” about emphasizing the message in the chart by adding small elements and/or comments? Perfect, I agree!
OK, the next step would be to show different “perspectives”. Now we are talking of dashboards, aren’t we? Good point!
I don’t think you need colours to “grab attention”. The “message” is everything you need: Give a statement, explain the situation and recommend what to do! The reporting-addressees – the decision-makers – will love you, I’m sure!
Instead: I think at the top of the pyramid there is “access”! Please no Emails with Excel-Files. They want to access their Reports via portals and their iPad.
the best, Lars
Jan: many good points.
I think it’s more of an interactive process. You go to the next level as soon as the previous one is filled, but levels above may trigger changes below.
I wasn’t sure about adding that first level. It is the question that defines if you need a chart or not… I’m just assuming that this question requires a visual answer.
You may not have the data, but you can have a question. It will also help you to identify gaps in your database.
“If a tree falls in a forest and no one is around to hear it, does it make a sound?” We may argue that there are no patterns unless they are revealed by a chart.Yes, you can use a pie chart if the pattern displayed in the pie chart is the best answer to your question. And since you are going to make multiple views from the same data you will be able to select the more effective ones.
I wanted to emphasize that most of the time you don’t need color. Color is a powerful pre-attentive variable so it can greatly enhance chart effectiveness. In that sense it would be included in Priorities. Here is a good example :).
Most of the time people use color for aesthetic purposes, not because they know anything about pre-attentive variables… How do you make a chart aesthetically pleasing? That’s in the eyes of the beholder, I guess… You can follow Tufte’s advice (“use colors found in Nature”) but no, I don’t think there’s a rule of thumb. I like minimalism but you may not. If you respect the needs below, the risk of reducing effectiveness by adding color is low. And they prevent you from adding textures, so no 3D effects as a bonus…
The data-ink ratio is a measure of chart junk levels. But you can have two similar charts with the same data-ink ratio and different levels of effectiveness. “Priorities” is about layering the objects in a chart and in that sense includes the judgment of the usefulness of junk objects.
Shouldn’t you have used a 3D stacked pyramid?
Seriously, it’s a good attempt to try to elucidate the data visualization environment. I don’t think there are specific levels in a specific order, as if you’re outlining the process. Partly it’s because of the iterative process discussed in the comments, partly the process runs in a differentorder in different cases, and partly because it’s a more amorphous environment. I envision more of a venn diagram, with the intersection of all those circles being something like “attention”.
Jon: What about a 3D Venn diagram with “real” bubbles? And animation, too! Well, Maslow was also criticized because there are many examples that don’t fit the hierarchical model.
This started as a simple play around Maslow’s hierarchy and how to transpose it to our knowledge field, but maybe it could be useful after all. Let me see if I can come up with something entry-level but a little more sophisticated.
Lars: different charts reveal different patterns. That’s why I am not sure that we can say that a pattern is inherent to the data. Maybe it’s a philosophical question without a single answer. Yes, different Perspectives can include dashboards but also trellis displays or scatter plot matrices, for example.
I think that we cannot avoid aesthetics and it makes sense to put it at the top of the pyramid. Yes the message is important, but how you say it is important too. That’s one of the differences between Bertin and Tufte.
Good point about “access”. Let me see if I can added to the model.
Jorge – I really like this idea, and I’m enjoying the debate. I think Jan and Lars made some good points. I think that while patterns are inherent in the data, they’re not always apparent, which is why it is important to select the right method(s) of visualization (as you stated, Jorge).
In terms of aesthetics and use of color, I try to find a good balance between minimalism and aesthetics. I believe that half of the battle is getting people to pay attention in the first place. If it takes a little pizazz, I’m more than happy to abide, being careful not to obscure the data too much.
Tom
Tom: 90% of the glamor(!) in glossy 3D pie chart vanishes when you switch to black and white (and gray…). It becomes more abstract, like every single self-respecting chart should be. We just need to convince the users that they should make their chart b&w… and switch to color when they the chart is ready. I’m sure they would focus more on answering the question than playing with textures.
I’m naive, but want to believe that, if you get the foundations right, the top will take care of itself. The problem today is that most people take a top-down approach to chart making: first, aesthetics, then the data.
Jorge – I hear you. When I said “a little pizazz,” I really meant a little. Nothing too over the top for me.
That’s a good idea to have folks make charts in black and white, and THEN switch to color.
By the way, I just witnessed one of the worst charts of my life. A 3D conical column chart using a second series line chart for the dates. This was in an annual report of a department! Looks like I may have to do some training 🙂
Tom: You sound like character in the Lord of the Rings: hope beyond hope. Do you really believe training will make any difference? Such a chart requires something a little stronger than training, like, hummm. brainwash.
Just kidding,
Haha – I’m an optimist! Training, brainwashing – same thing, lol. I really enjoy creating charts/reports, so I actually like it when departments ask me to help with such projects. Unfortunately, the longer I’m here, the less time I have to be so charitable! It’s good to be skilled in this area – thanks to you and others for helping me become more so!