Ad
More Related Content
Viewers also liked (13)
Similar to Storytelling with data think broad, mine deep, explain simply (20)
Ad
More from Luciano Pesci, PhD (20)
Ad
Recently uploaded (20)
Storytelling with data think broad, mine deep, explain simply
- 1. Storytelling with Data: Think Broad. Mine Deep. Explain Simply. SLC|SEM Digital Marketing Conference August 25, 2016 www.emperitas.com / 801.810.5869 / 4609 South 2300 East Suite 204, Holladay, UT 84117
- 2. Data’s Role in Digital Marketing
- 3. Why You Need to Be Using Data • The digital revolution means it’s never been easier to generate or collect data. • Data is the competitive decider right now. • Any data you use should complement your gut intuition, not replace it. 3
- 4. The Problem Is… • 80% of data projects are failing right now. • It’s because the analytics lack context & translation. • The results are confusing, ugly, and too technical. • The analysis isn’t tied to a clear problem or strategic decision. 4
- 5. Purpose of This Presentation – Effective Storytelling with Data • Thinking Broadly – Capture all relevant information and data. • Mining Deeply – Use the most powerful analytics available. • Explaining Simply – Translate the results into plain English. 5
- 7. The Story I’m Telling Today • I picked a random topic to show as an example of storytelling with data. • Our protagonists are two white guys who wanted to build a global brand teaching history through rap. • After failing at live performances, they turned to YouTube and within five years created the most successful internet show ever… 7
- 8. Epic Rap Battles of History* 8 *Source: https://youtu.be/njos57IJf-0
- 9. Epic Rap Battles of History 9
- 10. Epic Rap Battles’ Impact Most successful internet show ever – now on it’s 5th season. YouTube is running traditional advertising (TV, Billboard, etc.) to promote the ERB channel. 10 64 Episodes 13.5mn Subscribers 3bn Views
- 11. Their Brand Promise – Fans Are the Ones Driving It • Each video ends with the same call to action: • “Who Won? Who’s Next?” You Decide.” • Highly visual production, meticulously researched, intellectually engaging, and irreverent. • Give their fans behind the scenes access to see how the show is made. 11
- 12. ERB’s Pop Culture Impact • Regularly feature other YouTubers & celebrities. • Spawned huge numbers of copy cat channels and fan sites, and has been featured in multiple “react to” videos (i.e. “Elders React” & “Teens React”). • Being used in India to teach English.* 12 *Source: https://dspace.mah.se/handle/2043/16234
- 13. Mining Deeply
- 14. Any Analytics Project Needs a Clear Target • We can use data to see what’s driving engagement, and if they’re living up to their brand promise. • Ideally we’d use ERB’s proprietary channel data, but they haven’t given us that…yet. • This meant we had to look for public data sources. 14
- 15. YouTube API (Version 3) • YouTube’s open API provided us with: • Video Title • Date Posted • Video Length (in seconds) • # of Views • # of Comments • # of Likes and # of Dislikes • Samples of Comments 15 *Data was pulled the afternoon of August 23rd, 2016
- 16. Creating New Variables for Our Analysis • From the YouTube data, we created new variables: • Net Likes (Likes - Dislikes) • Likes Ratio (Likes/Dislikes) • Days Since Posting (August 23rd 2016 - Day Posted) • Comments Per Days Since Posting (Comments/DSP) • Views Per Days Since Posting (Views/DSP) • Net Likes Per Days Since Posting (Net Likes/DSP) • Comments Per Video Length (Comments/Video Length) • Views Per Days Since Posting (Views/DSP) • Net Likes Per Days Since Posting (Net Likes/DSP) 16
- 17. Fan Voting Data from Fandom (Wikia) • The producers of the show read their video comments for future battle suggestions from fans. • The comments also contain votes, but there’s no tally. • Fandom runs a poll for each of the battles, so we merged this data with the YouTube API data. 17
- 18. NLP & Manual Quantification • Natural language processing allowed patterns to be discovered, such as the role of profanity in all episodes and the comments. • The gender of the challengers, and whether they were real or fictional, were other variables that we manually added to the data set. 18
- 19. Mining Deep Using Open Source Tools • This combined data set is available on our website. • We used R & RStudio (both open-source) to run the analysis, and Tableau to make the visualizations. • We focused on answering the questions of what drives fan engagement & if ERB is living up to its brand promise. 19
- 21. What’s Driving Fan Engagement? • How do the battles stack up against each other? • Views, Comments, Net Likes, Likes Ratio. • What about deflating these metrics by Time Since Posting? • What role does profanity play in the battles? 21
- 22. Comments & Views Across Episodes • Comments • Average (204,800) and the Median (170,900). • One major outlier (606,951) – Barack Obama vs Mitt Romney. • Views • Average (49,470,000) and the Median (41,360,000). • Same outlier (123,600,000) – Barack Obama vs Mitt Romney 22
- 23. Total Views by Episode (Chronologically Ordered) 23
- 24. Total Views by Episode (Chronologically Ordered & Combining Vader v Hitler Battles) 24
- 25. Net Likes & Likes Ratio Across Episodes • Net Likes • Average (352,300) and the Median (340,200). • Two outliers this time (871,720) – Barack Obama vs Mitt Romney and (763,799) – Steve Jobs vs Bill Gates. • Likes Ratio • Average (40x) and the Median (39x). • One outlier (104x) – J. R. R. Tolkien vs George R. R. Martin 25
- 26. Likes & Dislikes by Episode (Chronologically Ordered)* 26 *Source: y-axis increments are scaled differently
- 27. The Power & Prevalence of Profanity • They are speaking the same language as their fans, and it’s profane. • Average of 4 “traditional” profanities across battles. Hitler vs Vader #2 has the most profanity at 11. • Top profaner was Marilyn Monroe (at 8), though Darth Vader had the most profanities per second (7 in 33 seconds). • Eve had 2x the profanities (7) of Adam (3) 27
- 28. Are They Living up to Their Brand Promise? • Why are challengers winning their battles? • How does gender and fictional status affect it? • Are they getting better at picking fan battle ideas that increase engagement? • Views, Comments, Net Likes, Likes Ratio. 28
- 29. Predicting Battle Winners 29 • The longer a challenger raps, the higher the probability of winning the battle. • Gender didn’t seem to make a difference, but a real challenger is significantly more likely to win against a fictional opponent. • Each profanity increases the chance of winning by 11%.
- 30. Are They Getting Better over Time? • Since each video has been available for different amounts of time, we need to deflate everything by the number of days since posting. • This gives us a clearer picture of the relative performance of each battle over time and allows us to answer if they’re improving. 30
- 31. Total Views per Day Since Posting by Episode (Chronologically Ordered) 31
- 32. Likes Ratio per Day Since Posting by Episode (Chronologically Ordered) 32
- 33. Likes & Dislikes per Day Since Posting by Episode (Chronologically Ordered)* 33 *Source: y-axis increments are scaled differently
- 34. The End
- 35. Knowing Is Half the Battle • Now you know: • The story of a unique brand living up to its promise and engaging its fans. • The story of how the data collection and analysis was done to be able to tell this story. • This is a process you can replicate by following the same three steps: think broad, mine deep, explain simply. 35
- 36. The Conversation Doesn’t Have to End Here… [email protected] / 801-810-5869 / EmperitasSG / 4609 South 2300 East Suite 204, Holladay, UT 84117