Inspired Insights: The PICA Protocol for Actionable Data Viz
Thank you. What a wonderful introduction. I’m really honored and humbled to be included in the amazing lineup for this virtual summit and I want to say hello to everyone listening. I’m so thrilled that you took time out of your day to join me. That introduction really said it all and I want to jump right in. We have a lot of really valuable information to cover today, but I want to lick things off with a pretty melodramatic and bold statement.
After the next 30 minutes, you will never approach how you present your data the same way again. I know that is really bold, but I believe it to be true. And to find out if that’s true, I want to know if this situation sounds at all familiar to you:
One of your managers or VPs or clients calls you up on the phone and he says, “Hey, I need you to come present come campaign results at next week’s leadership meeting. Okay? Thanks, bye.” Any you’re like, “Oh my god!” Because everyone that matters to your career is going to be at this meeting. So, you want to knock this out of the park.
You are working around the clock, sifting through mountains of clickstream, and Big Data, and voice of customer, and all kinds of crazy data. And you’ve now distilled into some shiny new insights.
Now presentation day. You are showering your audience with nuggets of wisdom and you decide to take a quick look around the room and you suddenly noticed that…
Uh, oh. You don’t have your audience’s undivided attention. Glazed eyeballs, a little drool, it’s not pretty. By the end of that meeting, you’re not exactly sure that your insights had the impact that you were hoping for. And after days or weeks of waiting by your phone or email, and signs of life that that information is going to get acted upon… Nothing.
You start to worry that all of your hard work has flown into what I call the dreaded “data black hole.” Yikes. Now if this does sound familiar to you at all, you’re not alone, because this was actually my story. Today, I want to empower you with the tools that you need to understand why this happens and help you prevent it.
I call this session inspiring insights. What you need to learn to present results, to get the results you’re looking for in your organization. A little bit of backstory. As mentioned, I cut my teeth doing digital analytics for a long time. One time I was doing an analysis of our site search behavior and I noticed that the vast majority of people using site search were leaving after the first time they used it. This was a big problem, but I was determined to not let this valuable information fly into the data back hole.
So, I decided to go on a three-year philosophical journey to understand answers to this deep metaphysical question: Why do bad things happen to good data? Why? It really does. It’s baffling. I would go on to find all kinds of answers to this question, but the one I found at the route actually kind of surprised me.
I would like for you to meet your audience’s brain. Hello. This massive bundle of neurons and synapse using glial cells, all bundle up to be your best friend or your worst enemy during your presentations. And learning how this thing actually works and absorbs information is what’s going to get you ahead. Unfortunately, when we leave analyst school and enter the workforce, no one is empowering us with this information. I certainly didn’t have it. No one told me that after I’m done crunching numbers all day and I walk into a conference room to present, what my stakeholder is really thinking is…
“Hey there, I don’t actually care about what you do all day. I’m sorry to break it to you. I definitely don’t care how difficult your job was of how complicated this analysis was. This is what I do care about: I want you to show me how to move my business forward. I want you to show me in a way I understand quickly and easily, and here’s the key, in a way that makes me want to act. Oh, PS, give me a count of hits anyway because someone told me that was cool.” They don’t always know how to articulate what it is that they want. It’s our job to help make that connection for them. I saw this as a real problem when I was trying to get my visualizations noticed, understood, and acted upon. I decided to create a prescriptive approach for handling this.
I call it the Pica Protocol. Why Pica? Well, I’m terrible at creating acronyms and naming things, and the letters seem to fit.
But maybe, selfishly, I wanted to give my last name meaning other than a small spherical rodent that sometimes eats dead birds. Yum.
Or—this is actually my personal favorite—an appetite for substances that are not food like paint and dirt. You can see why I might want to give it a different meaning. What I’m hoping that you’ll see is that this is going to be a super practical and effective approach for driving action from your visualizations.
A little bit of Twitter housekeeping; I would love for you to cc me @LEAPICA on any tweets and please don’t forget to include the conference hashtag: AnalyticsSummit. Ready to get started? I’ll take that as a yes.
