Falcon's Invoice Discounting: Your Path to Prosperity
Good News about Bad News – New Findings on Word of Mouth
1. BSI Good News about Bad News – New Findings on Word of Mouth
Marketing 2.0 Conference, Hamburg 2005
2. BSI
Join the conversation
MARKETING 2.0 CONFERENCE
www.marketing2conference.com
3. Good News about Bad News
New Findings on Word of Mouth
Presentation to WOM Conference
Robert East, Kingston Business School
R.East@kingston.ac.uk
4. Word of Mouth (WOM)
Communication between consumers about products,
services, companies etc
– we include emails and telephone advice
– we exclude advice from sales personnel
5. Positioning Our WOM Research
In the past WOM was Adoption of New Categories
implicit in research on the
adoption of new categories
– led to focus on early adopters
– the marketing problem was to
identify and target these few
Sales
early adopters
We are interested in WOM
about brands in established
categories
– leads to focus on all category
users, how they behave and
what has to be done to induce Time
them to give WOM
6. WOM: Influential but Under-Researched
WOM is often the most powerful channel for bringing
about change
– non-commercial, quick and the advice may be interactive
– Keaveney (1995), 50% found new providers via WOM
But little research, WOM is difficult to study
So mistaken beliefs about WOM persist:
– that there is more NWOM than PWOM - false
– that an instance of NWOM has more impact than an instance
of PWOM - often false for brand choice
I present evidence on these plus work on how WOM
relates to market share
7. Previous Research Has Related WOM to
Satisfaction
But satisfaction is about 5 times as common as
dissatisfaction in the USA (Peterson and Wilson (1992)
And much WOM is unrelated to the satisfaction of the
person giving it
– we recommend what will suit the other person even if we do
not like it ourselves
– Anderson (1998) found that those who were neither satisfied
nor dissatisfied gave 80% of the recommendation rate of
those who were extremely satisfied or extremely dissatisfied
8. Customers Who Are Neither Satisfied nor
Dissatisfied Still Give WOM (Anderson 1998)
14
12
10
8
Mean WOM
Sweden
frequency
6
4
2
0
Dissatisfaction Satisfaction
9. To Collect Data on WOM ...
We should
– collect data from those who use a category
– cover all the brands in a category
– do this for many categories
– measure the proportion of respondents who give WOM
(penetration) and how often they do so (frequency)
The incidence of WOM is penetration × frequency
We want the relative incidence of PWOM and NWOM
10. Measures
How many times have you recommended any dentist in
the last six months?
Please write in (0, 1, 2 etc) …….
How many times have you advised against any dentist
in the last six months?
Please write in (0, 1, 2 etc) …….
11. POSITIVE AND NEGATIVE WORD-OF-MOUTH FINDINGS
Service/product N Positive Word of Mouth in last 6 mths Negative Word of Mouth in last 6 mths Ratio
Penetration Frequency Incidence Penetration Frequency Incidence Ip/In
% per 100 % per 100
Ip In
Dentist 208 44 1.7 75 9 1.3 12 6.2
12. THE INCIDENCE OF POSITIVE AND NEGATIVE WORD-OF-MOUTH IN 15 CATEGORIES
Positive Word of Mouth in last 6 Negative Word of Mouth in last 6 Ratio
months months
Penetration Frequency Incidence Penetration Frequency Incidence IP/IN
% per 100 % per 100
(IP) (IN)
Restaurant 80 3.6 285 40 2.0 78 3.7
Elementary school 73 3.4 248 12 1.6 19 13.1
Car (univ. sample) 68 3.5 235 10 2.4 23 10.2
Coffee shop 77 2.9 219 41 1.9 80 2.7
ISP (gen. public) 56 3.7 207 38 5.0 186 1.1
Car (general public) 48 2.6 123 11 1.4 15 8.2
Cell-phone airtime 40 2.3 90 23 2.2 51 1.8
Car service 37 2.3 87 10 2.4 24 3.6
Dentist 44 1.7 75 9 1.3 12 6.3
Dry cleaning 28 2.2 62 5 1.2 6 10.3
ISP (univ. sample) 33 1.9 61 7 2.3 15 4.1
Credit card 31 1.8 58 16 2.3 37 1.6
Optician 36 1.5 55 21 1.5 31 1.8
Car insurance 24 1.5 36 1 2.0 2 18.0
House cont. insur. 16 1.3 21 3 4.3 11 1.9
Means (unweighted) 46 2.4 124 17 2.3 40
Ratio 3:1
13. Why Is PWOM Incidence Greater Than
NWOM Incidence?
