My #IronViz process: Research, Analyze, Design
I recently submitted my IronViz for the first 2019 feeder. I hadn't intended on entering, but the Agricultural theme was too much to resist (I might be the only one).
A few weeks ago, deep in the design of development of my IronViz Tableau dashboard, I made a promise:
So, here we are.
This post was just going to be about feedback, but I realized that without the context of my process it's hard to convey my goals of each round of feedback. So, I present to you my 3 phase, 21 step process to making an IronViz! Here's a quick visual that shows a rough estimate of the amount of time I put into each step (sorry about the hasty formatting).
If it sounds overwhelming, well, the IronViz was a bit overwhelming. And, I didn't set out with 21 steps in mind, but more of a general flow that I'm retroactively breaking into steps for your reading pleasure.
I promise I'll go through the steps [relatively] quickly.
It's a lengthy process, but it was my process and I like it. Don't feel like mine is the right way. You'll find your own.
A couple interesting notes, though:
- you'll find 8 steps are gathering feedback. Also, you'll be able to see my first version (which I didn't get to until step 14-ish) vs. my final version.
- Only about a third of my time on the IronViz was actually in Tableau. The rest was spent in conversation, word editors, the ag census Excel file, or sorting through all the cats on the interwebs to find articles about the latest in farming.
Phase 1: Research
Step 1: Data prep (~2 hours)This step was easy, because Chris Love (@ChrisLuv) is awesome. Courtesy of the Information Lab, he published a blog on how to prep the data. I just followed along.
See, I told you this would go fast.
Step 2: Discovery (~ 20 hours)I started by looking at the variables sheet in the Excel file, just to figure out what is in there. I began highlighting a bunch of variables I'd like to explore. I also highlighted the things I already knew something about. I grew up in farmland in Ohio - specifically, on a lumber farm surrounded by both pastureland and cropland. I knew my experience could provide more refined insights, and they were nice "grounding points".
Starting in the metadata helped me understand how the ag census thinks about data. For example, from a food perspective, mixing veggies and melons seems peculiar; but, from an agricultural perspective, the way you sow and harvest are similar. It was helpful to get a high level framework before touching the data.
I also looked at articles on the census definitions. There's some interesting patterns in farming that I worried could confound the results, and a bit of research confirmed that there's controversy in the definitions.
I also spent a lot of time figuring out what the data couldn't tell me, and the kinds of conclusions I wasn't comfortable making. This data was an interesting slice where thousands of farms and 4 seasons are rolled into each data value. This meant that I couldn't accurately measure -- even as proxies -- a lot of the things I wanted to measure. Note: this was a spot where I almost gave up, because this list kept growing and I feared I wouldn't feel comfortable standing behind any meaningful insights.
But I decided to press on.
Finally, I took all of my research, settled in, and asked myself "What's important to me? What are truths that I want to explore and convey?" This topic is an important part of my identity. What do I want the world to know? Robert McKee, a storytelling consultant, says "We [storytellers] have only one responsibility: to tell the truth."
So, from all this research, I found the truth I wanted to tell: I wanted to show the world that agriculture is more complex than a redneck on a tractor, and that our sustainability efforts need to match this complexity. I get frustrated by media and pop culture marketing of agricultural trends (don't get me started on "free-range" chickens, where the minimum requirements are often poorly executed and actually cause a number of health issues in chickens).
Step 3: Exploration (~ 4 hours)I took all the columns I'd highlighted during my research phase and performed simple univariate analyses by plotting them all as maps and boxplots. This helped me understand the distributions of values (and also provided a quick quality assurance).
Then, I wanted to explore relationships I thought might be interesting, so I performed bivariate analyses by plotting variables against each other (a bunch of scatterplots with trend lines).
After I'd plotted nearly a couple hundred charts, I started ruthlessly deleting worksheets that didn't catch my interest. By the time I was done, I ended up with about 60 charts.
Step 4 [feedback]: The expert (farmer) eye (~ 2 hours)My youth in farmland gave me a number of hypotheses on what I was seeing, but I specifically grew up on a lumber farm and I'm a bit out of practice on farming, unless you count my vegetable and herb gardens. I don't bring deep experience.
So, I reached out to some farmers. Luckily, I work with two and meet another one in the gym most mornings, so it was easy to have some conversations.