If you are familiar with me at all, I have a podcast. A little, tiny podcast called “The Present Beyond Measure Show.” You should definitely check it out. There are some great guests on it including Krista Seiden who’s also part of this lineup. I include a couple of video makeovers as part of my podcast of listener-submitted visualizations that were feeling a little sick and needed a health-boost. I’ve adapted one of those examples here for you today. I think it’s going to show you a lot of interesting tips.
Here was the challenge: how to display trending in volume and performance targets at the same time. That is a lot of information for one space.
And even more challenging than that—and we’ll take a look at what the original was sent in—was all of those variables had to be represented across four different marketing channels. The first time I saw this, my brain kind of went, “Oh gosh. I don’t know where to look. I don’t know exactly what I’m looking for.” There was a particular story jumping out at me.
I want to take a closer look at just one of these channels and put it under the microscope of what exactly is going on here. First, we have current monthly revenue represented by these vertical blue bars. The purple line in here represents the projected percent year-over-year growth. This was one number calculated for the entire year. It didn’t change. This red line here represents actually percent year-over-year growth in revenue. One little caveat is when they sent this to me, they used dummy data and they didn’t have barring actual growth. I changed that for this example, but just in case you were curious.
No, I want to examine some of the issues in the area of opportunity for this visualization. The first is there is a moderate amount of something come cognitive load on this slide. Cognitive load is something I’m going to explain more deeply, but basically it means extraneous noise. It’s not too bad here, but there’s definite room for improvement. The next is that these different scale dual-axis with the bars and lines are confusing. This is a big thing we’ll be talking about in a little bit because we use dual axis charts a lot.
Then the next thing is, because it’s a dual axis chart, these target lines interrupt the bars. The bars within bar charts really need a nice smooth, unbroken surface for accurate interpretation. That is how our vision works. But here, they create these blocks that are really interfering. Let’s get started with the methodology.
The very first letter is P, and it stands for Purpose. Before I tackle any visualization or open any tool that I’m about to use, I sit with my data and I ask it a very important philosophical question:
Why do you exist? Think about that. What question are you trying to answer? What decision are you going to inform? Taking a minute to ask this of your visualizations is going to prevent a giant data puke of any numbers that you might have at your disposal and help you get to something focused that’s actually going to be valuable to your stakeholder.
Once you have the answer to this question, what you’re able to do from there is to choose the best visualization for that data. The question should directly impact the choice of visual. I’ll tell you something, our tools that we work with aren’t the best at making this final judgment call for you. They do use algorithms to suggest maybe five to six of the most appropriate choices, but ultimately, that’s a judgement call that’s going to come from practice. But you don’t have to go that path alone.
If you need help getting started on which charts to choose for your questions, I highly recommend this book by my colleague, Stephanie Evergreen of Evergreen Data, it’s called “Effective Data Visualization.” What I love about this book is it not only does it only talk about charts that are really the best practice charts to use—not every chart in the world, but most effective ones—and it instructs you on how to choose those charts based on business questions you might be trying to ask. It’s an invaluable resource and I highly recommend it.
After I sat down with my listener’s visual, I asked why do you exist. This is where I think the question came out as: how is each channel trending over time and how is each performing against its target? Two simple questions, but all in one view. The answer that I came up with was a little bit of a tweak to the original setup.
What I’m going to do is created a trended column chart, which we already have, but pair it with a separate small multiple lines chart. That’s not an often-known chart and I’ll be going through exactly how that can be used.
Here’s where we’re starting with in terms of this one visual where the story really isn’t clear.
Here’s what it looks like after. I’ve separated these charts, and believe it or not, sometimes separating values in separate charts from each other can increase their collaboration with each other. Here I have a trended column chart for volume. I like to use bars for volume metrics because it kind of represents a bucket that’s filled with something tangible. Now up here, we have a small multiples line chart. You’re not actually seeing the small multiples part of this—and I’ll be going through that in a moment—but it is a simple line chart. I like to use line charts for calculated metrics because it speaks to me of the sort of EKG heartbeat, which performance can be related to. You don’t have to abide by that, but it is a way of creating association from the types of charts you’re using. That is how you determine the purpose and choose the right visual.
The next step in the process is I for Insight. This word is thrown around gratuitously in our line of work. We always want to know what our key insights are. Some time ago, I decided to look up the definition to understand what that word actually mean when I was saying it.
