Three-and-a-half lessons learned from network diagrams

Once a month here in Atlanta, we invite a few folks from the data community together to discuss the "data value chain" and sharpen each other's thinking in the area of using data better. In a recent gathering we were discussing the merits and challenges of network diagrams. The stake I firmly planted in the middle of the table was this: for the vast majority of problems that folks have to deal with, network diagrams don’t help. Ever.

Ok, so maybe that was a little harsh. And as we discussed it, I had to soften my position. We concluded that there are most definitely situations where network diagrams can be successfully used. Here’s what we uncovered.

When most people think about network diagrams, this is what first pops into their heads:

Simple Network Diagram

It’s great for showing the hierarchy that would otherwise only be represented in some sort of over-bloated frankenstein of a table. And I think it works pretty well for a situation with a finite number of nodes that represent physical elements that can readily be counted such as “number of routers.” This is our first lesson:

Lesson Learned #1: If you can reasonably count the nodes, a network diagram can reasonably add clarity about relationships.

So, the concept of a network diagram feels like it makes tiered data more accessible. But let’s look at more complex relationships. Take your LinkedIn network. There are lots of layers of relationships that it seems a network diagram would seem to make sense of. In case you missed it, a couple of years ago, LinkedIn Labs made network maps available to LinkedIn members in their InMaps. Here’s mine:

LinkedIn map

It is beautiful. They’re using a Gephi-inspired in-house development to lay out the nodes, chose the colors and stuff (if you’re interested in more on this topic, check out the Quora post - oh yeah, that guy Sal Uryasev who worked on creating inMaps is a former Juicer. Nicely done, Sal!).

I love, love, love the groupings. In my opinion, this is the most useful part of the layout. At this number of nodes, it’s not the individuals that are meaningful, but rather how those nodes group together. The approach Sal et al. used nicely summarizes a good portion of my career in about 5 large chunks such as “7 years on the roller coaster” at a dot com, and “Todo: attend a reunion” for connections I made while at Georgia Tech (the labels are mine - wouldn’t that be cool if InMaps could do that?).

But, as far as network diagramming goes, you’ll see that they’re just plotting the first-generation relationships (of which I have 500-ish) and it’s still fairly dense. Imagine what would happen if second-generation+ relationships were added (there are supposedly 11 million “in my network”). Yuck. So here is our next lesson:

Lesson Learned #2: Network diagrams with many nodes are most useful when showing aggregated groupings and relationships.

And the corollary this quickly brings us to:

Lesson Learned #2.5: When many nodes are aggregated into a few relationships, network diagrams can be used as a presentation medium. Otherwise, stick to exploration.

Ok, we have time for one more lesson. Here’s another example offered by a small company you might have heard of:

If you think about it, this is nearly the perfect problem for a network diagram to solve: making it easy for a person to find images similar to the one they’re looking at. But, this offering, inspite of it’s well crafted-ness, went nowhere.

Why? Well, one reason might be because those of us who are visual pundits would love to see these complex relationships simplified by just the right visual representation. But the fact remains that for the vast majority of people out there, advanced visualizations are just not enticing enough -- and too complex feeling -- to incite broad use. There, I said it. So, finally:

Lesson Learned #3: Even for relationships that “normal people” can easily understand, network diagrams aren’t easily traversable by “normal people.”

So, there you have it. Three-and-a-half lessons we’ve learned with network diagrams. Apply them to your next network display challenge and see how they work for you. If you need some technology to help you, check out the wikipedia article on network diagramming tools. Let us know if you find any that reveal other lessons to you.

The Best Product Manager: Hustler, Designer, Hacker

Much of what makes a great product manager is empathy and a desire to serve others. Tulsi demonstrates these qualities better than most I’ve come across.  As you will see below, her passion for design as part of product management is only surpassed by that for her customers, products and causes.  Oh, and there is usually much laughter involved. Enjoy and feel free to reach out to her at

Even after years of product management experience at several companies, I still get frustrated when folks frequently say “So, you are a Project Manager”. I usually respond with a vehement “No!” and go on to describe what it is I actually do everyday.

With this in mind, let’s begin this discussion by describing what a Product Manager is.  A (good) Product Manager is the champion of the customer and the market: part product visionary and part liaison officer between external and internal needs, pressures, and limitations.