I showed them the plots (or described them, in the case of meeting at the gym), and provided some hypotheses and my assumptions -- and I'd have them provide their thoughts on what was happening, my hypotheses, and especially my assumptions.
This is where I found out I was wrong, a lot.
Step 5 [feedback]: The audience (layperson) eye (~ 2 hours)I got a solid piece of advice from a co-worker in data viz: I might dismiss a pattern because it is obvious to me, but it might be interesting to a layperson that doesn't know farming. Also, I might find something interesting, but they don't have the background to follow why it's interesting.
So, I showed my charts to about 5 laypeople that didn't work in data viz or data science or farming. I wanted to see what felt interesting to them.
Honestly, they found pretty much everything boring, at least initially. When I explained why things were interesting, they became more engaged.
So, I knew this would have to be a narrative-driven design.
Phase 2: Analysis + storytelling
Step 6: Focusing on a topic (~1 hour)All of the steps above helped me narrow my focus. I'd decided I wanted to focus on the complexity of agriculture and things we need to understand for real sustainability. So, I filtered to variables that might indicate sustainable farming. I knew that practices differ based on the type of farm and what is raised, whether livestock or crop. This even further focused my attention to just a handful of variables.
Step 7: Analyze (~ 5 hours)Now I was ready for a real analysis. This was about chasing down and really investigating the patterns I found during the earlier discovery step. My farming background gave me a bit of instinct on what to look for, but most important was a trained analyst / data scientist eye for peculiarities and knowing how to work with different kinds of data. This led me to diving deeper by categorizing continuous variables with cluster analysis so I could more efficiently explore distributions of crops and livestock for different kinds of farms.
A key thing here is that I tried really hard to turn my hypotheses into something testable (including the ones I'd heard from farmers). Surprise surprise, the farmers were usually right, and I was more right than wrong, but still wrong a lot.
Step 8: Construct a narrative + a "so what" (~3 hours)By this point I had my truth, a bunch of patterns, and some insights. As I had learned when showing the patterns to laypeople, I knew this had to be heavily narrative-driven; the data alone wouldn't suffice for interest.
I needed to explain what the patterns meant and why they mattered. The latter was the "so what", or the compelling part of my truth. Here's my truth, and here's why it's your truth, too. Here's what you should do about it.
So, I opened a note program and began typing an essay. No visuals, just text. A letter, if you will, to my audience.
Step 9 [feedback]: The expert (farmer) eye, part 2 (~2 hours)Without showing them any visuals, I conversationally walked through my narrative with a few of the aforementioned farmers. I had them shoot down a lot of my points, and they offered a lot of alternative explanations. Note that, even though I had done a ton of research, my own analysis, drawn from my own background, and already spoken with these farmers...sometimes I was still wrong. It was much less frequent, now, but they really helped set me on the right track for some of the nuances I was putting together.
Step 10 [feedback]: The audience (layperson) eye, part 2 (~2 hours)I did the exact same thing I did with the farmers, but with laypeople. Here, though, I sought feedback on how interesting and compelling my narrative was. I got a lot of great insight onto which parts I should really expand, and the less interesting parts I should minimize, and the things that might still be confusing to a layperson.
I refined my narrative essay until I felt like all the pieces (albeit with too many words) were there.
Phase 3: Design
Step 11: Design strategy (~2 hours)Now I knew my narrative -- it was time to think of my design strategy. By design strategy, I mean thinking through what I want to do for the users, and how that translates into design choices. I personally break my personal project design strategies into 3 pieces: what do I want the user to feel, what do I want them to learn, and what do I want them to do.
Feel: My analyses got a bit nuanced and complex, including cluster analysis and percentages that exceed 100, etc. So, I wanted to combat any overwhelming feelings my users might encounter -- anything that might get them to shut down and stop reading. Primarily, I wanted my users to feel calm. Secondarily, I wanted them to feel intrigued, curious, and, eventually, enlightened.
Learn: I wanted users to walk away understanding the complexities of sustainable farming, and how farming is a complex industry with nuanced relationships between practices.
Do: This was my "so what", or how I turned my truth into my audience's truth. I wanted them to walk away understanding the sustainable farming was complex, and efforts need to address not just one thing but farming as a whole. Hopefully this turns into more educated voting, purchasing, etc.