It kind of interested me. The definition of an insight is: “The capacity to gain an accurate and deep intuitive understanding of a person or thing.” Let that sit for a second. And I challenge you to ask yourselves: do your visualizations, on the regular basis, meet these criteria? Mine certainly were not before I went on this journey. That’s what this step in the protocol is all about.
It starts with three useful tools that I have to help you surface insights quickly in your visualizations. The first is target or benchmark lines. Luckily, we already have that, but I always recommend if you have a target, plot it. If you don’t have a target, create one. Look at the last six months or so of your data and develop a benchmark because you’re more likely to be motivated to move the needle on something if you’re seeing a goal line in the end.
The next one is intentional data labeling. There is a way to label different data points with the intention of telling a story. The third is a multiple chart layout. You saw that I separated those two charts, so how your layout your charts are a way of driving insights that you might not have otherwise seen.
The first tool I want to go through are these target lines. In that small multiples chart, I’ve included a target line here. The thing to notice about that is I’ve dotted it because I’m representing a theoretical figure. A solid line or a solid bar is great for representing something that is actual or tangible, but I like to change the pattern simply to represent something that’s theoretical or didn’t happen. That’s one way of creating distinction there.
The next is unintentional data labels. Sometimes we get a little enthusiastic about our data labels and it can actually obscure insights better than help us see them.
For me, intentional data labels mean not labeling every point, but selectively labeling them. So, for a line chart, what I prefer to do, especially in a dashboard is label the maximum, the minimum, and the most recent value. Those usually answer the three most common questions that a stakeholder might have during a presentation. You can play around with that.
In here, the bar chart, I also labeled the last bar for reference because, with the overall axis range and the last one, you can get a good sense of what that variation is throughout the months.
Here’s one of my favorite insight tools: multiple chart layout. Here we’ve crammed a lot of space and each channel is in its own space. They’re really not working together to help us understand. The layout is only here to help save space.
What happens when we change the layout? We are able to allow understating between channels because of the way we’ve laid things out. Just to go through what I did here, I created vertical tracks for each channel so that performance is first and foremost and then volume is next for context. This allows us to clearly see all of the various points and answer those questions that we need.
Now I’d like to focus a bit on the upper section here, which is the small multiples layout. What the small multiples layout means here is that I’ve separated these lines by segment in order to improve the clarity and comprehension around what’s going on.
What this allows you to do is avoid a phenomenon I call “line chart spaghetti,” where all of the lines are plastered on top of each other and they don’t allow you to see the nuance in the trends by the different segments. That is what those small multiples allow you to do.
The next thing you can play around with in terms of the bars on the bottom are value scales. The value scales that we use can influence how we interpret the information. In this first few, I’ve put all of my bars on relative scales to themselves so the top value across all four channels is relatively at the same place. This allows you to view trending performance more clearly within the channel. The trends follow the same pattern.
It may be more important to you to understand what the relative volume is between the channels. You could plot all of those values on the same scale. What this allow you to do is view trends within the channel as well as understand the volume between channels. If we had this view and we wanted to make a decision about channel D, it might not make that much of a difference. But if we make a change to channel C, it might make a drastic improvement or marked improvement that affects the entire portfolio performance. That’s just something to think about and that’s where asking the key questions about what’s best for your business are really important.
The third letter in the methodology is C for Context. This is super important. The questions that you’re asking around context are: Do I have all the information at my disposal to make a sound decision here or is there a bigger piece of the story that I might be missing?
What if I took a step back and I thought, if I’m a stakeholder for this visualization, I may want to know how is each channel’s growth performing against the others at a single point in time? Because that view only helps me understand what’s happening across time for all four channels.
I decided to add a visualization to this set called a “target variance bar chart.” Let me show you what that is exactly.
What this does is a very simple horizontal bar for each channel, but what it does is it sets the zero axis to each channel’s respective target. The calculation that you see is the percent variance of each channel’s [performance from its own target. It creates a baseline of performance. What this allows for is a bird’s eye view of the channel’s performance against each other in a single point in time. I think this will be a great way to set the stage for a more granular analysis of the four-channel view that we saw earlier. I would lead with something like this when presenting.