As Catherine Shyu, Product Manager at Send Grid puts so nicely:

“Much of a Product Manager’s responsibility is to juggle multiple streams of conversation and move them towards closure.”

Successful disruptive and innovative brands like Basecamp, Airbnb, Fab, and many others have proven that features alone don’t improve the sales. Instead the infusion of design and love into the products is what creates real customer engagement and advocates. That’s why many of the companies mentioned have consolidated Product Management and Design into single roles or departments. Now the Product Management role is evolving even further.

In the words of Gary Tan

“The ideal startup team consists of: a designer, a hustler, and a hacker.”

The most successful Product Managers I’ve worked with and learned from seem to embody the qualities of all of these three roles. Just consider what these roles bring to the table:


This role can seem as nebulous as the Product Manager’s, so it’s no wonder they’re coalescing. Whatever the type of designer, success is based on the ability to emphatize,  perceive deep customer needs, and anticipate customer behaviors.

That’s why Product Managers with design and usability skills are able to create experiences rather than the features, simplify the interactions, and sketch and wireframe ideas to tell stories that others can understand.


Contrary to any negative (and possibly cheeky) connotations, the hustler knows the market, knows how to sell, and knows how to work with what they have to turn a profit. In other words, she knows how to connect products with customer and market needs. The hustler’s skills can help a Product Manager think beyond product design to the critical marketing and sales activities that will make products and companies thrive.


Hackers can think creatively, come up with solutions quickly, and iterate through problems they encounter along the way. They are also curious about technology and how things work. Hacker instincts help Product Managers communicate well with engineering teams, and work lean to get the best possible outcomes with the least possible time and resources.

The bottom line: the days of the traditional product manager are gone. Lines are naturally blurring around the Product Management role and discipline, and that’s a good thing! The better you are at blending these three roles, the more equipped you will be at juggling the responsibilities that are on today’s product manager. So, hustle, design, and hack your product into shape. And then tell somebody what you do!

Many thanks to @Imusicmash and @apmcinnes for their comments and feedback.

Explore Sochi Olympic Medal Results

We’re huge sports fans and the Winter Olympics in Sochi is just the beginning of a busy and exciting 2014 sports year. If you’re like us you may not have a few hours a night to hang out with Bob Costas, so we created an interactive summary dashboard of medal results. You can view results by country or event and drill down to the individual athlete if needed. Come back often to see the updated results. Enjoy!

Keep an eye out for visualizations we plan to do for March Madness, MLB Spring Training, NFL Combine, and the FIFA 2014 World Cup.

Extreme Makeover: Edition

Have you ever watched one of those miracle-home-improvement shows where they take a house that is a good foundation, but that has been neglected for a bit too long? Nothing’s better than real world practical examples. So, we thought we’d take that approach and apply it to an existing report from the real estate industry and do a little makeover to see if we could make a dramatic improvement. You be the judge.

We found a particular report on and we said to ourselves "Self: this data shows a lot of potential!" We really loved it for a few reasons: There is a lot of great information, over 100 different markets, expert commentary and pretty interesting to anyone owning a home or investing in real estate in the U.S. Even still, it feels lacking. What if instead of just making the data available, this report answered some specific questions on the minds of homeowners and investors as well as provided it to them in an easy way to consume this rich information?

Selecting a real estate example wasn’t completely random for us.  With friends and family at Colliers, TelesIntelligence, Berkshire Hathaway Home Services and Keyes, we know there are a lot of opportunities to make real estate data more valuable.

Here is the link to’s existing report. Our version of the report can be found here. We downloaded the .csv file that they make available and took it from there. We applied some Juice design principles and help the data answer some specific questions.

Our approach was threefold:

  • Make the report more readable and attractive
  • Offer the reader more guided exploration
  • Offer visualizations that permitted comparisons across markets

Before we did anything we did do a little transformation on the data. We added a dimension for region and broke out city and state into separate columns. This permits another layer of data exploration.

Below are some screenshots of our report makeover.   You can see, and interact, with the makeover version here.

First, we gave the commentary section of the report a little help. Using the Simple Font Framework, we improved the titles, highlighted some of the key trends, downplayed some of the contextual information. Just below that we provided some overall metrics, so that users could compare markets to the U.S. overall averages.