First, there is about 3 times as much PWOM because
about 3 times as many people give it
– mean penetrations are 46 (PWOM) and 17 (NWOM)
– mean frequencies are much the same 2.4 (PWOM) and 2.3
(NWOM)
Many people may lack negative examples that they
could advise against
– this would reduce the NWOM penetration but not the
frequency, as we found,
– mostly PWOM is about their main provider
14. Are PWOM and NWOM About the Main
Provider?
Category, all UK Proportion of Proportion of NWOM
PWOM about about other providers
main provider than main provider
% %
Cars 98 87
Dentist 98 79
Bank 88 51
Mobile phone handset 88 87
Mobile-phone handset 80 80
Mobile-phone airtime 79 81
Mobile-phone airtime 74 52
Mobile-phone airtime 68 89
Restaurants 65 72
Restaurant 52 88
Mean (unweighted) 79 77
15. PWOM and NWOM Incidences Are
Positively Associated across Categories
Incidence of
NWOM
150.00
Linear Regression
R-Square = 0.28
100.00
50.00
0.00
50.00 100.00 150.00 200.00 250.00
Incidence of PWOM
16. Consumers Who Give NWOM Are Much
More Likely to Give PWOM
Across all categories 75% of all those who gave
NWOM gave PWOM
Those who gave NWOM were 3.5 times more likely to
be in the PWOM group than the non-PWOM group
(which were about equal in size)
17. Explaining WOM
What are the drivers of PWOM and NWOM?
– motivation: importance of category, relative evaluation of brands
– opportunity based on individual abilities: category knowledge,
verbal skill
– opportunity presented by the environment: salience of category
and brand, social contacts, others seeking advice etc
Motivation and opportunity boost both PWOM and
NWOM in categories - so they are positively associated
But lack of knowledge of negative cases is likely to
restrict NWOM more in some categories than others - so
ratio of PWOM to NWOM will vary across categories
18. We Corroborated the 3:1 Ratio with Two
Other Methods
Received WOM
– we measured received WOM as well as given WOM
Conditional Intention
– we asked respondents to state whether they would give
PWOM and NWOM if asked, or if the topic arose in
conversation
– this gives penetrations for conditional PWOM and NWOM
that can be compared with the penetrations for given PWOM
and NWOM
– the conditional WOM is elicited behaviour, not recall
19. The Relative Impact of PWOM and NWOM
on Brand Choice
A single instance of NWOM is thought to have more
impact than a single instance of PWOM
One reason could be that rare information has more
impact
– surprise causes attention (Berlyne and McDonnell (1965)
– rare info is more informative (Lynch, Marmorstein and
Weigold 1988)
–
But, although we have shown that NWOM is rarer than
PWOM, negative information about brands may not be
rarer than positive information
20. Rarity: An Example
People might say each of the following equally often
Waitrose has:
– higher quality (+)
– a more pleasant environment (+)
– a good delivery service (Ocado) (+)
– more expensive product (–)
– so, you could have 3 times as much PWOM as NWOM about
Waitrose although the NWOM and PWOM facts occur
equally often
Note: people assign fewer negative than positive
attributes to brands (Mittal, Ross, and Baldasare 1998)
21. Previous Work on Impact of PWOM and
NWOM
Arndt (1962) examined the real effect of WOM on
purchase of one new brand
– NWOM depressed purchase twice as much as PWOM raised it
– but brands/categories vary, so more evidence is needed
Alhuwalia (2002) used attitude change in lab studies
– if people were committed to brands, they discounted negative
data (but they accepted positive data)
– so negative data was not more influential
But we want to measure the impact on brand choice in
real settings
22. Procedure
To address the weakness of single-brand studies, we
conducted 15 studies on categories
To make the work less artificial, we asked “did the last
instance of PWOM/NWOM that you received affect
your decision?”
