Seeking Patterns in Survey Results
John DeRiggi
DAI
Patterns in Data
DAI projects collect a lot of data in the field.
Patterns in the data can sometimes help us identify more targeted approaches to the delivery of our development work.
This presentation demonstrates of some of the visualization and analysis techniques we use to identify those valuable patterns.
Case Study
In the spring of 2016, Adam Fivenson collected a data sample for DAI's Consumer Insights study in Honduras.
One slice of the data we found interesting was related to the use of texting, Facebook, and WhatsApp.
This analysis seeks to identify some patterns in the data using a few simple techniques.
Scatterplotting Facebook vs Whatsapp
(minutes) 213 responses
Let's start this analysis by looking at scatter plots of the survey responses
Scatterplotting Facebook vs Whatsapp
(minutes) 213 responses
Specifically we are looking at the amount of time (minutes) people said they spent on Facebook...
Scatterplotting Facebook vs Whatsapp
(minutes) 213 responses
...and the amount of time (minutes) reported on WhatsApp
Time Spent on Facebook vs Whatsapp
(minutes) 213 responses
It is difficult to discern any interesting patterns looking at this scatterplot alone
Time Spent on Facebook vs Whatsapp
(minutes) 213 responses
It is difficult to discern any interesting patterns looking at this scatterplot alone
Let's see if looking at extremes in the data reveals patterns
Time Spent on Facebook vs Whatsapp
(minutes) 213 responses
This user spends a massive amount of time on Facebook and WhatsApp (probably over-reporting)
Time Spent on Facebook vs Whatsapp
(minutes) 213 responses
This user spends a massive amount of time on Facebook and comparatively little on WhatsApp
Time Spent on Facebook vs Whatsapp
(minutes) 213 responses
Let's try to identify all the super users of these systems: those who spend a lot of time with WhatsApp and Facebook
To do that we need a way of classifying what counts for a lot of time
Identifying Super Users of Facebook and WhatsApp
Super users are those who spend a lot of time compared to the rest of the survey responses
Identifying Super Users of Facebook
(minutes) 213 responses
We want to define a cutoff line for the users who spend a lot of time on either system
Identifying Super Users of Facebook
(minutes) 213 responses
We want to define a cutoff line for the users who spend a lot of time on either system
Let's use the standard deviation plus the mean to define a class of super users
Identifying Super Users of Facebook
(minutes) 213 responses
This line is one standard deviation plus the mean =
270.26
Identifying Super Users of Facebook
(minutes) 213 responses
Responses above the red line are super users of Facebook.
They spend a lot of time on FB
compared to all
the survey responses
Identifying Super Users of WhatsApp
(minutes) 213 responses
The green line is one standard deviation plus the mean =
398.94
Identifying Super Users of WhatsApp
(minutes) 213 responses
Respondents to the right of the green line use WhatsApp a lot
Identifying Super Users of WhatsApp Or Facebook
(minutes) 213 responses
In this chart we have all segments of super users for both systems
Identifying Super Users of WhatsApp Or Facebook
(minutes) 213 responses
In this chart we have all segments of super users for both systems
facebook alone
Identifying Super Users of WhatsApp Or Facebook
(minutes) 213 responses
In this chart we have all segments of super users for both systems
facebook alone whatsapp alone
Identifying Super Users of WhatsApp And Facebook
(minutes) 213 responses
In this chart we have all segments of super users for both systems
facebook alone whatsapp alone
and both
Let's Segment the Data to See if Patterns Emerge
Responses are tagged with gender for the respondent and urban or rural depending the household location
Segmenting the Survey Results by Two Categories: Gender and Urban/Rural
Survey Results by Category
The survey sample is mostly urban and has slightly more female than male respondents
Segmenting the Survey Results by Two Categories: Gender and Urban/Rural
Survey Results by Category
Let's see how the responses are distributed based on these categories
The survey sample is mostly urban and has slightly more female than male respondents
Patterns in Urban and Rural Responses for Time Spent on Facebook and WhatsApp
Scatterplots for each of the four segments in the survey
Patterns in Urban and Rural Responses for Time Spent on Facebook and WhatsApp
PATTERN: Super users are all urbanites, with one exception (male rural)
Patterns in Urban and Rural Responses for Time Spent on Facebook and WhatsApp
PATTERN: Super users are all urbanites, with one exception (male rural)
Patterns in Urban and Rural Responses for Time Spent on Facebook and WhatsApp
PATTERN: Super users are all urbanites, with one exception (male rural)
PATTERN: All super users of WhatsApp are urbanites
Patterns in Urban and Rural Responses for Time Spent on Facebook and WhatsApp
Let's see if we can identify preferences in the responses:
Which systems do users spend a greater portion of their time with?
Patterns in Urban and Rural Responses for Time Spent on Facebook and WhatsApp
PATTERN: Responses show a greater portion of time spent on WhatsApp than FB, especially among urban men
Patterns in Urban and Rural Responses for Time Spent on Facebook and WhatsApp
PATTERN: Responses show a greater portion of time is spent on WhatsApp than FB, especially urban men
Background: The preference for group messaging over social feed platforms is a global trend
Let's Compare Texting to WhatsApp
Seeking patterns in messaging platforms
Patterns in Urban and Rural Responses for Use of Text Messaging and WhatsApp
PATTERN: Super users of text messaging exist in rural areas
Patterns in Urban and Rural Responses for Use of Text Messaging and WhatsApp
(minutes) 215 responses
PATTERN: Rare is the super user of both text messaging and WhatsApp
PATTERN: Super users of texting exist in rural areas
Patterns in Rural Responses for Use of Text Messaging and WhatsApp
TEXTING vs WHATSAPP
FACEBOOK vs WHATSAPP
PATTERN: Rural super users tend to be super users of SMS. Only one rural facebook super user in the responses
-
Super users of FB and WhatsApp are urbanites
- Superusers exist in rural areas but only use text messaging
- WhatsApp seems to displace the use of text messaging in urban areas
- All WhatsApp super users were in urban areas
- A greater portion of time spent on WhatsApp than FB - in line with the global trend towards group messaging
Wrap Up: Patterns in the data
Thanks!
digital@dai.com
John DeRiggi and Adam Fivenson