Chart Makeover: Clustered Bars to Lines
In “Pulling It All Together,” Chapter 8 of Storytelling with Data, author Cole Nussbaumer Knaflic takes you extensively through a makeover of a data story about the retail prices of hypothetical consumer products—unimaginatively called Product A, Product B, Product C, Product D, Product E, and the protagonist Product F.
In fact, the entire chapter is a six-lesson process on how to pull off Cole’s makeover: from ugly, non-actionable clustered bar charts to recommendation-ready, color-subdued line charts. From page 205 of her book:
Problems with the Original Multicolor, Clustered-Bar Data Story
I completely agree with Cole that her After is so much more effective than her Before. The Before suffers from:
A title that doesn’t answer the audience’s burning question
“How should we price Product F?” In fact, nowhere does the Before chart address it.
That background shading on the title. Those unnecessary borders. Those seven rainbow colors. Oh, the humanity!
Wrong chart choice
Time element illustrated as colors instead of a variable on the x-axis (one of my chart design red-flags). Most of all, a clustered bar chart. (You will hear more of my disdain for clustered bar charts in another blog post.)
NussBaumer’s Made-Over Data Story
Cole’s improvement is a nine-slide presentation (Figures 8.11 to 8.19 of Storytelling with Data) to be delivered in five minutes.
It is essentially the same line chart, shown eight times, each slide highlighting a different element of the chart to cleverly build the story.
Problems with the Data Story Makeover (After)
Though Cole’s makeover is at least 89 times better than the original (I measured; don’t ask how), there are a few things in her approach that make me feel queasy.
This level of analysis does not need multiple pages.
Let’s recognize that this is a simplistic example meant to illustrate data storytelling instead of pricing analysis. I get it. Real-life pricing analyses and recommendations are much deeper than this and involve more factors than other players’ retail prices.
But, to be completely honest, if my key takeaway is simply, “The average price in the market right now is X, so enter lower than X,” I’m not going to take nine slides to say it. I’m going to do it in one.
And if I have only five minutes to present (as Cole’s first slide says), I’m showing one page. Not nine. Crazy, right?
Especially since the final takeaway (“The average price right now is X, so enter lower than X”) makes absolutely no reference to the previous slides’ key takeaways (see these direct quotes from Storytelling with Data):
- “Products A and B were launched in 2008 at price points of $360+.”
- “They have been priced similarly over time, with B consistently slightly lower than A.”
- “In 2014, Products A and B were priced at $260 and $250, respectively.”
- “Products C, D, and E were each introduced later at much lower price points…”
- “…but all have increased in price since their respective launches.”
- “In fact, with the launch of a new product in this space, we tend to see an initial price increase, followed by a decrease over time.”
- “As of 2014, retail prices have converged, with an average retail price of $223, ranging from a low of $180 (C) to a high of $260 (A).”
Cole’s final takeaway, verbatim, is: “To be competitive, we recommend introducing our product below the $223 average price point in the $150−$200 range.”
In other words, the average price right now is $223, so enter lower at around $150−$200.
That recommendation did not need to know when and at what price points Products A and B launched. It did not need to know jack shit about how Products A and B’s prices have moved over time. That recommendation couldn’t care less about the entry of Products C, D, and E. It is oblivious to the launch−price increase−price decrease cycle of the market.
All that’s needed to make that final recommendation is knowledge of the average price right now, $223. Why take eight slides to get to it?
Those long and short lines in the same chart bother the hell out of me.
Most of all—and I’ve been holding my peace for a few months now since I’ve read Storytelling with Data—I’m extremely uncomfortable with the varying lengths of the lines in the chart.
Is it just me, or doesn’t it look disturbingly odd to have long and short lines all on the same chart? They kinda look like bugs. And they're bugging me out.
A Better Makeover of the Makeover
So instead of Cole’s nine-slide presentation leading to this:
I made this one-page data visualization and recommendation:
I think it also qualifies as a complete data story.
How I Came Up with this Alternative Data Story
If I know I’m headed towards a price recommendation that’s simply lower than the average retail price in the market, all I need to show is the movement of the average retail price over time.
And while I’m at it, I might as well show the highest and lowest prices in the market.
Cole Nussbaumer Knaflic’s data table must have looked like this (h/t Ben Jones for approximating Cole’s data values in this Tableau Story Points re-creation):
So I processed it into this:
(I added a very simplistic 2015 forecast of retail prices by reducing each product’s price by $10, since there was an overall downward price trend.)
I then simply made a stacked bar chart of minimum prices (colored transparent) and price spans (maximum price less minimum price, colored gray).
I also added smoothed lines of each year’s average retail prices (black) and maximum retail prices (orange).
The ensuing data visualization reminded me of the shape of an herbivorous dinosaur crouching to feed itself. Obviously, because of Ang Lee, whenever you have a crouching ‘one thing,’ you should have a hidden ‘another thing.’ That’s where the Crouching Dinosaur, Hidden Tiger comes from.
Justifying the dinosaur reference is the very mature consumer market with two legacy players and three times as many competitors as when it started.
The new entrant, Product F, about to pounce on the market justifies the tiger reference.
I decided to use a sensible legend at the bottom of the chart to display the rest of the analysis and data story.
Here's a downloadable PDF with better resolution.
There you have it—my makeover of Cole Nussbaumer’s makeover of a consumer pricing data story. Another case of aesthetic discomfort sparking creativity that leads to a more effective data visualization and data story.