Data that persuades Howto prove your point Ian Lurie CEO, Portent This is the ‘narrated’ version of my presentation from SES San Francisco, 2012. You’ll see explanations in little black boxes like this one. San Francisco| August 13–17
If you havequestions after you read this, shoot ‘em to me on Twitter. I’m happy to answer. ME: @PORTENTINT
Also, I’ve putlinks to books, examples, etc. on a bit.ly link bundle at this link. portent.co/sesdata Yes, like this, no ‘m’.
I’m not afan of bullets. They give slides a bad name. So I’ve banned them from this presentation. This presentation is certified 100% bullet-free.
I came upwith the idea for this presentation after receiving a graph that looked like this: 7000 6000 5000 4000 Jun-‐12 3000 Jul-‐12 2000 Aug-‐12 1000 0 Alfalfa Timothy Chewy Pigloos Hay Treats @portentint
Why the @#$!would you do this? 7000 6000 5000 4000 Jun-‐12 3000 Jul-‐12 2000 Aug-‐12 1000 0 I couldn’t figure out why people would do this to innocent, Timothy Chewy Alfalfa Pigloos helpless data. This graph is Hay Treats awful. @portentint
Why the @#$!would you do this? 7000 6000 5000 4000 Jun-‐12 3000 Jul-‐12 2000 Aug-‐12 1000 0 Even worse, it was created by a Alfalfa Timothy Chewy Pigloos marketing ‘professional’ to give to Hay Treats her boss (I changed the products to protect the innocent) @portentint
Why the @#$!would you do this? 7000 6000 5000 4000 Jun-‐12 3000 Jul-‐12 2000 Aug-‐12 1000 0 Alfalfa Timothy Chewy Pigloos Would you take this seriously? Can Hay Treats even figure out what the story you is behind this data? @portentint
Why the @#$!would you do this? 7000 6000 5000 4000 Jun-‐12 3000 Jul-‐12 2000 Aug-‐12 1000 0 Alfalfa Timothy Chewy Pigloos Hay Treats Then I found out her boss asked for it this way to show to his boss. Grrrrrrr… @portentint
Of course, thisisn’t really our problem, right? I mean, we’re SEOs and PPC’ers and such. Our job isn’t data presentation – it’s giving good advice about rankings and ads and stuff. WHY SHOULD, TOO AND YOU I CARE
WRONG. SO WRONG.We face a few huge problems we can only fix if we know how to present data in a compelling way. WHY SHOULD, TOO AND YOU I CARE
The problem rightnow is that we approach data the same way we approach food. DATA = FOOD
…more’s better! Sowe pig out. We create these awful spreadsheets and charts, export dozens of pages from analytics software and let our clients and bosses stuff themselves. But they don’t get any useful info from it. More data better!!!
Whoa, hey now, let's not get personal. So our clients with end up fatter than Grizzlies, but with no idea what to do next. The data is totally overwhelming and actionably worthless.
We need toserve our clients little, well-prepared servings of super-dense, useful information. Not 32 Big Macs. I’ll lay off the food metaphor now.
YOUR CAREER ISN’TABOUT BEING RIGHT Because here’s the thing: Being right doesn’t get you paid. It doesn’t get you promoted. It doesn’t get you business. Being right is how you make sure you don’t screw your clients/bosses with bad information.
YOUR CAREER ISABOUT BEING UNDERSTOOD If you want the fact that you’re right to matter – if you want to really succeed in your career as a marketer – you need to be understood. In our business, that comes down to data, and good presentation of that data.
IF YOU WANTPERSUASIVE You have to be persuasive. But that doesn’t mean the prettiest graph, or the loudest argument, or the best skills as a debater.
BE THE PERSONTHEY ACTUALLY UNDERSTAND You want to be persuasive? Be the one person that your client can actually understand. That’s how you grow your career, and become awesome.
HERE’S HOW YOUDO IT 3 STRATEGIC TIPS First off, three basic rules you have to know.
LIZARD. APE. HUMAN. Beforeyou prepare your report, repeat the phrase “Lizard, Ape, Human”. It’ll help you remember how people consume information, and keep your report focused.