And the final step in the protocol—last but absolutely not least—is A for Aesthetics. Aesthetics in data visualization is not about looking pretty, looking flashy, being snazzy, or adding pizazz. I’ve banished these descriptors from my data viz vocabulary because none of them will get you to a visual that is clear and easy to interpret accurately. This is something we really have to work with our stakeholders because they tend to ask for things in that fashion, so education is important in this part.
What you can tell them is that data visualization aesthetics is really about two things. It’s about cutting the clutter in your charts and embracing simplicity to allow your data story to come through clearly without interference.
One of my favorite tools that I take all of my private workshop students through is something called the “Data Viz Detox.” This is one of my most popular methods and I’m so excited to take you through that now.
This is what the charts could look like when I create them in Excel without changing any of the formatting defaults. Unfortunately, some of our data visualization tools add a lot of stuff that the brain doesn’t need in order to understand things. So, we have to take a few steps to clean things up. I’ll show you what this looks like after the detox.
Doesn’t that look so much clearer? I think so. I’ll quickly go through the steps of exactly what I did here. First, I reduced the bar gap width to 50 percent. I increased the linewidth to five points. I minimized extraneous decimal points, up until the point that allows me to create distinction between my data points. I removed gridlines and minimized the axis lines. I coordinated actual versus target colors, so you can see that in both charts the light gray is actually and black represents the target. It’s very important.
Speaking of color, here’s a really important point: I changed all the data to emotionally neutral grays, which allows me to send all of that data to a backdrop so that now I can use strategic color to tell my specific data story. In this case, I may want to emphasize a big win we had in May and June or I may want to use different colors to point out some loses that we had in March and April. This is really, really, powerful.
Even in my variance bar chart, I’m using color to call out, “Guys, this is where we’re struggling a bit.” This is what I call intentional use of color. Color effects the interpretation of data in our minds, but red and green are part of the pallets we often find in default settings.
This is a really important mantra to repeat over and over, please: don’t let your data visualization tool decides what matters to your decision makers. Take that power back with intentional color.
We’ve made a lot of changes. Let’s take a look at where we were before. Oh, and I do want to mention that I know I went through that checklist quickly, there will be a full data detox checklist available for you to download from ObservePoint with this webinar, so no need to worry if you missed all that. Here’s where we were before.
Here’s where we were after. Using color to even call out at the title what we want people to notice and then representing that in the chart. I hope you agree that this a much more visually clear way to understand what’s happening and answer those key business questions.
To quickly recap the PICA Protocol, we have P: as your visualization, why you exist. What questions are you going to answer? And choose the best visualization. I for Insight: what tools am I using to surface my key points quickly, easily, and intuitively? C for Context: do I have all the information I need at my disposal or is there a bigger story? and A for Aesthetics: it’s not about being pretty, it’s about cutting the clutter and embracing simplicity.
Whether you are just getting out of the gate in terms of telling your data story or you’re a veteran at presenting your data, I highly recommend this amazing book by my friend Cole Nussbaumer Knaflic, called “Storytelling with Data.” It, for me, is one of the best guides I’ve found that has distilled all of the best principles from the major data visualization books and distilled it into a really practical approach that is not dogmatic and is compassionate, and it also pulls storytelling tactics from Hollywood. Really interesting read. I can’t recommend it enough for your journey.
Here at Search Discovery, we’ve actually created a special page just for ObservePoint’s Summit attendees. If you’re like a printable handout of this presentation, a list of all of the resources I’ve mentioned, and the tips I’ve talked about, please visit searchdoscovery.com/OP17—as in ObservePoint 17. We hope to see you there. You can also learn more about how we can help you make better decisions with your data and how we can also help you present them for maximum impact.
I’d like to leave you with a little bit of wisdom from the great Aldous Huxley. And that is: “Facts do not cease to exist because they are ignored.” While that might be true, I think getting ignored sucks. I don’t want that for you. Please don’t let bad things happen to your good data. You work so hard to hunt down and distill your insights. It’s time that we stand up and pave the way for them to really start making a difference for you, your customers, and your organizations, starting tomorrow.
With that, on behalf of myself and Search Discovery—viz responsibly, my friends.
Now I’d actually like to turn it over to you and find out what kinds of challenges or questions do you have in terms of presenting your insights to get acted upon?