Social Listening and Ratings

Question 1

Are people more likely to be influenced up or down in social ratings?

Answer 1

Social Influence Bias: Positive herding is more likely than negative herding

Question 2

How to leverage positive herding for social ratings?

Answer 2

Encourage customers who had a positive experience to rate and rate EARLY because it will influence all the other ratings to follow!

Question 3

Social Influence Bias

Answer 3

Customers tend to rate more positive when they see others have. Positive herding is more likely than negative herding
* Do positive polls predict or influence election results?

Question 4

Social Influence Bias by Topic

Answer 4

Businesses and Culture / Society highest impact

Question 5

Incentives to review

Answer 5

Ok as long as you don’t tell them what to write

When yo write your book, launch your business, send emails to customers: If you’ve read the book, please review…

Question 6

Why Online Reviews are Critical

Answer 6

2nd Most Trusted Source of Brand Information
2/3rds of Consumers Trust Reviews, up 15%
92% Read Reviews
Of which:
46% Influenced to Purchase
43% Deterred from Purchase
3% of Decisions “Unaffected”

Question 7

Reviews with poor grammar

Answer 7

Zappos used Mechanical Turk to edit grammar and spelling in reviews at $0.10 per review. Sales went up across the board.

Question 8

Amazon Mechanical Turk

Answer 8

Can go hire real people to do human intelligence tasks that a computer isn’t good at. Pay cents to look at images and review items humans are better at.
Zappos used this at $0.10 per review. Sales went up across the board.

Question 9

Captcha

Answer 9

Crowd Sources for people to inform machine learning for Google and others to recognize things. like street signs etc.

Question 10

Percent of discussions

Answer 10

Don’t necessarily correlate to sales impact

Not enough to understand what people are talking about, need to look at what they do.

Question 11

UGC

Answer 11

User Generated Content

Question 12

Reputation & Pricing Power

Answer 12

Impacts willingness to pay

• Number of Stars
• Number of Past Transactions
• Text!

Question 13

o How to decompose scoring reputation

Answer 13

 Use price premium as true reputation score and build regression
 Take number of mentions of a keyword or phrase
 Y = sellers price / avg price of all items OR sales OR price
 X = star rating, number of reviews, number of past transactions, “fast delivery”, “good packaging”, etc..

Question 14

Potential text analytics tools

Answer 14

Google engrams?

Question 15

How to decompose scoring reputation and unstructured text

Answer 15

 Use price premium as true reputation score and build regression
 Take number of mentions of a keyword or phrase
 Y = sellers price / avg price of all items OR sales OR price
 X = star rating, number of reviews, number of past transactions, “fast delivery”, “good packaging”, etc..