TED2010
Sebastian Wernicke: Lies, damned lies and statistics (about TEDTalks)
Filmed:
Readability: 4.2
2,510,120 views
In a brilliantly tongue-in-cheek analysis, Sebastian Wernicke turns the tools of statistical analysis on TEDTalks, to come up with a metric for creating "the optimum TEDTalk" based on user ratings. How do you rate it? "Jaw-dropping"? "Unconvincing"? Or just plain "Funny"?
Sebastian Wernicke - Data scientist
After making a splash in the field of bioinformatics, Sebastian Wernicke moved on to the corporate sphere, where he motivates and manages multidimensional projects. Full bio
After making a splash in the field of bioinformatics, Sebastian Wernicke moved on to the corporate sphere, where he motivates and manages multidimensional projects. Full bio
Double-click the English transcript below to play the video.
00:16
If you go on the TED website,
0
1000
3000
00:19
you can currently find there
1
4000
2000
00:21
over a full week of TEDTalk videos,
2
6000
3000
00:24
over 1.3 million
3
9000
2000
00:26
words of transcripts
4
11000
2000
00:28
and millions of user ratings.
5
13000
2000
00:30
And that's a huge amount of data.
6
15000
3000
00:33
And it got me wondering:
7
18000
2000
00:35
If you took all this data
8
20000
2000
00:37
and put it through statistical analysis,
9
22000
2000
00:39
could you reverse engineer a TEDTalk?
10
24000
2000
00:41
Could you create
11
26000
2000
00:43
the ultimate TEDTalk?
12
28000
2000
00:45
(Laughter) (Applause)
13
30000
2000
00:47
And also, could you create
14
32000
2000
00:49
the worst possible TEDTalk
15
34000
2000
00:51
that they would still let you get away with?
16
36000
2000
00:53
To find this out, I looked at three things:
17
38000
2000
00:55
I looked at the topic that you should choose,
18
40000
2000
00:57
I looked at how you should deliver it
19
42000
3000
01:00
and the visuals onstage.
20
45000
2000
01:02
Now, with the topic: There's a whole range of topics you can choose,
21
47000
3000
01:05
but you should choose wisely,
22
50000
2000
01:07
because your topic strongly correlates
23
52000
2000
01:09
with how users will react to your talk.
24
54000
3000
01:12
Now, to make this more concrete,
25
57000
2000
01:14
let's look at the list of top 10 words
26
59000
3000
01:17
that statistically stick out
27
62000
2000
01:19
in the most favorite TEDTalks
28
64000
2000
01:21
and in the least favorite TEDTalks.
29
66000
3000
01:24
So if you came here
30
69000
2000
01:26
to talk about how French coffee
31
71000
2000
01:28
will spread happiness in our brains,
32
73000
3000
01:31
that's a go.
33
76000
2000
01:33
(Laughter) (Applause)
34
78000
2000
01:35
Whereas, if you wanted to talk about
35
80000
2000
01:37
your project involving
36
82000
2000
01:39
oxygen, girls, aircraft --
37
84000
2000
01:41
actually, I would like to hear that talk, (Laughter)
38
86000
2000
01:43
but statistics say it's not so good.
39
88000
2000
01:45
Oh, well.
40
90000
2000
01:47
If you generalize this,
41
92000
2000
01:49
the most favorite TEDTalks are those
42
94000
2000
01:51
that feature topics we can connect with,
43
96000
3000
01:54
both easily and deeply,
44
99000
2000
01:56
such as happiness, our own body,
45
101000
2000
01:58
food, emotions.
46
103000
2000
02:00
And the more technical topics,
47
105000
2000
02:02
such as architecture, materials and, strangely enough, men,
48
107000
3000
02:05
those are not good topics to talk about.
49
110000
3000
02:08
How should you deliver your talk?
50
113000
2000
02:10
TED is famous for keeping
51
115000
2000
02:12
a very sharp eye on the clock,
52
117000
2000
02:14
so they're going to hate me
53
119000
2000
02:16
for revealing this, because, actually,
54
121000
2000
02:18
you should talk as long as they will let you. (Laughter)
55
123000
2000
02:20
Because the most favorite TEDTalks
56
125000
3000
02:23
are, on average, over 50 percent longer
57
128000
2000
02:25
than the least favorite ones.
58
130000
2000
02:28
And this holds true for all ranking lists on TED.com
59
133000
2000
02:30
except if you want to have a talk
60
135000
2000
02:32
that's beautiful, inspiring or funny.
61
137000
2000
02:34
Then, you should be brief. (Laughter) But other than that,
62
139000
2000
02:36
talk until they drag you off the stage.
63
141000
3000
02:39
(Laughter)
64
144000
2000
02:41
Now, while ...
65
146000
2000
02:43
(Applause)
66
148000
6000
02:49
While you're pushing the clock, there's a few rules to obey.
67
154000
3000
02:52
I found these rules out by comparing the statistics
68
157000
2000
02:54
of four-word phrases
69
159000
2000
02:56
that appear more often in the most favorite TEDTalks
70
161000
3000
02:59
as opposed to the least favorite TEDTalks.
