Calling the Data Viz Curious

Ever spot those gorgeous graphs in the New York Times or the Washington Post and wish you could make them? You can. And you don’t need fancy software to do it.

Good graphs, at their core, are based on a few fundamental principles of data visualization design, a structured sequence of steps to follow, and a little bit of knowing what buttons to push on your computer. We’ve bundled those core lessons together into the Chart Starter Series.

The Chart Starter Series is a set of 10 video-based tutorials that will change your life. I’ll walk you through my tried-and-true, works-every-time, structured process for making great graphs. I’ll talk you through choosing the right chart type and how to format your chart so your story shines through. And I’ll show you how to do it all in Excel. Yes, Excel. Microsoft Excel. The program you already have on your computer.

You won’t learn how to make interactive dashboards in Excel (though we show you how to do that in the Evergreen Data Visualization Academy) or how to run trig functions (more power to you if you can). You WILL learn how to use the tools you own to make high-power, newspaper-worthy visuals for your slides, handouts, and reports.

Through January 15, 2020, you can take the Chart Starter Series for 15% off. Just use the code starter15 when you enroll.

If you have been curious about how to get started graphing, this course will take you to the next level. You’ll be the one offering gentle suggestions to other people at the office. This is how you become to go-to guru.

PS. Later this Spring we’ll release a Chart Starter Series for Tableau and R, too.

Decade of Data Viz

Well, that just flew by. Ten years ago, I was in the throes of writing my dissertation on Presenting Data Effectively, knowing I had focused on a topic that the world was hungry for, even if they didn’t know it yet. I had collected enough literature and data to know that a better way was possible, despite the fact that there were no viz conferences, journals, or twitter channels to serve as repositories for me. I had Tufte, Reynolds, and Few, none of whom cited research to form their opinions, and a lot of published work spread across disjointed fields to pull together. But even that much made it clear to me that the way we were talking about our data was boring AF and it was a major impediment to use.

Even so, I was presenting at conferences (on other topics) and trying SO HARD to convince my academic co-presenters to try something new. It was hit or miss.

I, in fact, presented this total Death by PowerPoint a decade ago:

I successfully convinced other colleagues to let the visuals work their magic, so that same year, I also presented this slide:

Which I now use as an example of terribly cliché imagery that undermines our credibility.

Looking back at the graphs I was producing a decade ago, I can see that I was already incorporating what I had learned in my research:

Clean lines, gray plus an action color… I was figuring some thing out. But I had not yet figured out how to directly say the insights in the data. I was still assuming others would just look at the data and figure them out.

In another example from that same time frame, it is obvious that the lessons on color had not consistently hit home yet. It looks like I was decorating a kindergarten classroom.

I can see that at this point I knew enough to answer questions with clear, headline-style statements. But I didn’t use color to connect the headline to the graph. And I hadn’t yet realized that the sort order of the graph was an important storytelling tool. I clearly only had bar charts in my graphing repertoire but didn’t know I needed to make those bars thick and sexy for maximum impact.

My design evolution can also be tracked through my annual reports, though they aren’t quiiiiiite a full decade:

Some things – like using a grid – stay the same. But I learned more about chart types, supporting narrative, colors (yikes), and design. I can’t be too hard on myself for what I was doing a decade (ish) ago. When you know better, you do better.

The Stephanie Evergreen of 10 years ago had a pretty good idea that the data viz world was about to explode, though I had no idea it was going to be so cool. We have tools! We have conferences! We have books and checklists and workshops! We have even reached that initial peak of “how cool is data viz”  and moved on to important issues like “how does this help bring social justice” and “how can I make this more accessible” and its really been something to see.

Particularly in this climate where there’s so much to explore and learn, it is difficult to pause and look back at what was happening 10 years ago. But it is so important to do so because it brings about this notable mix of embarrassment and pride. Did I really think that was awesome? Heck yeah, I did. I think the same thing about what I’m making today. Which means, I wonder what my graphs will look like in another 10 years.

If you are reading this, thinking to yourself, “Geez the stuff I’m making right now looks like what Stephanie Evergreen was making a decade ago,” don’t get sad, get hopeful. Everyone evolves. Keep going and keep growing.