According to DonaldNorman, we respond to information on three levels: Visceral (fight or flight), Behavioral (like or hate), and Reflective (meaning). VISCERAL BEHAVIORAL REFLECTIVE From The Design of Everyday Things, Donald Norman
it's true. I'ma visceral guy. When you drop a report on my desk (I am a boss, after all), my first reaction is visceral. It comes from my lizard brain, and it’s a very simple yes/no response. A ‘yes’ means I’ll look long enough to form an opinion. A ‘no’ means I’m probably going to do everything I can to ignore the data you so lovingly prepared for me. Choice of color, layout, report clutter and such can affect this response.
not as smart as humans? Whatever. Then I’ll look a little more closely. This is closer to the ape-driven part of my brain. It’s an emotional response, but not totally driven by a look at the data. A well-designed bit of data graphics can make a huge difference here, because the ape-brain isn’t quite as smart as the human-brain, but it’s very nuanced. At this point, I’m looking for something to ponder.
does this mean I'm an apex predator? Sweet! Now, finally, I hit the reflective response. The visceral and emotional response shape the reflective one. If my first reaction was “ew” and my second was “I can’t believe I paid for this,” my reflective response is going to be “Who can I hire to replace you?” If my visceral reaction was “huh!” and my behavioral was “I had no idea!” my reflective response will be “I need to dig deeper! Get in here!”
WHAT. WHY. WHATNOW? Getting the right visceral response is about information design. I’ll talk about that in minute. Getting the right behavioral and reflective response is about answering the right questions: What story does the data tell? Why did that story happen? What should we as a company do next, based on the what and why?
WHAT: SEO &PPC are on an upswing. PPC SEO E-‐mail Begging May-‐12 $669.00 $643.00 $1,005.00 $767.00 Jun-‐12 $850.00 $1,500.00 $647.00 $1,074.00 Jul-‐12 $2,500.00 $2,340.00 $619.00 $687.00 Understand that data isn’t a bunch of numbers. It’s a story – a narrative – of how your marketing, product and company are working. It’s up to you to dig out that story.
New landers launchedJuly 1. SEO content strategy ongoing as of May 1. PPC SEO E-‐mail Begging May-‐12 $669.00 $643.00 $1,005.00 $767.00 Jun-‐12 $850.00 $1,500.00 $647.00 $1,074.00 Jul-‐12 $2,500.00 $2,340.00 $619.00 $687.00 It’s not enough to just say “SEO and PPC are doing well!” You need to explain why. Otherwise, you’re no better than a Google Analytics report, and I don’t need to pay you.
Why? PPC Conversion rate 6% Landers go live June 1 5.00% 5% What explains the change? Look 4.00% 4% at the data. Think this through, thoroughly, before you even try 3% to create a report or presentation. Do the research. 2% 1.10% 1.00% 1.00% 1% 0.75% 0.40% 0% Jan-‐12 Feb-‐12 Mar-‐12 Apr-‐12 May-‐12 Jun-‐12 Jul-‐12
Next Test more landers. ContinueSEO. There’s a lag. Then, be ready with next steps. These examples are over- simplified, obviously. But you get the idea. Because any boss/client will ask what to do next.
WHAT. WHY. NEXT. Don’tforget: Answer these three questions. Yes, this is a repeat slide. It’s that important.
SEO & PPC performance $1,400.00 $1,200.00 $1,000.00 $800.00 While this chart tells me SEO $600.00 and PPC are improving, it doesn’t $400.00 tell me how they’re doing compared to my other channels. $200.00 I read it and say ‘so what’? $-‐ May-‐12 Jun-‐12 Jul-‐12
Sales by channel $1,200.00 Ah HAH. Now I see the whole equation. SEO’s out-performing all other channels. This is data in context. It helps me make a smart decision. May-‐12 $600.00 Jun-‐12 Jul-‐12 $-‐ PPC SEO E-‐mail Begging
Lizard. Ape. Human. What.Why. What next. Provide context. Three big strategic tips that’ll add depth to your reports. But you still have to get past that first, visceral assessment. And, you have to provide clarity when the reader looks more closely.
13 WAYS TOGIVE GREAT DATA Time to get tactical. Here are a bunch of little tips that’ll help make your reports more consumable.