This gives the percentages of respondents whose
decisions were affected by PWOM and NWOM
23. The rates of PWOM and NWOM and their reported impacts
Category Number in sample Percent claiming Sig. of Impact Relative
receiving effect on decision difference ratio rarity
PWOM NWOM PWOM NWOM (4/5) (2/3)
1 2 3 4 5 6 7 8
Restaurant (1) 96 25 80 80 ns 1.0 3.8
24. Reported Impacts of PWOM and NWOM
Category Number in Percent claiming effect Impact Rarity
sample receiving on decision of … ratio ratio
PWOM NWOM PWOM NWOM (4/5) (2/3)
1 2 3 4 5 7 8
Restaurants, Iranian 79 58 100 73 1.4 1.4
Restaurants, type of cuisine 85 43 83 48 1.7 2.0
Restaurant, main 97 25 80 80 1.0 3.8
Restaurant, main 161 117 78 74 1.1 1.4
Restaurant, main 68 38 72 87 0.8 1.8
Holiday destination 131 92 71 64 1.1 1.4
Cell-phone handset 157 155 70 35 2.0 1.0
Optician, UK 48 28 67 43 1.6 1.7
Luxury brands 72 36 64 44 1.4 2.0
Cell-phone airtime 149 152 61 53 1.2 1.0
Coffee shop 74 60 54 72 0.8 1.2
Cell-phone airtime 49 54 47 26 1.8 0.9
Credit card, UK 92 85 40 42 1.0 1.1
Car, UK 134 110 34 36 0.9 1.2
Supermarkets 42 35 33 54 0.6 1.2
Mean (unweighted) 96 73 64 55 1.2 1.5
• Categories vary but, overall, PWOM has a little more effect than NWOM
(64% versus 55%, ns)
• PWOM and NWOM impacts are correlated (cols 4 and 5, r = 0.52; p < 0.05)
• The impact ratio is not explained by relative rarity (r = –.11, p = 0.69)
25. Alternative Measurement of Impact
“NWOM affected my decision” may indicate less
impact than “PWOM affected my decision”
– if most brands have little chance of purchase, NWOM cannot
reduce this chance much but PWOM might increase it a lot
– the measure should look at the shift in purchase probability
26. Preliminary Findings
Relative Impact of PWOM and NWOM
Category Probability of purchase Shift in probability of
purchase
Prior to Prior to After After
PWOM NWOM PWOM NWOM
Rest. Iranian 0.44 0.22 .31 –.02
Restaurant, fav. 0.35 0.59 .39 –.48
Restaurant type 0.36 0.41 .34 –.22
Supermarket 0.43 0.39 .16 –.16
Mobile phone 0.39 0.36 .20 –.07
Mobile airtime 0.32 0.41 .19 –.10
Luxury brand 0.54 0.56 .16 –.16
Means 0.40 0.42 .25 –.17
• Prior probabilities higher than we expected
• It appears that PWOM again has marginally more effect overall
but it varies by category
27. WOM in Relation to Market Share
Mobile Handsets
Brand MS % PWOM % NWOM %
Nokia 40 40 24
Sony-Eric 24 21 22
Motorola 14 20 15
Samsung 10 11 11
Siemens 4 2 9
Panasonic 2 1 6
Others 4 3 15
• PWOM follows market share
• NWOM for mobiles relates to MS but is flattened
28. WOM in Relation to Market Share
Mobile Airtime
Brand MS % PWOM % NWOM %
T-Mobile 25 23 14
Vodafone 21 24 8
Orange 19 15 29
O2 12 18 8
3 11 8 33
BTcell 8 8 5
Virgin, Fresh 2 4 3
• PWOM follows market share
• NWOM for airtime shows spikes for Orange and 3.
- Orange: 58% about coverage, 18% about reliability
- 3: 39% about coverage, 32% reliability and service
29. What Brand Did They Talk About?
Handset Airtime
PWOM NWOM PWOM NWOM
Current brand 81 20 79 19
Brand before current brand 5 45 4 32
Owned before that 10 26 5 24
Never owned 5 9 12 26
30. Applications
With a database
– reduce NWOM by focusing on complainants, ex-customers
and current customers (with different messages)
(remember, the person who gives NWOM is also likely to be
giving PWOM on another brand).
– increase PWOM by targeting current and, usually, recently
acquired customers
Develop metric for WOM performance
Identify the “talking points” used in PWOM and
NWOM and how these can be transmitted in
advertising on which media http://subaru.com.au/awd/
31. Finally
If NWOM is “bad news”, the “good news” is that it is
not that common and usually has less influence than
PWOM when it occurs
But it is a pity that this research has not been done
before
We need evidence-based marketing not marketing
folklore