Four Ways to Show Projections

Of course we all want to know what will happen in the future. These days folks are looking at data like it has a crystal ball. To the extent that we provide our audiences with projected data, let’s talk about ways to visualize the projected data. Because here’s the thing: The data that isn’t real yet can’t look like the data that is real.

It needs to be immediately clear that there is something different about some of the data so audiences can give the projected data the proper mental caution. Here are 4 ideas. You might even think of others!

Note: This is not about showing uncertainty, such as confidence intervals. I discussed that over here.

Dash That Line

While I usually recommend against any sort of pattern fill in our graphs, this is the one time when it could make sense. Changing the projected line segments from solid to dashed match the story because the dashes indicate “incomplete.”

Fade It

Similarly, a lighter color can be read as “not as strong yet.” So fade the projected data by just using a lighter shade of what you used on the rest of the graph.

Gray It

If you end up with a lot of data in your chart and the thought of changing a lot of individual line segments into a lighter color makes you want to get up and go for a walk, well, go for the walk. It’s good for you. Then come back and add a semi-transparent overlay to those sections as another way to mark that data as different.

Range of Possibility

When your projected data could fall within a range, show the entirety of the range in the graph. I made this one by combining a line graph with an area graph. As always, be sure your title is strong, clearly telling your story.

Whichever method you use (maybe one you are now inspired to invent), you gotta do SOMETHING to make it clear that the not-yet-real data is visibly distinct.

Better Graphs Tell Clearer Stories: The Breast Cancer Example

Half of getting people to become more aware of issues like breast cancer and change their behavior is to make an emotional pitch that gets in their guts or their hearts. Recent ad campaigns, like Know Your Girls from the Ad Council, have been really successful here.

Ad showing three Black women with copy that says "How well do you know these girls?" which is positioned around one woman's breasts.

But the other half of raising awareness and creating behavior change is appealing to people’s logic using data. And we can do so much more here. Research related to increasing awareness of climate change tells us that we win people over when we speak to both their emotions and their logic. Breast cancer awareness needs more of the hard data logic. 

Susan G. Komen does have some data on how breast cancer disproportionately affects women of different races.

Graph of breast cancer incidence rates from 2012-2016, by race.
White 130.5
Black 124.0
Asian/Pacific Islander 100.1
Hispanic 97.2
American Indian/Alaska Native 79.5
Data available at linked URL

If you dig deep enough into the data section of their website, you’ll be able to find data on how breast cancer rates breakdown over time – and there’s a profound story in here, particularly if we could figure out how to present it better. Can you see the story?

Breast cancer incidence and mortality rates over time for Black and White women. Data available at linked URL.

Let me spell it out: This graph is showing that, while white women have a higher incidence of breast cancer than black women, black women are more likely to die from the disease. 

Now that I say that, can you see it in the Komen graph? It is easier, right? And that’s why strong, courageous, clear titles are important. But that’s not all. A better graph type on the same data would paint a picture that better matches this urgent, important story.

Overlapping bar chart of breast cancer incidence and mortality data for Black and White women. Data available at previous link.

An overlapping bar chart is a more straightforward way of graphing data when the two variables are inherently interrelated, like incidence and mortality. This better captures the relationship, showing that those who die from breast cancer are a subset of those who contract the disease in the first place. The larger set of blue bars in the graph for Black women is key here.

If we wanted to zoom in on the racial disparity angle of this story, we could take this even further. You see, with the overlapping bar chart, it is easy to spot the relationship between incidence and mortality rates for each race but it is not so easy to compare incidence rates between Black women and White women because (1) the data are in two separate graphs and (2) the scale stretches all the way down to zero so as to include the mortality data. I mean, the gray bars don’t look all that different, do they?

Let’s reshuffle the data so incidence rates are paired together and mortality rates are paired together, using scales that are more appropriate for each set of data. And this time, let’s use a graph type that better tells stories about disparity – a dumbbell dot plot.

Two graphs showing the disparity in breast cancer incidence and mortality rates among Black and White women. Data available at earlier link.

When we cram the data on the same scale, we make it difficult to see the important disparities in each data set. Adjusted scales, appropriate graph types, and clear titles (and – heaven help me – losing pink) create a set of visuals that tell the story about breast cancer disparity with the concern and urgency that are warranted.

We teach folks how to make smarter graphing choices in our Academy and Graph Guides programs.

Learn in the Academy!