1 DON’T MAKETHEM READ Pretend your brain is a computer. Every individual ‘chunk’ of information you have to process is a task. More tasks = more work. More work = grumpy boss. @portentint
PPC SEO E-‐mail Begging May-‐12 $669.00 $643.00 $1,005.00 $767.00 Jun-‐12 $850.00 $1,500.00 $647.00 $1,074.00 Jul-‐12 $2,500.00 $2,340.00 $619.00 $687.00 Tabular data like this forces me to perform at least 12, and as many as 19, individual tasks, just to get the story that PPC and SEO are moving up.
A simple linechart requires at most 2 tasks. Phew. Much better. The moral? Don’t make me read. Reading was the last thing we learned to do. We’ve been looking for patterns, though, for hundreds of thousands of years. Our brains are better at it, and we can take in a pattern shown in an image in a few seconds. Visceral, ftw! SEO & PPC performance $1,400.00 $1,200.00 $1,000.00 $800.00 $600.00 $400.00 $200.00 $-‐ May-‐12 Jun-‐12 Jul-‐12
FOLLOW TUFTE’S LAWS Aboveall else, show the data. Maximize data-ink ratio. Erase non-data ink. Erase redundant data ink. Revise and edit. From The Visual Display of Quantitative Data, by Edward Tufte Next, a bunch of fantastic tips to follow, from the top data visualization guru on the planet, Edward Tufte. Read his books! @portentint
2 ABOVE ALLELSE, SHOW THE DATA His best, first rule: Show the damned data!!!! @portentint
Remember, my brainhas to perform tasks. Every additional bit of junk on the page interferes with my ability to ‘see’ the data. $1,200.00 $1,000.00 $800.00 May-‐12 $600.00 Jun-‐12 Jul-‐12 $400.00 $200.00 $-‐ PPC SEO E-‐mail Begging
By simply deletingthe weird vegetable icons, you’ll get one step closer to just showing the data. Nice. $1,200.00 $1,000.00 $800.00 May-‐12 $600.00 Jun-‐12 Jul-‐12 $400.00 $200.00 $-‐ PPC SEO E-‐mail Begging
One way toshow the data first is to maximize the ratio of data to ink. 3 MAXIMIZE DATA-INK RATIO @portentint
All the inkused to show the 3D effects are a total waste. They in no way help me understand this data. In fact, they may get in the way, since the blue ‘May’ bars are partly hidden. $1,200.00 $1,000.00 $800.00 May-‐12 $600.00 Jun-‐12 Jul-‐12 $400.00 $200.00 $-‐ PPC SEO E-‐mail Begging
See how muchbetter that is? I’m using fewer pixels for decoration, and more for data. $1,200.00 $1,000.00 $800.00 May-‐12 $600.00 Jun-‐12 Jul-‐12 $400.00 $200.00 $-‐ PPC SEO E-‐mail Begging
Take another stepby removing more non-data ink. 4 REMOVE NON-DATA INK @portentint
For example, thegridlines in this graph aren’t really helping. I can tell you that, as your boss, I’m not going to trace across them. Remember the story – you’re trying to show a trend, and relative performance. The grid isn’t $1,200.00 helping with that. $1,000.00 $800.00 May-‐12 $600.00 Jun-‐12 Jul-‐12 $400.00 $200.00 $-‐ PPC SEO E-‐mail Begging
Bye-bye, grid. Niceknowing you. $1,200.00 $1,000.00 $800.00 May-‐12 $600.00 Jun-‐12 Jul-‐12 $400.00 $200.00 $-‐ PPC SEO E-‐mail Begging
5 REMOVE REDUNDANTDATA INK Remove data ink (pixels, if you’re an onscreen person) that repeats the message, too. In this example: I don’t need to show the Y-axis values in $200 increments. Again, this is about trending. The small increments introduce a lot of redundant ink. @portentint
This tells thedata’s story just as well, but my visceral reaction is a lot better now. My fevered little brain doesn’t have to perform as many tasks. $1,200.00 May-‐12 $600.00 Jun-‐12 Jul-‐12 $-‐ PPC SEO E-‐mail Begging
As a reminder,there is where we started. $1,200.00 $1,000.00 $800.00 May-‐12 $600.00 Jun-‐12 Jul-‐12 $400.00 $200.00 $-‐ PPC SEO E-‐mail Begging
