1

Trigger-Action Programming

in the Wild:

An Analysis of 200,000 IFTTT Recipes

Blase Ur, Melwyn Pak Yong Ho, Stephen Brawner, Jiyun Lee,

Sarah Mennicken, Noah Picard, Diane Schulze, Michael Littman

2

3

Trigger-Action Programming

4

Trigger-Action Programming

5

Trigger-Action Programming

6

Trigger-Action Programming

7

Trigger-Action Programming

8

Trigger-Action Programming

9

Trigger-Action Programming

10

Trigger-Action Programming

11

Trigger-Action Programming

12

What are people

creating on ?

13

Methodology

• Collect all publicly shared IFTTT recipes

14

Methodology

• Collect all publicly shared IFTTT recipes

15

Methodology

• Collect all publicly shared IFTTT recipes

– Using Selenium & ChromeDriver

16

Methodology

• Collect all publicly shared IFTTT recipes

– Using Selenium & ChromeDriver

– As of September 6, 2015

17

Methodology

• Collect all publicly shared IFTTT recipes

– Using Selenium & ChromeDriver

– As of September 6, 2015

– Compare to 2013 dataset (CHI 2014)

18

Methodology

• Collect all publicly shared IFTTT recipes

– Using Selenium & ChromeDriver

– As of September 6, 2015

– Compare to 2013 dataset (CHI 2014)

• Characterize recipes & ecosystem

19

Methodology

• Collect all publicly shared IFTTT recipes

– Using Selenium & ChromeDriver

– As of September 6, 2015

– Compare to 2013 dataset (CHI 2014)

• Characterize recipes & ecosystem

• Share dataset with other researchers

20

Example

21

Example: Trigger Channel

22

Example: Trigger

23

Example: Action Channel

24

Example: Action

25

Example: Author

26

Example: Title

27

Example: Adoptions

28

Key Characteristics

29

Key Characteristics

• Huge growth from 2013 2015

30

Key Characteristics

• Huge growth from 2013 2015

– 67,820 recipes 224,590 recipes

31

Key Characteristics

• Huge growth from 2013 2015

– 67,820 recipes 224,590 recipes

– 35,495 authors 106,452 authors

32

Key Characteristics

• Huge growth from 2013 2015

– 67,820 recipes 224,590 recipes

– 35,495 authors 106,452 authors

• Many authors, but few are prolific

– Only 2.5% of authors shared 10+ recipes

33

Key Characteristics

• Huge growth from 2013 2015

– 67,820 recipes 224,590 recipes

– 35,495 authors 106,452 authors

• Many authors, but few are prolific

– Only 2.5% of authors shared 10+ recipes

• Many connections are being made

34

Key Characteristics

• Huge growth from 2013 2015

– 67,820 recipes 224,590 recipes

– 35,495 authors 106,452 authors

• Many authors, but few are prolific

– Only 2.5% of authors shared 10+ recipes

• Many connections are being made

– 15,961 unique trigger-action combinations

35

Key Insights

36

Key Insights

• IFTTT is big…and growing

37

Key Insights

• IFTTT is big…and growing

38

Key Insights

• IFTTT is big…and growing

• Adoptions vs. duplicated functionality

39

Key Insights

• IFTTT is big…and growing

• Adoptions vs. duplicated functionality

40

Key Insights

• IFTTT is big…and growing

• Adoptions vs. duplicated functionality

• Many diverse connections

41

Key Insights

• IFTTT is big…and growing

• Adoptions vs. duplicated functionality

• Many diverse connections

Action channels

Trigger

channels

42

Key Insights

• IFTTT is big…and growing

• Adoptions vs. duplicated functionality

• Many diverse connections

• ????????????

43

Download the dataset at

www.upod.io/datasets

Trigger-Action Programming

in the Wild:

An Analysis of 200,000 IFTTT Recipes

Blase Ur, Melwyn Pak Yong Ho, Stephen Brawner, Jiyun Lee,

Sarah Mennicken, Noah Picard, Diane Schulze, Michael Littman