Tuesday, October 29, 2013

storytelling with data in 140 characters or less

A few weeks ago, I ran a storytelling with data workshop for the IMPACT Planning Council in Milwaukee. It was a fun session (hosted by the School of Public Health at the University of Wisconsin, which is housed by a beautifully renovated former Pabst brewery building) with a super engaged group (plus my mother-in-law in attendance!). Last week, IMPACT shared with me a reconstruction of the workshop using the tweets published live during the event.

I found this pretty cool, so thought I'd share the bite sized morsels from my session here.
  • Cole Nussbaumer, kenote speaker at data viz wrkshp takes podium
  • Nussbaumer blogs about data viz at Storytelling w Data
  • Cole says understanding the situational context of the data is key
  • Who do you want to communicate to? What do you want to communicate? How can you communicate?
  • Keep in mind what background info is relevant? What sound bite could you use to clearly articulate ur message? No more than 3 minutes
  • Ur big idea must be a complete sentence that tells the audience abt the context of ur data
  • V imp to choose the right way to display ur data. Sometimes plain text is best option
  • Table or graph? Tables interact w our verbal system. Graphs interact w our visual system
  • Line graphs are for continuous data, usually across time. Bar graphs are for noncontinuous data
  • Bar charts shld always start from zero
  • Exploding 3D pie charts misrepresent the relationships btwn the sections. Don't use them. Our eyes have difficulty judging size of areas
  • There are many diff types of graphs. Always use the type that makes the most sense for your data and audience.
  • @laurynbb: And kill the 3D bar chart - data viz wkshp “@planningcouncil: Give your graph to a friend to see if they understand it”
  • You know you've achieved perfection in design when you have nothing more to take away
  • Gestalt principles of visual perception: proximity, similarity, enclosure, closure, continuity, connection
  • Eliminate clutter from your graphs.
  • Get rid of anything that makes the audience work . Make it easy for audience to understand the point you're trying to make with the data
  • @HelenBaderFound:  RT @planningcouncil: Eliminate clutter from your graphs. #zentweets
  • To focus attention where you want it, understand how ppl perceive info
  • @HelenBaderFound:  Follow @planningcouncil for tips on presenting your nonprofit's data, beautifully and effectively.
  • Nussbaumer recommends Stephen Few's book as a good resource on data viz
  • Ppl can keep abt 4 pieces of data in their memory at one time so design your graphs accordingly
  • @HelenBaderFound Thx for following our live tweets of Cole Nussbaumer & Thx for your support of this wrkshp!
  • @laurynbb:  VizComm tenets “@planningcouncil: You have 8 secs to get audience's attention”
  • The most effective data viz will still fall flat if you don't have a story to go with it.
  • Stories stick in our minds in a different way than facts do.
  • Use text to highlight key points in your graph
  • Plot, twists and ending are components of your data story. If there's no twist - if it's not interesting, don't share the data
  • Tactics for making story clear: horizontal & vertical logic, repetition, reverse storyboarding, fresh perspective
  • Your graph is the evidence that backs up your story
  • Cole says Excel can be used to make good graphs. It's not the default charts but a good user can make excel work
  • Light backgrounds on graphs are easier on the eyes (and ink) than dark backgrounds
  • Any time you cut out info be sure to think abt what context you might be losing
  • Everything in a data viz (text & visuals) needs to reinforce the same message
  • Rather than hope the data will tell you what it's about, be clear what your question is & then organize the data to answer the question
  • Cole asks what distinction do you make btwn data viz and infographics?
  • Infographics came out of journalism but have changed over time so now they are more glitz than data
  • Just putting in a graph or infographic in to fill white space is not a good reason
  • Donut graphs are even more difficult to read than pie charts; don't use them
One of the makeovers we discussed in the workshop.
  • @MilwaukeeStat:  @planningcouncil and @storywithdata – thanks for putting on the best data viz gathering Milwaukee has ever seen. Well done!
  • @storywithdata Thanks, Cole, for inspiring 50 Milwaukee data geeks today with your advice on good data visualization!
  • @storywithdata:  @planningcouncil Thank you for the invitation to speak to your group - an engaged and lively bunch - I had a fantastic time with you all!
Another visual makeover from the workshop.
The full PDF (that puts my session back into the context of the rest of the afternoon, including a brief segment by Milwaukee mayor Tom Barrett) can be found here. Big thanks to the IMPACT Planning Council for hosting the event and allowing me to post their recap here!

Wednesday, October 23, 2013

the right amount of detail

In my workshops, we spend time discussing the importance of context when it comes to putting yourself in a position to create an effective data visualization. In my mind, there are two flavors of context that are important. First, there is the context that you as the data visualizer must know: who you want to communicate to (your audience) and what you need them to know or do (the call to action). When it comes to creating the visual communication, the second type of context also becomes important: the context that your audience needs to know in order to make sense of the information you are attempting to communicate. This often leads to the question: how much detail is the right amount of detail to show?