Step 12: Layout (~2 hours)Next came the organization. My essay provided a natural organization of the visuals, and I put general placeholders for text. I focused heavily on space and placement. Because I wanted my users to feel calm, I worked from a strategy of emphasizing white space -- especially at the beginning, where I was inviting users in and introducing a lot of new concepts. I knew that, later on, my analysis became iterative and I was reusing a lot of variables and chart types, so I could worry less about white space there. Last, to get to that curiosity and intrigue, I wanted lots of small opportunities for the audience to further engage the data, so I made space for a lot of copy that were interesting bonuses but not critical to the narrative (like the copy I have around the maps).
Step 13: Formatting (~2 hours)I also wanted to create curiosity and intrigue with some unique color palettes that reinforced the themes, but with something that was bright (after all, we've evolved to pay attention to bright things). This landed me on the dark sea-green to dark brown palette, with white in the middle that created a sort of "shiny" effect, but also helped de-emphasize the middle ground early in my narrative, where I wanted my audience to focus on the extremes.
As for the font, well, I was limited in choice: Tableau's proprietary fonts, Arial for sans-serif and Times New Roman for serif. Times New Roman really felt right. I knew I would have a lot of text for the narrative, and the serifs make it easier to read text when they are small. Also, it's such a common font that I knew it wouldn't distract from anything.
I then de-emphasized the chart clutter with smaller axis labels, and either eliminated my gridlines or moved them to a barely noticeable color, all toward that calm feeling.
Step 14 [feedback]: The data viz eye (~3 hours)This is the first time I showed my work to people in the data viz community. I was very particular on the number of people I asked and who I asked -- not because I think I'm above certain people, or better than them; but, given my design strategy, I knew there were some people that were incredibly talented at executing similar strategies: two examples being Michelle Gaudette (@LearnVizWithMe), who carries an incredibly strong expertise in education; and Ludovic Tavernier (@ltavernier7), who is the master of calmness with white space.
Here's the first version I sent out for design feedback (it didn't have quite everything yet, but I sent it out a bit early to make sure they had time to read and send feedback).
I got a ton of great feedback; I implemented some of it. I wasn't looking to just apply a bunch of quick thoughts: I collected the feedback, and compared it to my design strategy. I implemented the feedback that aligned with my goals. For the feedback that didn't, I used it to reevaluate my execution of my strategy, and see if I needed to try something different to be more successful.
Step 15: Copywriting (~5 hours)This is where I wanted to really flesh out my narrative. In the version above, you'll see that some of it was already filled out, but a lot of it changed pretty dramatically. At this point, I was looking at the essay I wrote earlier in the process, figuring out what sections could be replaced with visuals, and then translating the rest into the text placeholders I had in there.
At the end of this step, I had my first, true version 1 complete, where everything I wanted was there.
Step 16 [feedback]: Audience eye, part 3 (~2 hours)I took my viz and sent it out to a number of laypeople for feedback. I got a ton of contradictory feedback, which was really good. My audience was going to be diverse, and so my test audience was also diverse. This meant I had to accommodate a lot of different preferences. Some were very data literate, others hated numbers.
Again, I didn't implement all feedback - I compared it to my design strategy to ensure every iteration moved me closer to my goals.
Step 17: Copywriting, part 2 (~3 hours)I'm describing this as a repeat step because I spent so much time on it. As I mentioned, I knew the analysis had to be driven home with a strong narrative. However, I knew that large bodies of text would turn people off, so I wanted to make sure I was concise and precise - but that I also preserved a bit of personality, including my own story of growing up in farmland. This meant a ton of editing and revising just in this area; honestly, way more than I spent on the charts and the formatting.
One of the really common things I heard from laypeople was "so what?" Despite all my efforts, my truth wasn't becoming my audience's truth -- mostly because of the copywriting and their lack of agricultural expertise. This led me to emphasize the key takeaways at the beginning and reenforce the takeaways throughout and especially at the end. I also built a bit more context around agriculture so they could follow things.
Step 18 [feedback]: Repeat 15 and 16 (~ 4 hours)Iterating is fun, but also kinda makes you want to pull your hair out.
Step 19: Forget about it (~ 5 days)This is a really important step. I [mostly] dropped this for about 5 days. I wanted a fresh perspective. I tried to forget about it.
I ended up getting a ton of great, new ideas during this time, especially on how to word things.