You can find step-by-step instructions on how to make 60+ awesome visuals in my Evergreen Data Visualization Academy.

Video tutorials, worksheets, templates, fun, and a big-hearted super-supportive community. Learn Excel, Tableau, R or all three. Come join us.

Enrollment opens to a limited number of students only twice a year. Our next enrollment window opens April 1. Get on the wait list for access a week earlier than everyone else!

Master Dataviz with Graph Guides!

Our newest program, Graph Guides, is a custom-built, year-long sprint through 50 Academy tutorials.

When you enroll, we’ll assess your current data viz skill set, build you a customized learning path, and hold your hand as you blaze your way to new talents.

We open enrollment to 12 students at a time and only twice a year. Get on the waitlist for early access to our next enrollment window.

When Will Bad Infographics End?

Back in November of **2010** I set up a Google Alert for “infographic.” Oh, I thought, these are an interesting development in how we communicate data. But either the quality of the design will drastically improve or these things will be a fad that disappears within a couple years.

Nine years later and it is clear that both of my predictions were wrong. My Google Alert for “infographic” contains hundreds of hits every single day and the designs are seriously bumming me out.

Let’s take a spin through a random selection of the infographics that landed in my inbox. I don’t mean to shame individual designers here, so I’m purposely not linking back to the original post nor including authorship.

This first one was so long, I had to max my zoom out and this is still only a portion of it. Just be a website.

And note how it is heavy on the graphics but light on the info. That’s become a mainstay of the bad infographic. Here’s another example of this trend:

It feels “over-graphicked,” you know? And the graphics aren’t even good! Look at that 3D pie chart. Come on.

Weak data visualization is one of the biggest faults that undercuts an infographic’s viability. This one uses hard-to-notice donut charts and bar charts with a disappearing baseline (if you can even spot them).

One might be tempted to blame weak data visualization on graphic designers, who tend to avoid a deep understanding of the topic. A graphic designer with a strong data visualization game is as rare as a unicorn. But this infographic also violates *basic* graphic design rules like color contrast.

An increasing number of the infographics in my daily digest from Google are just a single graph, like this one:

And, again, the design needs some love. The legend is in sentences at the bottom of the figure, for cryin out loud. But is this even an infographic? Or are we calling single graphs infographics now to get on board the trend train?

Why Weak Infographics Are Still Happening

This is actually a sign of success. The proliferation of infographics means that the world is waking up to the need to effectively communicate data. Can I get an amen? We are democratizing data and this is a good thing.

And that means that a lot more people are tasked with making infographics who do not have backgrounds in graphic design OR data visualization (when what they need is both). These are folks who have jobs like data analyst or marketing associate.

With the democratization of data we also need a correlating democratization of data design. A lot of people need models, step-by-step guidance, tools, and – most importantly – the strategic thinking that helps one know how to tell a story with data.

Thankfully, we are seeing more and more classes emerge that teach folks how to DIY good infographics, but their rise has not kept pace with the need. The Data Visualization Academy is a good place to start.

What Makes a Good Infographic

  1. A balance of narrative and graphics. You can’t convey information without some words. And you can’t qualify as an infographic without some graphics.
  2. Adherence to graphic design best practices. That means attending to stuff like page layout, reading flow, color contrast, and font size.
  3. Adherence to data visualization best practices. Don’t over-rely on showing data as a percentage in a large font. Learn graph options. Don’t distort the data or make it hard to read the graph.
  4. Coherence. Meaning, the infographic should provide some insight in a story-ish fashion. All the parts should hang together into a whole that teaches the audience something.

This is what I’ve got so far. What else would you add to this list?

We do not often make infographics (we prefer to make handouts) but if you want to contrast the ones in this post to others that are more effective, check out our examples here and here.

Excel vs. Tableau vs. R

We are hard core believers that you should become the master of the tools you own. If your company relies on Microsoft, figure out how to use Excel to make amazing data visualizations. If your company invested in a site-wide license for Tableau, climb over that learning curve and master it. If you are working within a round number in your budget, like ZERO, and relying on free software such as R, become its master. Work whatchur mama gave you.

It doesn’t matter to us what software you use, so long as the visuals you create from it are awesome. We show folks how to make awesome graphs and dashboards in all three of those platforms at the Evergreen Data Visualization Academy and in our Graph Guides program.