This is theend result. Much better, yes? $1,200.00 May-‐12 $600.00 Jun-‐12 Jul-‐12 $-‐ PPC SEO E-‐mail Begging
Most important: Don’ttreat your reports and data graphics as fragile models. You can tear them apart, revise them and generally tweak them to make ‘em more readable. Revise and edit! 6 REVISE & EDIT @portentint
For example, Idecided the grid was helpful after all, because it made it easier for folks to compare bars on the far-right of the chart. I didn’t want to add ink, so I added a white line across at the $600 increment. Revise and edit. $1,200.00 May-‐12 $600.00 Jun-‐12 Jul-‐12 $-‐ PPC SEO E-‐mail Begging
FOLLOW TUFTE’S LAWS Aboveall else, show the data. Maximize data-ink ratio. Erase non-data ink. Erase redundant data ink. Revise and edit. From The Visual Display of Quantitative Data, by Edward Tufte There’s no way I can do Tufte justice in a few slides. Read his stuff. Follow his 5 laws. Join the data nerds. It’s nice here. They have cookies.* @portentint
*Not crappy cookies,like chocolate chip cookies that turn out to be oatmeal raisin (shudder) or those nasty, chocolate-free sugar cookies. We’re talking cookies with honkin’ huge chocolate chunks in them so large they make your cheeks hurt. Data nerd-dom is a nice place. @portentint
7 USE SMALLMULTIPLES Small multiples help you break up complex datasets and compare them. Tufte and a bunch of other writers talk about them. I’m not sure who came up with the idea. It wasn’t me, though. @portentint
Sales by month $1,200.00 $1,000.00 $800.00 PPC $600.00 SEO E-‐mail $400.00 Take this example. It’s a little Begging confusing. Which line is which? I have to do a lot of work to interpret it. $200.00 $-‐ May-‐12 Jun-‐12 Jul-‐12
Sales by channel PPC SEO $1,400.00 $1,400.00 $1,200.00 $1,200.00 $1,000.00 $1,000.00 $800.00 $800.00 $600.00 $600.00 $400.00 $400.00 $200.00 $200.00 $-‐ $-‐ May-‐12 Jun-‐12 Jul-‐12 May-‐12 Jun-‐12 Jul-‐12 E-‐mail Begging $1,400.00 $1,400.00 $1,200.00 $1,200.00 $1,000.00 $1,000.00 $800.00 $800.00 $600.00 $600.00 $400.00 $400.00 $200.00 $200.00 $-‐ $-‐ May-‐12 Jun-‐12 Jul-‐12 May-‐12 Jun-‐12 Jul-‐12 Small multiples means breaking those 4 series into a grid of matching graphs. Much easier to compare.
Sales by channel PPC SEO $1,050.00 $1,250.00 $1,000.00 $1,150.00 $950.00 $1,050.00 $900.00 $950.00 $850.00 $800.00 $850.00 $750.00 $750.00 May-‐12 Jun-‐12 Jul-‐12 May-‐12 Jun-‐12 Jul-‐12 E-‐mail Begging $1,140.00 $1,160.00 $1,120.00 $1,150.00 $1,100.00 $1,140.00 $1,080.00 $1,130.00 $1,060.00 $1,040.00 $1,120.00 May-‐12 Jun-‐12 Jul-‐12 May-‐12 Jun-‐12 Jul-‐12 You can compress the y-axes if you want to, to emphasize change…
Sales by channel PPC SEO $1,350.00 $1,350.00 $1,250.00 $1,250.00 $1,150.00 $1,150.00 $1,050.00 $1,050.00 $950.00 $950.00 $850.00 $850.00 $750.00 $750.00 May-‐12 Jun-‐12 Jul-‐12 May-‐12 Jun-‐12 Jul-‐12 E-‐mail Begging $1,350.00 $1,350.00 $1,250.00 $1,250.00 $1,150.00 $1,150.00 $1,050.00 $1,050.00 $950.00 $950.00 $850.00 $850.00 $750.00 $750.00 May-‐12 Jun-‐12 Jul-‐12 May-‐12 Jun-‐12 Jul-‐12
Sales by channel PPC SEO $1,350.00 $1,350.00 $1,150.00 $1,150.00 $950.00 $950.00 $750.00 $750.00 May-‐12 Jun-‐12 Jul-‐12 May-‐12 Jun-‐12 Jul-‐12 E-‐mail Begging $1,350.00 $1,350.00 $1,150.00 $1,150.00 $950.00 $950.00 $750.00 But I recommend using the same range on every $750.00 May-‐12 Jun-‐12 graph. Here, I have each chart Jun-‐12 Jul-‐12 May-‐12 starting Jul-‐12 $750, at instead of 0, because it makes the changes more obvious. But it’s still apples-to-apples.