Step 20 [feedback]: Full feedback + iterative modifications (~4 hours)Step 20 was like step 18, but with the full audience. It was near complete, but I was blasting it out to everyone who had already provided feedback, including laypeople, farmers, and data viz experts. I also opened this up to a number of other people that hadn't provided feedback. At this stage, I loosened up my restrictions, although I still sought out people that I thought could provide the specific type of feedback I was looking for.
I would collect a day's worth of feedback, go home, compare all the feedback and compare the feedback to my design strategy, implement tweaks, and repeat. This was the last week of the IronViz. These were all a TON of small tweaks, but cumulatively they had some *huge* impacts.
Step 21: Publish! (a couple minutes)All done. Deep breath. Relax.
And then I went farming...but on a very tiny, yet largely delicious, scale.
Let's bring it home
There's a few things from my process I want to emphasize -- things I don't think get enough vocal value. First, you'll note that I had Tableau open for less than half of my process. That's because of the emphasis I put into research, design strategy, and feedback.
A total of 30 hours just for research. Almost half the time I put into the entire project. Yes, I had a strong background in farming; but I still did a ton of research. I happen to know a lot of farmers, but I didn't just limit my research to talking with them. I read. A lot. In fact, about 50% of the articles Google recommends now are about agriculture, and Amazon is totally convinced I need a how-to guide so my new farm is successful.
This was important because I was wrong about a lot of things, and because the data itself was, honestly, a bit misleading. It was easy to read too much into the data. For example, there were only a few fields that indicated trends -- so unless I used those, I knew I couldn't say anything about trends. Also, I had to be careful about comparing counties to each other without including geographical context, given that farming in Ohio (where I'm from) is very different than farming in the South or the West (I don't need to worry about insects nearly as much as they do in Southern farms, or irrigation nearly as much as Western farms).
Design strategy was important, and it influenced every decision I made, all the way down to the color choice of the gridlines. Getting this right from the get-go was very important. Again, I wanted to devote strong though to my truth, and how to make it my audience's truth. I gave significant thought to how I wanted them to feel, what I wanted them to learn, and what I wanted them to do. Whenever I design a viz, either professionally or for personal expression, I try to start with this, and then every choice needs to be a reflection of that.
Feedback21 hours devoted to seeking out, gathering, and curating feedback -- about a third of the entire project time. Honestly, I'm not very interested in directive, shoot-from-the-hip feedback. Someone may tell me to use a different color, or a different font, etc., but if they haven't thought or asked about my design strategy, their feedback isn't designed to get me toward my goals. For example, someone might say "don't use red, it means bad". But maybe I want to alarm someone, or maybe I want to call to mind more violent imagery to drive home an impact. In that case, their feedback would actually work against my design strategy.
That doesn't mean I give everyone a briefing on my design strategy before I let them give me feedback, but it does mean that I look for feedback from people that either already exemplify my particular design strategy for this project, or I look for people that try to understand my design strategy before providing the feedback.
Even with these people, I still compare their feedback to my design strategy. In the example above, someone tells me not to use red. I shouldn't just dig my heels in and say "they don't understand my design strategy, their feedback sucks". Instead, I should think about how to turn the volume up on my execution, so that the reason I'm using red becomes more apparent and, consequently, more powerful.
The people that provided feedback knew that I wasn't going to apply all their suggestions, at least not directly. But everything that everyone told me turned into some sort of tweak, even if that tweak was very different than what they suggested.
One last point: I emphasized feedback from non-data viz experts. Even though this was a data viz competition judged by data viz experts, this project was very important to me. This is part of my identity. This is a dear truth.
I wanted my work to extend beyond just the IronViz. I wanted this to serve others beyond our community. In that sense, I wanted a critical eye for accuracy from farmers, and I wanted to make sure it was valuable, interesting, and informative to people that weren't invested in numbers or in agriculture.
Regardless of how the judging turns out, I'm extremely happy with this. I feel like I added a strong viz to my portfolio, but more importantly, I found an opportunity for personal expression that represents an important part of who I am, a part of the Jackalope brand. I got an opportunity to merge my current world with my past world. I originally never had any intention of competing in an IronViz (coding a dashboard in 20 minutes is enough of a nightmare, let alone in front of 15k+ people). But, when I saw the topic, I felt compelled to share. I'm thrilled to have had this opportunity, and I hope you enjoyed the experience of reading about it.