71
164000
2000
03:01
I'll give you three examples.
72
166000
2000
03:03
First of all, I must, as a speaker,
73
168000
2000
03:05
provide a service to the audience and talk about what I will give you,
74
170000
3000
03:08
instead of saying what I can't have.
75
173000
2000
03:10
Secondly, it's imperative
76
175000
2000
03:12
that you do not cite The New York Times.
77
177000
2000
03:14
(Laughter)
78
179000
2000
03:16
And finally, it's okay for the speaker -- that's the good news --
79
181000
3000
03:19
to fake intellectual capacity.
80
184000
2000
03:21
If I don't understand something, I can just say, "etc., etc."
81
186000
3000
03:24
You'll all stay with me.
82
189000
2000
03:26
It's perfectly fine.
83
191000
2000
03:28
(Applause)
84
193000
4000
03:32
Now, let's go to the visuals.
85
197000
2000
03:34
The most obvious visual thing on stage is the speaker.
86
199000
3000
03:37
And analysis shows if you want to be
87
202000
2000
03:39
among the most favorite TED speakers,
88
204000
2000
03:41
you should let your hair grow a little bit longer than average,
89
206000
3000
03:44
make sure you wear your glasses and be slightly more dressed-up
90
209000
3000
03:47
than the average TED speaker.
91
212000
2000
03:49
Slides are okay, though you might consider going for props.
92
214000
3000
03:52
And now the most important thing,
93
217000
2000
03:54
that is the mood onstage.
94
219000
2000
03:56
Color plays a very important role.
95
221000
2000
03:58
Color closely correlates
96
223000
2000
04:00
with the ratings that talks get on the website.
97
225000
3000
04:03
(Applause)
98
228000
2000
04:05
For example, fascinating talks
99
230000
2000
04:07
contain a statistically high amount
100
232000
2000
04:09
of exactly this blue color, (Laughter)
101
234000
2000
04:11
much more than the average TEDTalk.
102
236000
2000
04:13
Ingenious TEDTalks, much more this green color,
103
238000
2000
04:15
etc., et.
104
240000
2000
04:17
(Laughter) (Applause)
105
242000
7000
04:24
Now, personally, I think
106
249000
2000
04:26
I'm not the first one who has done this analysis,
107
251000
2000
04:28
but I'll leave this
108
253000
2000
04:30
to your good judgment.
109
255000
3000
04:33
So, now it's time to put it all together
110
258000
2000
04:35
and design the ultimate TEDTalk.
111
260000
2000
04:37
Now, since this is TEDActive,
112
262000
2000
04:39
and I learned from my analysis
113
264000
2000
04:41
that I should actually give you something,
114
266000
2000
04:43
I will not impose the ultimate
115
268000
2000
04:45
or worst TEDTalk on you,
116
270000
2000
04:47
but rather give you a tool to create your own.
117
272000
2000
04:49
And I call this tool the TEDPad.
118
274000
3000
04:52
(Laughter)
119
277000
3000
04:55
And the TEDPad is a matrix
120
280000
2000
04:57
of 100 specifically selected,
121
282000
2000
04:59
highly curated sentences
122
284000
3000
05:02
that you can easily piece together to get your own TEDTalk.
123
287000
3000
05:07
You only have to make one decision,
124
292000
2000
05:09
and that is: Are you going to use the white version
125
294000
2000
05:11
for very good TEDTalks,
126
296000
2000
05:13
about creativity, human genius?
127
298000
2000
05:15
Or are you going to go with a black version,
128
300000
2000
05:17
which will allow you to create really bad TEDTalks,
129
302000
2000
05:19
mostly about blogs,
130
304000
2000
05:21
politics and stuff?
131
306000
2000
05:23
So, download it and have fun with it.
132
308000
2000
05:25
Now I hope you enjoy the session.
133
310000
3000
05:28
I hope you enjoy designing your own
134
313000
2000
05:30
ultimate and worst possible TEDTalks.
135
315000
2000
05:32
And I hope some of you will be inspired for next year
136
317000
3000
05:35
to create this, which I really want to see.
137
320000
3000
05:38
Thank you very much.
138
323000
2000
05:40
(Applause) Thanks.
139
325000
10000
ABOUT THE SPEAKER
Sebastian Wernicke - Data scientistAfter making a splash in the field of bioinformatics, Sebastian Wernicke moved on to the corporate sphere, where he motivates and manages multidimensional projects.
Why you should listen
Dr. Sebastian Wernicke is the Chief Data Scientist of ONE LOGIC, a data science boutique that supports organizations across industries to make sense of their vast data collections to improve operations and gain strategic advantages. Wernicke originally studied bioinformatics and previously led the strategy and growth of Seven Bridges Genomics, a Cambridge-based startup that builds platforms for genetic analysis.
Before his career in statistics began, Wernicke worked stints as both a paramedic and successful short animated filmmaker. He's also the author of the TEDPad app, an irreverent tool for creating an infinite number of "amazing and really bad" and mostly completely meaningless talks. He's the author of the statistically authoritative and yet completely ridiculous "How to Give the Perfect TEDTalk."
Sebastian Wernicke | Speaker | TED.com