BSG despair-o-meter Season 1 Season 2 9 9 9 9 9 9 8 Episode quality 1-‐5 6-‐10 11-‐15 16-‐20 1-‐5 6-‐10 11-‐15 16-‐20 Season 3 Season 4 10 9 9 7 7 6 4 2 1-‐5 6-‐10 11-‐15 16-‐20 1-‐5 6-‐10 11-‐15 16-‐20 A personal favorite: My Battlestar Galactica Despair-O-Meter compares episode quality season by season. Small multiples fight the Cylon menace.
8 USE NATURALCOLORS Anything you can do to tap our instinctive ability to respond to stimuli makes data more digestible. That includes using natural colors, instead of… @portentint
…eye-gougingly bad colors. Sales by channel $1,200.00 May-‐12 $600.00 Jun-‐12 Jul-‐12 $-‐ PPC SEO E-‐mail Begging
Ah. Much better,yes? $1,200.00 Sales by channel May-‐12 $600.00 Jun-‐12 Jul-‐12 $-‐ PPC SEO E-‐mail Begging
9 USE CONTRASTING NATURAL COLORS Yes, but still a little hard to read, because all the blues ran together. So, try to use contrasting natural colors. @portentint
From Tim Beard http://portent.co/NHy5QD Tim Beard has this great palette you can use. It’s a whole set of natural colors.
This isn’t perfect,but it’s a lot easier to look at, and it’s easy to distinguish one series from another. Sales by channel $1,200.00 May-‐12 $600.00 Jun-‐12 Jul-‐12 $-‐ PPC SEO E-‐mail Begging
Pie charts suck.Seriously. I’ve never seen a pie chart that couldn’t be improved by using another type of chart. 10 SAY NO TO PIE CHARTS @portentint
How exactly amI supposed to see the difference between, Orders by channel say, 4% and 8%? Do they really look like they’re separated by a factor of 2? No. Pie charts tend to conceal distinctions.
Orders by channel 70 Turn the same thing into a bar chart, and suddenly it’s a lot 60 easier to see the entire range of results. 50 40 30 20 10 0 SEO PPC E-‐mail Display Direct
11 USE CONSISTENTPROPORTIONS If you’re using small multiples, or even presenting several pages of charts, use the same proportions on all of them. @portentint
Sales by channel PPC SEO $1,350.00 $1,350.00 $1,150.00 $1,150.00 $950.00 $950.00 $750.00 $750.00 May-‐12 Jun-‐12 Jul-‐12 May-‐12 Jun-‐12 Jul-‐12 E-‐mail Begging $1,350.00 $1,350.00 $1,150.00 $1,150.00 $950.00 Here, using different proportions $950.00 makes it look like the change over $750.00 time for PPC and SEO$750.00 is about May-‐12 the same as Jul-‐12 Jun-‐12 e-mail and begging. I May-‐12 Jun-‐12 Jul-‐12 won’t look at the numbers, first. I’ll look at the lines. Be sure they tell an accurate story.
Sales by channel PPC SEO $1,350.00 $1,350.00 $1,150.00 $1,150.00 $950.00 $950.00 $750.00 $750.00 May-‐12 Jun-‐12 Jul-‐12 May-‐12 Jun-‐12 Jul-‐12 E-‐mail Begging $1,350.00 $1,350.00 $1,150.00 $1,150.00 $950.00 $950.00 $750.00 $750.00 May-‐12 Jun-‐12 Jul-‐12 May-‐12 Jun-‐12 Jul-‐12 See the difference?
12 GO 45DEGREES Stephen Few talks about trending data in his book, Now You See It. A lot of people ask him what the best size/shape is for a line graph. @portentint
Sales $450,000 $400,000 $350,000 $300,000 $250,000 $200,000 $150,000 $100,000 His answer: Don’t worry about $50,000 the size. Instead, set the $-‐ proportions so the average slope Jan-‐11 Feb-‐11 Mar-‐11 45 degrees. That tends to be Aug-‐11 Sep-‐11 is Apr-‐11 May-‐11 Jun-‐11 Jul-‐11 easiest to read.
