Kickstarter & NYTTP: Kickstarter Data-Driven University

Transcript

1. Data-Driven University Beyond Coding
2. I. Thinking Skeptically with Data (Ben) II. Data Visualization (Jeremy) III. Storytelling with Data (Kevin) What we’ll be covering today
3. THINKING SKEPTICALLY WITH DATA BEYOND CODING
4. why am I here?
5. - always be curious - data requires context - validate everything (twice)
6. PART 1: exploratory data analysis
8. set mean_x mean_y var_x var_y correlation 1 9 7.500909 11 4.127269 0.8164205 2 9 7.500909 11 4.127629 0.8162365 3 9 7.500000 11 4.122620 0.8162867 4 9 7.500909 11 4.123249 0.8165214
10. LESSON: don’t oversimplify
11. PART 2: confounding results
12. Treatment A Treatment B 78% (273/350) 83% (289/350) Comparison of treatment of renal calculi by open surgery, percutaneous nephrolithotomy, and extracorporeal shockwave lithotripsy. - C. R. Charig, D. R. Webb, S. R. Payne, J. E. Wickham (29 March 1986)
13. Treatment A Treatment B Small Stones 93% (81/87) 91% (234/270) Large Stones 73% (192/263) 69% (55/80) Both 78% (273/350) 83% (289/350)
14. as a doctor, which treatment do you recommend?
15. Treatment A Treatment B Small Stones 93% (81/87) 91% (234/270) Large Stones 73% (192/263) 69% (55/80) Both 78% (273/350) 83% (289/350)
16. tech-company example
17. Email A Email B Week of 1/4 93% (81/87) 91% (234/270) Week of 1/11 73% (192/263) 69% (55/80) Combined 78% (273/350) 83% (289/350)
18. LESSON: context is everything
19. PART 3: causality
20. http://gizmodo.com/5977989/internet-explorer-vs-murder-rate- will-be-your-favorite-chart-today
21. what if it’s not that obvious?
22. causality comes in many forms
26. Variant A Variant B
27. LESSON 3: do SCIENCE when you can be careful when you can’t
29. deep breaths
30. how do you know that your result is wrong?
31. how do you know that your result is right?
32. how do you validate your results?
33. if you think something is weird it probably is
34. come at things from multiple angles
35. - always be curious - data requires context - validate everything (twice)
36. DATA VISUALIZATION BEYOND CODING
37. Why are you here?
38. Part 1: Learning to Critique
40. http://www.wnyc. org/story/284042-vignelli- designer-of-famous-subway- map-defends-his-version-over- these-others-images/
42. Part 2: Design Principles
43. What's your purpose?
44. What's your purpose? Communicate? Explore? Something else?
45. How do you achieve your purpose?
46. How do you achieve your purpose? Graphical excellence. Obviously.
47. “Graphical excellence is that which gives to the viewer the greatest number of ideas in the shortest time with the least ink in the smallest space.” Edward R. Tufte The Visual Display of Quantitative Information (1983)
49. Part 3: Visual Components
50. Position (encode numbers)
51. Size: Length (encode numbers)
52. Size: Area jk length
53. Size: Area (encode numbers)
54. The problem with pies Image from http://en.wikipedia.org/wiki/Pie_chart
55. The problem with pies
56. Value (encode ordered relationships)
57. Value vs. color http://www.poynter.org/uncategorized/224413/why-rainbow-colors-arent-always-the-best-options-for-data-visualizations/
58. Color (encode categories)
59. Texture (encode categories)
60. Orientation (encode ¯\_(ツ)_/¯)
61. Shape (encode categories)
62. Double encoding
63. Part 4: Data Types
64. What kind of data do you have?
65. What kind of data do you have?
66. What kind of data do you have?
67. What kind of data do you have?
69. DISCRETE
70. CONTINUOUS
71. CATEGORICAL
72. TEMPORAL
73. GEOGRAPHIC
74. ORDERED
75. Part 5: Example
76. Number of successful projects over time
82. Wrapping up
83. Design principles
84. Design principles Visual components
85. Design principles Visual components Data types
86. Design principles Visual components Data types Common visualizations
87. http://extremepresentation.typepad.com/blog/2006/09/choosing_a_good.html
88. Always be learning Good starting point: http://flowingdata.com/2013/07/12/getting-started- with-visualization-after-getting-started-with-visualization/ Books and other resources: http://fellinlovewithdata.com/guides/data-vis-beginners-toolkit-1 Visualization tools: http://fellinlovewithdata.com/guides/data-vis-beginners-toolkit-2