The answer is... it depends.

I started off writing this post with the assumption that the right amount of detail for a given situation depends on a lot of things, but as I sit here and think about how to articulate it, I'm landing on just two things that I believe are most important: 1) the method of communication and 2) the level of trust your audience has in you. Let's discuss each of these briefly and then we'll look at an example.

The method of communication. When you present something in a live setting, you are in nearly complete control. You can respond to audience cues and questions. You can get away with having less context and detail physically written down or shown because you are there to fill in any blanks. This is not the case in written communication, where the level of explicit detail must be higher because you are not there to facilitate the flow of information but rather must count on your words and visuals to paint the full picture (and address at least first-order questions). Because of this, the level of detail necessary for written communication is much higher than for a presentation.

The level of trust your audience has in you. If you have an established, trusting relationship with your audience, you can get away with showing less detail without your audience questioning it. On the other hand, if you haven't established a basis of trust with your audience, you may need to show more detail explicitly so that your audience doesn't feel you are somehow trying to mislead by only showing the parts that back up your story and ignoring the rest.

When thinking about the audience's level of trust, be sure to keep the ultimate audience in mind. If you're preparing something that someone else (let's call her Mary) will be presenting, think about not only how much trust Mary has in you, but also how much trust Mary's audience has in her and use this as your guide for determining the appropriate level of detail.

One thing to try to anticipate when considering the right amount of context and level of detail is what questions your audience will have. Of these, think about which need to be answered explicitly in your communication directly vs. which can be addressed on the fly. In a live presentation, one feature you can leverage is the hyperlink. If you can anticipate potential questions and have visuals or data you can use to address but only want to leverage them if the conversation turns that direction, you can include that detail in an appendix, then create a hyperlink on the given slide in the main presentation that points to the relevant slide in the appendix. If the question comes up, this gives you a seamless way to jump to the appropriate slide (but doesn't force it in the progression in the event that the question doesn't come up and you don't have to address it). Make sure to also make a link from the appendix page back to your main presentation so you can continue with the planned flow.

Let's look at a quick example of how we might think through the above as we're creating a specific visualization. The following is a generalized example from a recent consulting project. Let's assume here that this is one slide that's part of a broader presentation to the head of Marketing for our company and that the story we want to tell is around the most important features based on the survey feedback.


When we're thinking about what level of detail to show here, there are a couple different decision points:
  1. Can we combine response categories? Currently, there are five response categories that range from Extremely important at one end, to Not important at all at the other. Could we combine some of the categories to make the information easier to consume? What information would we lose by doing so? Even if we don't combine categories, there are things we can do to draw more attention to some and de-emphasize others that will allow us to keep the detail there without it becoming visually overwhelming.
  2. Can we show fewer features? There are currently 15 features shown - this is probably the complete picture of all of the features we asked about in the survey. We could think about reducing this to just the features we want to focus on and get rid of the rest. Or again, if we don't feel comfortable eliminating them entirely, perhaps there are things we can do to draw attention to some and de-emphasize others that will allow us to preserve the detail without it feeling so cumbersome to our audience to consume.
Let's look at a couple possible solutions. 

If our audience trusts us and we're comfortable with a minimalist approach, we could reduce the visual above to something like the following:


If the above feels like we've taken too much away, below is an alternate view that preserves nearly all of the detail from the original visual, but uses emphasis and de-emphasis to make it easier to consume and less visually overwhelming for our audience.


Note that in both cases above, I've also added a footnote with some details about the survey. This is detail that needs to be there, but can be de-emphasized (make it small, grey, put it at the bottom).

If we're presenting this information live, I'd suggest going with the minimalist version: there's less detail there to process, which means your audience can listen to what you're saying vs. focus their attention on deciphering the graph. Keep the full version on hand so you can address questions that might come up (having a printed version of it you can refer to if necessary can work well), or you could put the detailed version in the appendix and leverage the hyperlink strategy discussed above to jump to it only if there are questions on survey feedback on the other features. This could work well in a written report as well, where you can put a footnote that says "Survey feedback on full feature suite can be found in the appendix on page x" or something to that effect.

In case you want to look at this example more closely, the PowerPoint file that contains the above visuals and underlying data can be downloaded here.

What's your reaction to the above? What other considerations should we have in mind when it comes to determining the right amount of detail to show?

Note: it struck me in the middle of writing this post that I started out writing about context but ended up with discussion focused more on the level of detail (vs. broader context). Rather go back and rewrite, I'm going to go ahead and post, with some minor edits and the hope that the discussion above isn't confusing. There are certainly other aspects of context that are important when it comes to specifically understanding the data that is shown as well, but I think discussion there is probably more situation-specific so haven't addressed that at all here but rather focused my discussion on level of detail.