We know from experiences developing graphs and dashboards in multiple platforms that each program has its strengths and its confusing idiosyncrasies that make it a pain in the neck. It’s funny, some people like to think “oh this software makes doing X really hard” so they’ll switch to another software where now X is easy but Y and Z are mysteriously horrid. Every program has its trade offs.

To help you know what you are getting in to, at the Academy, we give every tutorial in a ninja level rating. That’s our way of telling you how hard it is going to be to make a particular graph in the program you chose. A 1 means it is super easy to make that graph, and a 10 means it is pretty hard (but possible!).

The table below lists our ninja ratings for some of the many tutorials in the Academy. Listen, ninja ratings are subjective. And rated by those of us who have some experience at this, thinking about what it would be like for a newbie. Even with those caveats, you can sort and compare the ratings to see where things are harder and easier in each program.

TutorialExcel Tableau R CodeR
Advanced Dashboards779
Auto-Populate Office Documents3coming soon7
Back to Back384
Bar + Vertical Benchmark822
Beeswarm833
Beginner Dashboards679
Bullet Chart725
Bump Chart437
Choropleth, Tile, and Hex Maps3410
Connected Scatterplot367
Customize Colors222
Dashboard Automation9coming soon9
Dashboard with Filters749
Diverging Stacked Bar Graphs946
Dot Plot884
Dynamic Titles844
Embed Legends213
Filled Intersect Line Graph869
Gauge Chart663
Horizontal Dumbbell Dot Plot9105
Horizontal Lollipop534
Interactive Heat Maps839
Line to Area133
Lollipop Gantt764
Nested Area Graph3coming soon8
One Page Handout357
Overlapping Bar Charts547
Pictograph384
Projections257
Pull In Pictures834
Sankey858
Setting the Scale521
Slide Handouts32not possible?
Slope Graph575
Small Multiple Bars327
Sparklines & Indicator Dots178
Tables225
Basic Bar125
Treemap426
Using Icons353
Vertical Dumbbell Dot Plot4104
Waffle Chart1055
Waterfall348

When folks join the Academy, they watch a few of our orientation videos, learn my 4 step process for making graphs, and then dive in to lessons with a ninja level they are comfortable with, tackling more complex tutorials as they gain more skill. Everyone gains several ninja points with us, sometimes racking up points in different software.

In our Graph Guides program, we build you a customized path through these tutorials (and the dozen or so we’ve added since we published this post) and test you at the end – you master graphing in one year.

Want in on the action?

Learn in the Academy!

You can find step-by-step instructions on how to make 60+ awesome visuals in my Evergreen Data Visualization Academy.

Video tutorials, worksheets, templates, fun, and a big-hearted super-supportive community. Learn Excel, Tableau, R or all three. Come join us.

Enrollment opens to a limited number of students only twice a year. Our next enrollment window opens April 1. Get on the wait list for access a week earlier than everyone else!

Master Dataviz with Graph Guides!

Our newest program, Graph Guides, is a custom-built, year-long sprint through 50 Academy tutorials.

When you enroll, we’ll assess your current data viz skill set, build you a customized learning path, and hold your hand as you blaze your way to new talents.

We open enrollment to 12 students at a time and only twice a year. Get on the waitlist for early access to our next enrollment window.

Audience Engagement Strategies

Dr. Sheila Robinson is a master at engaging an audience. She’s honed this skill through decades of work in education where you have to know how to take the pulse of a room and determine whether a change in direction is needed to keep the group tuned in. You might think I’m talking about middle school students but teaching teachers is just as rough. We know this to be true first hand.

That’s how Sheila and I bonded – our mutual background in education and our shared interest in audience engagement strategies.

Even though Sheila and I are both done with our days ruled by the dismissal bell, we still both use our teaching skills REGULARLY in our workshops. In addition to her own workshops on survey design, Sheila has joined Evergreen Data and now delivers data visualization and reporting workshops for me. We are typically on opposite sides of the country at the same time.

One of the things I love about how well Dr. Robinson represents Evergreen Data on the road is her masterful ability to keep an audience engaged for 6 entire hours. She wrote up a guide on 20 audience engagement strategies that we incorporate in my workshops all the time.

She can sense when it is time to break from the lecture and start an activity, like a Turn and Talk.

She can smell in the air when people just need to talk to their neighbors for a minute, so she’ll launch something like a Think-Pair-Share.