Next, we posed questions that we thought the audience might be most interested in knowing and then applied a visual that would answer that question. In the case of the map the user can toggle the metrics between the month to month and the year over year change in median price. As you notice Maryland stands out on the map, while finding this on the original report takes a little effort.

As with most decisions, homeowners and investors can’t rely on just a single measure to inform their opinion. The leaderboard visualization below allows the reader to view rankings across multiple metrics. In the highlighted example below we compared Denver, Colorado to Seattle, Washington.

The finished version (found here) also provides some additional visualizations and a global filter at the top, but you get the idea. Not brain surgery or something we would use to make actual real estate investments; however a quick 60 minute makeover to call out some of the beauty trapped in this data.

So, what do ya think? Worthy of the Extreme Makeover moniker?

Reading Visualizations for Beginners

A skilled author of data presentation will choose the right visualization to emphasize a message. The data, chart, and supporting descriptions will work in harmony to point out what is interesting. The reader simply goes along for the ride. Unfortunately this is the exception more than the rule. Many data products are a muddled mess of chart choices, obscure labeling, and arbitrary layout. In essence, the author has passed responsibility to their audience to find the meaning.

If you are to carry this burden of rooting out the insight in a data visualization, you need to know where to look. The best place to start is by focusing on the unexpected. Does the world work the way you think it does? Or does the data show you something that challenges assumptions of expected values? Let's take a look at a few ways to find the unexpected.

Unexpected distributions

Pie charts are designed to show how something breaks into its constituent pieces. The slices add up to the whole, and the volume of each slice indicates its piece of the pie. The primary insight offered in a pie chart comes from slices that are smaller or larger than you would expect. One weakness of the pie chart is that to discover slices that are bigger or smaller than expected, the reader needs to compare the actual chart to what they imagined it might look like. For example, in this pie chart the reader might be surprised to find that confections are nearly half their diet by volume (that’s not healthy eating).

Unexpected patterns or relationships

Plotting data in a scatterplot or bubble chart is a way to show relationships between two or more variables. The pattern of the points may express a correlation that is either expected or surprising. Furthermore, outliers from this pattern are interesting because they break the mold.


Source: Data from Natural History Magazine, March 1974

This scatterplot shows animal size versus weight. The data indicates a positive relationship between size of the animal and its top speed. Bigger is faster, but with a lot of variation. The cheetah is an outlier with an unusually high ratio of speed to body mass. That's interesting. (Also, it’s good to see that humans are faster than bears. Unfortunately, a careful reading of the underlying data reveals that the human data point is Usain Bolt, world record holder in the 100 meter sprint.)

Unexpected trends

Trends across time are another common place to look for insights. Line charts can make obvious the deviations compared to expected patterns or trends. Like the pie chart, the reader needs to overlay their assumptions on the shape of the lines. Do you expect there to be an upward trend? Should the values remain steady over time, or is it normal to see substantial fluctuations?


Cardiologists use Electrocardiograms (aka EKGs) to trace the electrical activity from the heart. A healthy heart demonstrates familiar patterns in the lines; changes to these patterns indicate problems. An experienced cardiologist can see abnormal heart rhythms, chamber enlargement, and signs of impaired blood flow through changes in the shape of the lines.


Data without context may offer little meaning. But adding a comparison value—whether an industry benchmark, an organizational goal, or a regulatory standard—brings values into focus. Comparisons across time periods can communicate improvement or regression. Direct comparisons can show how two or more entities rank compared to each other. Numerous specialized data visualizations have been designed to enable quick comparison, including bullet charts, “stop-lighting,” and leaderboards.

This dashboard compares bank brands by a series of survey questions. Rankings and side-by-side comparison make it obvious who is performing better for each brand performance measure.

Find a starting point

A dashboard, report, or data visualization can feel like an ocean of information competing for your attention – like a Where’s Waldo™ puzzle. Rather than trying to take in the whole picture at once, it’s a good idea to focus your attention on a small piece of the picture. Focusing on a single element can help you grasp the nature of the data, the dimensions and metrics being displayed, and eventually how a small piece fits into the whole. Take the following data visualization comparing hospitals by patient experience as an example.