Sales $450,000 $400,000 $350,000 $300,000 $250,000 $200,000 $150,000 It’s not easy to see this is an $100,000 upward trend. $50,000 $-‐ Jan-‐11 Feb-‐11 Mar-‐11 Apr-‐11 May-‐11 Jun-‐11 Jul-‐11 Aug-‐11 Sep-‐11
Sales $450,000 $400,000 $350,000 $300,000 $250,000 $200,000 $150,000 Add a trendline. Now I can see $100,000 the trend in a second. $50,000 $-‐ Jan-‐11 Feb-‐11 Mar-‐11 Apr-‐11 May-‐11 Jun-‐11 Jul-‐11 Aug-‐11 Sep-‐11
DON’T BE THESEGUYS There are the tactics. Now, a couple of examples of what can happen when you ignore the tactics and the strategy of great data visualization, and just crank out one crappy report after another.
No clear story.No what, when, what’s next. Huge amounts of non-data ink and just visual trash.
This is actuallyso bad that it looks equally informative backwards…
And even upside-down.If your report looks just as useful(less) upside-down, start again.
Total Clicks Comparedto Total Monthly Spending $7,000.00 6,000 $6,000.00 5,000 $5,000.00 4,000 Spending $4,000.00 Clicks 3,000 $3,000.00 2,000 $2,000.00 $1,000.00 1,000 If I just pull out two graphs, I $- 0 Dec-10 think there’s a story. The story of Jan-11 Feb-11 Mar-11 Apr-11 May-11 Jun-11 Jul-11 Aug-11 Sep-11 Oct-11 Nov-11 a very bored analyst, playing Historical Average with shading filters in Excel. Clicks Conversions Compared to Total Monthly Spending $7,000.00 18 $6,000.00 16 14 $5,000.00 12 Conversions Spending $4,000.00 10 $3,000.00 8 6 $2,000.00 4 $1,000.00 2 $- 0 Dec-10 Jan-11 Feb-11 Mar-11 Apr-11 May-11 Jun-11 Jul-11 Aug-11 Sep-11 Oct-11 Nov-11 Spending Conversions
20 Conversions 15 10 5 0 Take 2 minutes to remove non- Dec-10 Feb-11 Apr-11 Jun-11 Aug-11 Oct-11 data ink, apply small multiples and such, and suddenly there’s a bit more information. $8,000 Spend $6,000 $4,000 $2,000 $- Dec-10 Feb-11 Apr-11 Jun-11 Aug-11 Oct-11
20 Conversions 15 10 5 0 Dec-10 Feb-11 Apr-11 Jun-11 Aug-11 Oct-11 The data isn’t bad. The presentation was awful. $8,000 Spend $6,000 $4,000 $2,000 $- Dec-10 Feb-11 Apr-11 Jun-11 Aug-11 Oct-11
DON’T BE THESEGUYS Sometimes, though, the data is bad, too. Take this last example.
Monthly PPC visitors 1,200 1,000 800 Unique Visitors 600 400 200 This report is supposed to show my PPC campaign performance. But it 0 only shows me clicks!!! I don’t care Jan Feb about clicks!!! IMay March April wantJune know what to July Aug Sep Oct Nov Dec 2008 319 405 667 926 1,121 913 817 520 433 331 267 204 2009 239 those clicks cost and earn. This 722 a 780 992 1,073 889 862 is 567 199 133 86 2010 253 397 100% useless crap graphic. 618 741 573 456 463 402 370 333 192 153 2011 217 322 511 640 931 793 603 526 390 357 250 262 2012 339 405 581
AN OBLIGATION I hopeyou find this helpful. My point, again: Your job is as much explainer as expert. Learn to find and tell the story in the data, and learn to do it well. It’ll put you head-and-shoulders above your competition.
AN OBLIGATION And, tome, there’s an ethical obligation for us as experts. It’s our job to provide good, actionable advice. Not ‘good’ and ‘actionable’ from our own perspective, but from our clients’.
AN OBLIGATION The businessesof the people you work for depend on that. Take it seriously. Plus: Cookies. We have them. Data nerds, unite.