She intuits that, mid-afternoon, audiences need to get up and move as part of the workshop, so our workshops often include a Gallery Walk after a slide makeover activity.

Sheila keeps our workshops deeply educational and highly interactive because we know that those things, put together, create life-altering professional development.

Find out more about a workshop with Sheila (or me – hey, I’m pretty good) and book your next high-impact skill-building session on our Workshops page.

The Dashboard Sketch Process

“I vote no.” This short answer speeds up the dashboard development process significantly. And I developed the question my client voted on in roughly 3 minutes, by showing her a quick sketch of some possible graph options.

My client is in charge of a dashboard that tracks how high school students in Mississippi apply for college financial aid. She built a simple dashboard in Tableau but brought me in to refine the design and rethink the graph types.

In total, the dashboard had 5 metrics. So I used a page from The Data Visualization Sketchbook – the dashboard template with space for 6 metrics – and I sketched out what I thought the 5 graphs should look like.

In the sketching process, I realized I had outstanding questions like “where is the data for this metric?” Tiny stuff, you know 😉 And I made note of what to ask my client in our next email.

This stage of the sketching process was strictly for me, to help me get organized. It took me 15 minutes. I emailed my list of questions to my client and opened Tableau to build everything else.

Most of the metrics are simple, just showing change between two years, sometimes with associated benchmarks each year. There are LOTS of ways to visualize this kind of data. So as I’m thinking through the logistics of building the primary graph for the dashboard, I realize that I want to provide the client with several options of graph type, so she can decide which one is easiest for her to read. But some of the graph options I want to show her require a total overhaul of their in-office workflow, completely changing how they plug the data into the Tableau data source. My goal with dashboard development is never to make the work life of my client more complicated afterwards.

Before going down that path, I whipped out The Data Visualization Sketchbook once more and turned to a graph grid paper page. I sketched out the four graph types I was considering and sent a pic to my client, along with notes about each graph type’s readability and implications.

This time, my sketching was for quick feedback from the client, before I sunk time into data manipulation processes that they might not even desire.

She replied within the hour with this:
Graph 1: I really like this idea.
Graph 2: Works.
Graph 3: I vote no.
Graph 4: I would rather not use this one either.

AWESOME! Two possibilities she likes. One of which does NOT require major rearrangement of their workflow. Decision made. Onward.

And this is exactly why I published The Data Visualization Sketchbook. Because sketching saves so much time and money. We iterate through options and land on the right answer so fast.

Everyone’s Opinions About Your Dashboard

The primary struggle with dashboard development is not identifying the right key performance indicators, building the graphs in your favorite software, or even getting people excited about the idea of using a dashboard. That’s all a cake walk compared to managing people and their reactions to the dashboard.

Let me tell you a couple stories about what is happening with some current clients. I guarantee you will nod your head in agreement.

I’m coaching the data team of a large nonprofit as they venture, for the first time, into creating a data dashboard. It is supposed to show data for program locations around the United States. They are building it in Tableau with the talent of some awesome interns. We have a weekly phone call where they show me the dashboard-in-progress and I give feedback about design tweaks. Then, last week’s phone call was not about design. It was about people.

They said, “The dashboard development process seems to be going pretty well and everyone is excited to start using it. But the problem we have is that whenever we show a draft to the rest of the staff, they all have their OPINIONS.”

Oh girl, indeed they do. That’s when I went from Dr. Evergreen the data visualization expert to Dr. Evergreen the therapist.

When it feels like everyone has OPINIONS about the dashboard, it usually means we haven’t prepped them properly for the review they are undertaking. Here’s my advice:

  1. Be very clear that the draft will have typos, misalignments, and mismatched colors. The purpose is not to look for those things at this point. They will be corrected during the next design phase. It might be hard, but ignore them. (If you don’t state this upfront, you WILL get a dozen points of feedback about the same typo and you will NOT have received feedback you really need at this stage.)
  2. Send the draft with specific review questions. The best questions are related to the actual use of the dashboard. Can you do the things you’ll need to do? Can you access the information you need to make decisions? Is any data missing? Is there a way of seeing the data that is missing? I also like to add in questions like Should these y-axes be the same or different? and How many decimals are necessary in this data label? (Without these guiding questions, you’ll get feedback about the look, not the substance, of the dashboard.)
  3. Speak directly to the purpose of the dashboard and its intended users. Not everyone who is reviewing your draft is likely to be the end user. Or, maybe they are – but they have different uses. This point deserves some elaboration with a story from another client.