There is a lot going on here. The bubble chart shows three separate metrics about each hospital. The meaning, size, and positioning of the bubbles requires a new reader to carefully review the axes and legend to get his bearings. The connection between the bubbles and the bar chart on the right is not immediately obvious.

It’s a lot easier to tell the story of a single bubble.

The highlighted bubble represents a hospital, and the three metrics used to size and position the bubble are shown in the tooltip. In fact, the connection to the bar chart becomes obvious as Russell Hospital is identified as one of the largest hospitals by bed size. This particular data point may not be the most interesting or unexpected in this chart, yet now you how a much better sense of what the data product is trying to convey about hospitals.

Turning Data into Words

It is often said that you know you have become fluent in a foreign language when you dream in the language. Short of this inflection point, language students have to translate the words back into their native tongue. And so it is with data. Without instant recognition of the meaning of a data visual, it can be useful to convert the information into a language in which you are familiar.

Take the example above: to understanding the message in the data, you might translate a data point into a descriptive sentence such as “Russell Hospital has 730 beds, tied with three other hospitals” or “Russell Hospital’s patience experience score is near the top of all hospitals shown.” This is a way of capturing and testing your understanding of what you are seeing in the data.

By breaking a complex data product into its smallest pieces and finding something comprehensible, you will start to understand both what the author is trying to show and how to read the content.

Data Storytelling, The Pixar Way

Pixar's Rules for Storytelling -- as shared by Emma Coats, Pixar’s Story Artist -- are almost as relevant for communicating with data as they are to filmmaking. Whether you are a creator of data-rich presentations, infographics, or data visualizations, here's an abridge list of the most relevant Pixar principles for data storytelling: #2 You gotta keep in mind what's interesting to you as an audience, not what's fun to do as a writer. They can be very different.

#3 Trying for theme is important, but you won't see what the story is actually about til you're at the end of it. Now rewrite.

#5 Simplify. Focus. Combine characters. Hop over detours. You’ll feel like you’re losing valuable stuff but it sets you free.

#8 Finish your story, let go even if it’s not perfect. In an ideal world you have both, but move on. Do better next time.

#11 Putting it on paper lets you start fixing it. If it stays in your head, a perfect idea, you’ll never share it with anyone.

#12 Discount the 1st thing that comes to mind. And the 2nd, 3rd, 4th, 5th – get the obvious out of the way. Surprise yourself.

#13 Give your characters opinions. Passive/malleable might seem likable to you as you write, but it’s poison to the audience.

#14 Why must you tell THIS story? What’s the belief burning within you that your story feeds off of? That’s the heart of it.

#17 No work is ever wasted. If it’s not working, let go and move on – it’ll come back around to be useful later.

#22 What’s the essence of your story? Most economical telling of it? If you know that, you can build out from there.

Video: Treat your data like a gift

We recently had a series of webinars in which we discussed a few simple ways to package your data like it's a thoughtful gift. Here is recording and slides from one of those webinars. It's just under 30 minutes, so if you missed the original it might be something interesting to watch while waiting in line to return that sweater or in between nibbles of stale gingerbread house.

Have a wonderful Christmas!

Oh yeah... and here are the slides for Treat your Data Like a Gift.

Treat data like a gift; A Juice webinar


It felt like it was about time to share some of what we’ve been working on as well as review some of the Juice design principles and how they relate to sharing information. In light of the holiday season, we’ve packaged a webinar to share some of how we think you can gift wrap data for your clients. There are four sessions next week, so feel free to join us at a time most convenient for you. We'll keep these sessions to 30 minutes in length so you'll have a plenty of time to check on your other gifts.

Hope we see you there.

Tuesday, December 10 1pm ET

Tuesday, December 10th 4pm ET

Friday December 13 9am ET

Friday, December 13th at 1pm ET

Juice x Meals with a Mission


We're going to have a little pre-emptive thanks-giving around a meal by doing something called a Meal with a Mission, and you're invited. We'll be eating around this ridiculously earthy, awesome table at ROAM in Alpharetta, Ga.

When? Thursday, Nov. 21st, 12pm-1pm

What Makes This Meal Special? We're going to have fun and do some good together! Here's how it works:

There's $20 donation required to participate. Everyone will bring a nonprofit cause to share, an organization whose work you admire. After the meal, we'll take 3-4 minutes each to talk about our causes. We'll all vote, and the cause receiving the most votes wins the entire donation pool.