89. DATA STORYTELLING BEYOND CODING
90. So far today, we’ve learned to: Think skeptically... ...and visually (the right way!) + GOOD BAD
91. Why are we here today? Communicate
92. Talking about data often feels like...
93. We’ll be learning some simple rules to help you: • Make your message memorable and compelling • Keep it interesting and influence others • Improve speed and quality of decision making with data
94. The principles apply to all communication... • Presentations • Emails • Homework assignments • Classroom discussions • Public speaking • ...and more
95. Storytelling with data insights “It’s not about the data… …it’s about what the data means”
96. 3 rules we’ll be covering • Convey a clear “so what” • Know your audience • Caveman speak: less is more
97. 3 rules we’ll be covering • Convey a clear “so what” • Know your audience • Caveman speak: less is more
98. How to do it: convey your objective in a way that makes its significance perfectly clear
99. I need info from you I need to give info to you We need to make a decision Inform Decide Inspire I want to influence or guide broad action/attitudes Inquire How to do it: ask yourself - “what is my objective in communicating?”
100. Example 1: What is the objective of this chart? Kevin’s first “analysis” at LinkedIn (circa 2010) :-( BEFORE X% Y% Z%
101. Example 1: What is the objective of this chart? AFTER X% Y% Z%
102. 3 rules we’ll be covering • Convey a clear “so what” • Know your audience • Caveman speak: less is more
103. How to do it: ask the right questions • What does your audience actually care about? • What’s already on their mind? What are they expecting? • What level of detail would be most effective? CEO’s frame of reference New employee’s frame of reference Your frame of reference The goal is not to show how smart you are or how much data you have!
104. Example 1: who is the audience for this email? X
105. Example 2: and this email?
106. Example 3: what about this pull request?
107. How does the audience affect the “so what?” So What First So What Last Our recommendation is X. We believe this is true because of: ▪ Finding A ▪ Finding B ▪ Finding C Our analysis has led us to the following findings: ▪ Finding A ▪ Finding B ▪ Finding C Therefore, we recommend X. When is “So What First” better than “So What Last?” Vice versa?
108. 3 rules we’ll be covering • Convey a clear “so what” • Know your audience • Caveman speak: less is more
109. How to do it: Less is more (filter out noise!)
110. Embrace your inner caveman/cavewoman Emotion ▪ Good or bad? Happy or sad? Response ▪ Keep course or change course? ▪ Take new action?
111. Make it easy for your audience to follow along! ▪ Ensure there’s a flow of the story you’re conveying ▪ Bullets are your friend – both written and spoken ▪ So is bolding for emphasis – again, written and spoken Reminder: goal is convey understanding with clarity & simplicity, not to “impress” with detail & complexity
112. Example: New Discover Research • The new Discover revamp prioritizes category discovery. We find that globally, legacy ref_tags unimpacted by the change - such as Social & Popular - still perform the best. By studying conversion rates from the newly-created Category pages, we find that the new referrer_tags implemented with the change (category_popular, category_newest, etc.) are driving higher than average pledges (conversion rate) but lower than average pledge dollars than their legacy counterparts. Our hypothesis is that the new experience is attracting more casual backers to projects by directing them immediately to categories of interest as opposed to generic sorts (like Newest). We also observe that the exit rate - or rate of visitors ending their session at the Discover page - is lower for the new experience. Once we decide on a direction, we need to focus on optimizing the current Category pages. Given backers of certain categories might display different preferences (What’ s Popular, Newest), we should be thoughtful about which sorts we show on which categories. Version 1
113. Example: New Discover Research ▪ Social, time (Ending Soon, Newest), and intent-based referrers (Search) perform best in driving pledges – Search (~X%), Social (~Y%), Ending Soon (~Z%) ▪ Category-redesign is converting more Discover users to Category pages – Clicks to category pages have increased ~M% while exit rate has dropped ~Z%... – Q3 recommendation is to optimize modules on category / discover pages for CvR – Higher prominence for “Popular”, “Ending Soon” sorts could lead to more pledges Version 2
114. Caveman speak for slides / charts: the 3-5 second rule Take the guess work out: The 3-5 second rule Can you look at a data slide for 3-5 seconds and see – at a caveman level-- what it’s telling you?
115. Example 1: What does this slide tell us?
116. Example 1: and this slide?
117. Example 2: What story does the data tell? Double Fine Launches
118. Tip: Choose your chart based on the story you want to tell ▪ 100% bar: Mix, Mix comparisons, Mix shift over time ▪ Stacked bar: Absolute comparisons, Absolute value growth ▪ Line: Absolute value trend, Trend comparisons
119. 3 rules we’ll be covering • Convey a clear “so what” • Know your audience • Caveman speak: less is more
120. Recap: Storytelling with data insights “It’s not about the data… …it’s about what the data means”
121. Homework Find a graph or presentation slide that's confusing, misleading, or just plain bad. https://www.beyondcoding.io/#homework
122. Thank you!