He told me he has 4 different company vice presidents who all want a dashboard. They have different roles, different interests, different burning questions. At first, he thought he could make one dashboard that would satisfy them all. LOLOLOLOLOLLLLLLLLL no. That will never happen. The sooner you realize different audiences need different dashboards, the less time you’ll waste trying to make everyone happy and the less therapy you’ll need.

Stephen Few said dashboards typically have one of three possible purposes:

Strategic – used by management, 30,000 foot view of key performance indicators, updated weekly or monthly

Analytical – used by analysts, drill-down ability to explore the details

Operational – used to keep a pulse on organizational behaviors, updated constantly

Sure, there might be some overlap in purposes and audiences but most dashboards fall into the trap of trying to do all things for all people. Even in the case of my client with 4 VPs, though they might all be at the strategic level, they need to see totally different metrics, for the areas they oversee.

So in the dashboard coaching-turned-therapy session, I also recommended identifying the purpose and intended audience for this dashboard, noting that other dashboards for other audiences could come in the future (ahem, job security) and asking any reviewers to only review from the perspective of the identified intended audience.

Establishing these three parameters at the beginning of a dashboard review process helps manage people, their expectations, and their reactions so that they stay focused on the substance of the dashboard, rather than turning into instant design experts.

My therapy bill for this 5-minute read is $250 (invoice in the mail) but you’ll save yourself thousands in time and headaches.

Data Fortune Tellers

Pick a color, any color. You know how this game goes. You pick a color, your friend opens and closes the fortune teller, spelling out the name of the color you selected. You pick a number from the visible choices, your friend opens and closes the fortune teller, until you eventually pick a flap that your friend opens to reveal your destiny.

When I was a kid, we wrote stupid fortunes like “Marry a rich husband” and “Have 10 children,” both of which would make me run away screaming. But what if our fortunes had been more like “Graduate with a degree that will earn you $14,000 more than your peers in your first job” or “Land an engineering scholarship”?

Those are the fortunes available today – scratch that – those are the futures readily available today to the young women and men who take part in FIRST®. Ever heard of FIRST? FIRST hosts robotics competitions for kids and creates an inclusive experience where kids of all ability levels learn by doing and have a lot of fun at the same time. In the long run, they are bridging the STEM skill gap.  How cool is that?

So FIRST just completed year 5 of a multi-year study where they tracked their program participants into college and HOLY COW did they find out that these students were doing some cool stuff. FIRST participants are much more likely to land a STEM-related internship than the comparison group (26% vs. 15%) by their sophomore year of college and are more likely to be involved in computer and engineering clubs and competitions in college than their peers. 89% of FIRST alumni declare majors in STEM versus 59% of the comparison group.

Slides designed by FIRST staff based on our template

And they are seeing a pronounced impact on young women. Female alumni are 5.3 times more likely to take computer courses and 3.7 times more likely to take engineering courses in college than their peers. 59% are declaring majors in engineering or computer science versus 12% of comparison group females!  

Naturally, FIRST is really proud of these results and they hired Evergreen Data to help them package up their findings into a slideshow and handouts they could distribute to parents, principals, and coaches. And students? Yeah, no. Sitting seventh graders down for a slideshow on the likelihood of landing a tech-related summer job is not going to get them excited to stay after school and build robots. We needed a different way of reaching the students. So we made them FIRST Future Tellers. FUN!

Each potential future came right from the data in the study.

The Future Teller, by nature, must be made from a square so we used the rest of the real estate on the page to talk a little about FIRST and provide some directions.

Sometimes physical objects are the most effective way to engage key audiences with your data.

A FIRST staffer has already been putting this into practice:

“I used the FIRST Future Teller as a warm-up activity with students, parents, and team mentors before starting a presentation reviewing the key study findings. The audience partnered up to play with the future tellers and the energy level in the room grew significantly. I think many of the audience were more interested in the presentation as a result. It was a fun way to present findings. One mentor asked if he could bring the future teller back to his classroom to use with middle school students!”

Customize your own Future Teller here.

Just download the file, write in your own futures, print, and play with data.

From the blog