We'll be providing some 'ol fashioned homemade goodness from Cue Barbecue, an Atlanta favorite.

Sign-up here to donate and participate.

We can fit up to 16 folks around the table, but if more sneak in, we'll try to squeeze! If you're in Atlanta or passing through, we would love to see you.

Thanks! Can't wait to see you!

Everyday Visualizations: Cupcake Menu Comparison

A few weeks ago, while visiting in Virginia for work, I ended up on M Street in D.C. with a couple of co-workers. One of us was going to a meet-up and the other two of us had time to kill. (One thing you should know about me before we proceed is that I am a big—and I mean BIG—cake/cupcake fan. Have you heard of DC cupcakes? I don't watch it regularly, but I am very aware of it. I'm also aware of a lot of other big cupcake joints, like Sprinkles Cupcakes. So...)

Once I realized where we'd ended up with time to kill, I was immediately on a mission to find DC cupcakes. In the process we discovered a bonus: Sprinkles Cupcakes was only a few blocks away. One of my favorite things to do is compare flavors of cake from different places and see which I like better. Clearly, I need CAA (Cupcake Addicts Anonymous).

I won't bore you with all the details of the cupcake tasting (though it was delicious fun) but we did find some great everyday visualizations to share. Given that we are visualization nerds, it's only natural that while taste comparing the cupcakes, we also compared their visual menus that you can take with you on their visualization merit.

Let's start with the Georgetown Cupcakes menu. It's pink - I dig that. But beyond girly color, it has some other great features. Though it's not really a "typical" visualization in that it doesn't replace text or anything with visual elements, but it does have kind of a visual layout. To outline a few things it does well:


1. The way it's broken up by day (columns) makes it easy to search for the day you are there and find the cupcakes available.

2. It's broken into three "rows" so to speak, with a "row" for everyday, special and advance flavors. This is helpful to quickly find the break between everyday flavors and the other flavors.

Things it could do better:

1. The everyday and advance order cupcake flavors are the same for every day. It seems you could focus more on playing up the special flavors, since that is where the variation lies. Plus, you could save space and ink.

2. Within the special flavors it can be tough to find which days a particular flavor falls on easily. In the everyday rows, the same flavor is always on the same row. Not the case in the special section so your eye does a lot of jumping around trying to search for the same flavor on different days.

Moving on to the Sprinkles Cupcakes menu. It's more of a visual menu than the other.


1. It's easy to find a flavor you want and figure out which day it's available.

2. It's slightly less easy but still not hard to find the day you are there and see which cupcakes are available. There's a little more eye movement involved since there are more flavors.

Things it could do better:

1. I don't know that it really adds anything really valuable to the visualization, but the colored dots do correspond to the dots on top of the actual cupcakes. If you were having trouble reading the signs in front of the cupcakes in the case, that could be helpful for cupcake identification I suppose. A way to make the dots more helpful would be to correspond their colors more closely to the flavor they represent. The colors are pretty though.

2. No seasonal flavors are listed.

Aside from the visual aspects of the menus, another thing I like about the Georgetown Cupcakes menu is that on the backside, they have great descriptions of each cupcake flavor. As a cupcake lover, I find that helpful. Just sayin'.

And if you love cupcakes like I do and are curious which cupcakes are actually the best, you might be disappointed. They are both delicious—preference between the two comes down to...well...preference.

If you want a little cake with your icing and believe a cupcake is not a cupcake without a great icing, Georgetown Cupcakes might be your pick. Their icings were off the chart awesome! Just give me a shot of the icing, hold the cake.

If you dig cake and could take or leave the icing, you definitely want Sprinkles Cupcakes. Not only do they put less icing on each cupcake, but the cake (and how moist it is) really stands out above the icing.

In a perfect world, I'd combine the two. I need both to feel truly cupcake satisfied. What can I say? I want it all. I guess you could say it's the same with the menus: combine the find-what's-available-today easiness of the Georgetown menu with the when-are-they-serving-my-fav icon layout of Sprinkles and it'd be a match made in heaven. Until then, I guess I'll have to keep searching. And tasting. It's tough work, but someone's gotta do it.

So there you have it! A breakdown of cupcake menus! Who knew cupcake hunting could be so visual?