This study examined how the content and style of online product reviews influence their adoption by consumers. [STUDY 1] analyzed over 700 reviews of Amazon Kindle e-readers to identify factors that increased or decreased review helpfulness. Reviews seen as more helpful used short sentences, provided evidence, discussed multiple product features, and came from experienced users with a positive but not overly emotional tone. [STUDY 2] then experimentally tested these findings. The results provide marketing implications like focusing online messages and identifying influential community members to create more impactful testimonials.

1. Online Persuasion:
How the Written Word Drives WOM
Evidence from Consumer-Generated Product
Reviews(Mar,2011)
Jin Li
Lingjing Zhan

2. Introduction & Background
• Importance of word of mouth in marketing
• Development of Word of mouth on Internet

3. PURPOSE
• To examine how the language style,
organizational structure, and other content
features of consumer-generated, online
product reviews affect review adoption.

4. Studies
• Study 1 : the authors analyzed a data set of
online product reviews regarding a consumer
electronic device and identified the content
features that positively or negatively
influenced review adoption.
• Study 2 : used an experimental approach to
probe the boundary conditions under which
some effects observed in study 1 may or may
not have occurred.

5. Study 1

6. Methodology
• STEP 1 Determine the platform and product
that can apply the research
PLATFORM Product

7. WHY AMAZON?
• Each reviewer can provided an overall
evaluation using a 1- to 5-star scale and a
qualitative product review. Readers also could
evaluate the helpfulness of a review on a
dichotomous scale (“Yes” or “No”), and they
could post comments after the review as well

8. Review
usefulness
Review content

9. Why kindle?
•People are more likely to consult other buyers’ views when they
consider purchasing high involved product(Riegner, 2007).
•Kindle is a brand new product

10. Methodology
STEP 2 SAMPLEING
• Sampling
They use the review on Amazon. But only the reviews that had been
evaluated by at least 50 readers (i.e., in the statement “m of n people
found the following review helpful,” n ≥ 50) were selected, for a total of
737 reviews for analysis.
• The distribution of early product ratings
The distribution of early product ratings was fairly balanced. Of the 737
reviews studied by the authors, 217 rated the Kindle with “five stars,”
67 with “four stars,” 78 with “three stars,” 100 with “two stars,” and
275 with “one star.”
217 67 78 100 275

11. Methodology
•STEP 3 determine Theoretical Framework of wording
ELM
MODEL
(Petty and Cacioppo,1986)

12. ELM MODEL
Ease of
comprehension
Evidence
Presence ARGUMENT
Opposing QUALITY
Viewpoints
Comprehens- Perceived
iveness review
helpfulness
Product Usage
SOURCE
Language CREDIBILITY
Intensity

13. ARGUMENT QUALITY1：論點品質的衡量
Study
Ease of
comprehension
– Average paragraph length
– Average sentence length
– Point format used (yes or no)
Evidence
presence
Evidence presence (yes or no)

14. ARGUMENT QUALITY
Balance
Opposing
viewpoints
Positive
Unbalance
Negative
Opposing viewpoints ( D1 , D2 )
Positive (1,0)
Negative (0,1)
Balance (0,0)

15. ARGUMENT QUALITY1：論點品質的衡量
Study
Comprehensiveness
• Review length expressed in number of
words
• Number of product features discussed
in the review

16. Ease of
comprehension
Evidence
Presence ARGUMENT
Opposing QUALITY
Viewpoints
Comprehens- Perceived
iveness review
helpfulness
Product Usage
SOURCE
CREDIBILITY
Language
Intensity

17. Study 1：來源可信度的衡量
SOURCE CREDIBILITY
Product Usage
Whether use kindle or not
Language Intensity
• Exclaim icon usage
• Positive attitude
• Negative attitude

18. Develop experiment
Independent variable
• Argument quality
• Source credibility
Dependent variable
• Review helpfulness

19. Methodology
STEP 4 ENCODING the review
•WHY?
Have to determine whether the independent variable appeared or not
(like emotion, evidence appeared, opposing view point)
•HOW?
Two native-English speakers were recruited as independent judges.

21. STEP 5 determine statistic model
Linear Predictor
j
ηi β0 β1 x1i ... β j x ji βk xki ...(1)
k 0
The independent variables are written as x ji (i = 1, …, 737 ，j = 1, …, 13)
Distribution Assumption) ai
ai yi ~ binomial (ai , pi ), i 1,..., 737 ...(2)
yi
yi：Number of readers
ai：Percentage of helpfulness
Link Function
e ηi
pi ηi
....(3)
1 e

22. OVERVIEW OF THE WHOLE RESEARCH
Encoding
Decide Sampling the product
product& product review into Run statistic
platform review theoratical
framework

23. RESULT

24. Ease of
comprehension
Independent variables parameter
Average paragraph length –0.244***
Average sentence length –0.212
Point format used 0.120
Evidence presence
Independent variables parameter
Evidence presence 0.992**
Means significant

25. Opposing
Viewpoints
Independent variables parameter
Positive 1.118***
Negative –0.107
Comprehensiveness
Independent variables parameter
Review length 0.677***
Product feature 0.280**
Means significant

26. Product Usage
Independent variables parameter
Product Usage 1.038***
Language
Intensity
Independent variables parameter
Exclaim icon usage –0.381***
Positive attitude 0.545**
Negative attitude –0.358***
Means significant

27. FINALIZED THE RESULT
The results indicate that helpful reviews exhibit five discernable
characteristics:
• They were comprehensive and easy to read.
• They were provided by reviewers who have usage experience with
the product.
• They presented supporting evidence for arguments.
• They provided positive information about the product.
• They may have contained strong positive emotions but not strong
negative emotions.

28. Conclusion

29. ADVANTAGE of the research method
1. Happened in a totally natural situation

30. DISADVANTAGE of the research method
1.Sample not drawn from the regular population
2.Encoding subjectively
3.Inability theoretical framework
4.Argument quality and source credibility are
not observable from online reviews
5. Ethic issue(?)

31. Marketing implication
• Companies can increase the usefulness of message on their
Web sites
• Companies can identifying the most influential voices in the
community.
• Help to create a good testimonial ads
• Use short sentences
• Offer evidence for claims
• Provide information on as many important
features as possible
• Avoid the use of exclamation marks.

33. What should a good

35. Thank for your you attention!