Monday, October 21, 2013

another advertising graph

I feel conflicted when it comes to the use of graphs in advertising. I like it in theory. But in practice, I tend to be disappointed with what I see. Almost like the designers couldn't come up with anything better, so they threw in a graph. Perhaps it's just my nature, but also when I see a graph in an advertisement, I'm immediately skeptical - it's like I start with the hypothesis that the creator is trying to mislead me. I'm not sure what drives that. Obviously, I like the use of graphs to communicate information; what is it about graphs in advertising that gets under my skin?

Last month, I enjoyed reading your comments on the vacuum graph. Last week, I came across another advertising graph in a fashion magazine:


I can outline specific things I would change in the above visual. But beyond that, it's interesting to me that my first reaction to data in advertising like that depicted above is skepticism vs. improved understanding.

What is your reaction to the Neutrogena graph? Have you seen examples of data and data visuals used successfully in advertising? Leave a comment with your thoughts.

Tuesday, October 15, 2013

how I would visualize this data

A couple weeks ago, I posted the following visual (prompted by a question that came up at a recent workshop), along with the challenge to come up with a better way to show this data.


Big thanks to everyone who spent time with this data and shared what they came up with (see full original post, plus comments linking to reader solutions here).

I found myself with some unexpected free time this afternoon (that hasn't been happening so much lately!) and spent a bit of it playing around with this data. There were a couple challenges I was looking to address in my solution:

  1. First, the inclusion of Unaided Awareness. I can understand why it's important, but it doesn't seem to quite fit with the other categories, as many of the reader comments pointed out. Rather than complicate my visual with it, I omitted it entirely. Perhaps it belongs on another page of the deck that walks through this data?
  2. The other challenge from my perspective is that the remaining categories are each a sort of subset of the prior category: those who prefer a given brand are a sub-portion of those who would consider the brand, which is a sub-portion of those who are aware of the brand. I think this means that it doesn't make sense to stack the bars, or to compare only a single category at a time across the different brands as some of the user makeovers suggested.
For me, the goal here is to get a quick visual understanding of how the three categories for OUR COMPANY compare to various competitors, as well as how the three categories compare to each other. The goal is for our bars to be big to the right. There are many competitors with bigger bars than ours. But the ratio of Consider to Aware is actually pretty good for us (better than many competitors). Here is how I chose to visualize this data:


It isn't particularly creative. But perhaps it doesn't need to be. In the original visual, the two main things I wanted to change were 1) the idea that you should compare only downward - the layout of the original charts left me wanting to compare competitor A to E to H (first column), whereas really I should be comparing OUR COMPANY to each of the various competitors (not only those within the same column); and 2) center aligning the bars as is done in the original makes the relative values hard to compare, so I wanted to align each element to a single baseline to allow for easier comparison both within a specific brand as well as across the brands.

To those who read this blog regularly, my choice of a horizontal bar chart here probably isn't a surprise. I use them a lot. I use them a lot because I think they are easy to read. Note that this example isn't totally complete - there's still a story to be woven around this data to make it relevant to our audience. Given that I've anonymized the example so much here, I decided to focus just on the data viz in this case.

I think this works here. What do you think?

In case you're interested, the Excel file for the above visual can be downloaded here.

Tuesday, October 1, 2013

how would you visualize this data?

I had a follow up session yesterday with a group that recently went through one of my custom workshops. It was really great to see and hear how participants have been able to put what we learned to use and to discuss where they are facing challenges and strategies to overcome them. One participant recapped what they took away from the session: enhance what's important, tone down what isn't, and question the question!

We also looked at some specific data they were having challenges visualizing and had some fun experimenting with possible solutions on the whiteboard. There was one that we didn't get to that I want to share here for input:

Click here to download the data in Excel.
A few comments on the data:
  • In the original, "Competitor A," "Competitor B," etc. were actual competitor names, which I've generalized here (and in the downloadable data) to preserve confidentiality. 
  • Unaided Awareness, Awareness, Consideration, and Preference are in regards to the given company's brand. This is sometimes known as the "purchase funnel". The data is collected through a survey; percents represent the percent of the total sample population. 
  • Unaided Awareness is the % of people who know the brand off the top of their heads (for example, What brands come to mind when you think of shampoo?).
  • Awareness is in response to a list of brands: Which of the following brands of shampoo are you aware of?
  • Consideration is which brand a consumer would consider purchasing.
  • Preference is which brand a consumer is most likely to purchase.
  • Conversion (highlighted in title) is the percent of people who move from one funnel step to the next (what percent of aware people consider you, etc.)

The intent is to be able to compare the brand of "OUR COMPANY" along these dimensions to the various competitors, as well as get a sense of where our opportunity lies.

I think this could be done more effectively than it is here. I have a couple ideas, but am lacking time at the moment to try them out. So I thought I'd solicit your help. How would you tackle this visualization challenge? Leave a comment with your thoughts!

Note: please link to your visual if you create one; you can embed a link in your text comment with the following HTML (replace the parts in blue).
<a href="http://www.link_you_want_to_insert.com/" target="_blank" >text to link</a>