Transcript Document

New Dimensions of Quality in Online Panels
Jacqueline Lorch
Vice President, Global Knowledge Management
Survey Sampling International
The World Is Moving On…
Why Does It Matter?
Questionnaire Design
Fieldwork
Sampling
Interpretation
What Are The Key Quality Elements?
 A broadly-recruited, representative, well-managed pool of
respondents

Respondents who respond honestly and conscientiously:
 What guarantees are there to guard against bad data, i.e. respondent cheating or
not concentrating/caring in their responses… ?

A well-designed survey instrument
How Do Respondents Fit In?
 Google
“Market Research”
464 million entries
“Market Research Respondents”
2 million entries
What Is A ‘Professional Respondent’?
 Some assumptions:
 Quality of responses will be lower
 Motivated to maximize incentives, not by intrinsic
interest
 Do too many surveys – become conditioned
 But what definition?
 Someone who gets paid?
Honesty Is The Issue
“I am an IT Director with 7 wives…”
New Typology Of “Bad” Respondents
 Hyperactive Respondents
 Too many surveys, too many panels
 Fraudulent Respondents
 Misrepresent themselves
 Inattentive Respondents
 Don’t put thought into answers
 Conditioned Respondents
 Have learned from past surveys
Smith and Hofma Brown, Harris Interactive
Hyperactive Respondents
 Do busy panelists provide bad data?
 Do the most responsive panelists take
the surveys the fastest?
 Are the fastest surveys the ‘worst’?
Hyperactive Respondents
 US SurveySpot panel, 3 month internal tracker
 Two groups:
– SurveySpot only
– One or more other panels
 No difference in average time taken by each group
 Similar answers on motivation for joining:
 Motivated by chance to influence decisions
– SurveySpot-only: 50%
– More than one panel membership: 47%
Fraudulent Respondents
 How can you tell if someone is who they say they are?
 How can you stop multiple panel memberships using different
identities?
 How can you tell if someone is just making up the answers?
Who Are You Today?
Ask me something only I
would/should know…..
Catch The Cheat…
 Respondents…
 Who have 23 different ailments
 Who report using non-existent brands
 Whose education doesn’t match their profession
Inattentive Respondents
 Fatigue leads them to skip
questions
 Don’t pay attention to
instructions
 May be inevitable after a
certain length of survey
Inattentive Respondents
 Related to interview length
Length of interview, minutes
14
19
21
Items
15
5
15
Skipped all
5%
8%
11%
 5% baseline of inattention?
Respondents Speak Out
 Repetitive questions “sorry this survey was just too long.”
 “Sometimes it becomes so repetitive you say, ‘to hell with it, I
don’t need this.’”
 “You think you are about done
and the same questions start
all over again.”
Inattentive Respondents
 Time use > 7300 hours = 4 hours of sleep a night
 Time use > 8760 hours = no sleep, ever.. more hours than in a year
Length of interview, minutes
14
19
21
>7300 hours
5%
6%
6%
>8760 hours
3%
4%
4%
 5% baseline of inattention?
 A really tough question set?
Respondents Speak Out
 “…absolutely ridiculous… [questions like] ‘if this
pizza was a person.’ ”
 “Why is this bottled water like your favorite pet?”
 “In filling out this survey it asked when I would buy a new house. I
said “never”. The next several questions were regarding my new
house and it required an answer…So I quit the survey and didn’t
finish it.”
Conditioned Respondents
 Only give the answers they
do because of what they
have learned from previous
surveys
 Or change behavior as a
result of information from
surveys
 Have no redeeming features
 But do such people exist…..?
TNS Experiment
 3 groups of respondents
 High frequency – interviewed 5 times (1.2.3.4.5)
 Medium frequency – interviewed 3 times (1.3.5)
 Low frequency – interviewed 2 times (1.5)
 + Control group at wave 5
 Same questionnaire
 Survey frequency not yet released by TNS
 UK, France, Germany
 n = 1202 (control = 1470)
Product Usage
High
Low
Product Purchase
High
Low
Brand Awareness
High
Low
Conditioned Respondents
 Evidence is hard to find
 Maybe surveys aren’t quite so important to
respondents as they are to researchers!
 Could we be over-reacting and losing
good respondents?
Respondents Speak Out
 “I would like to know why it is that practically every time I give my age
I am refused the access to the survey? Is there something wrong with
being 75 and in good health, mentally and physically?”
 Why do I spend 10 minutes answering questions on one of your
surveys before I am given message ‘Sorry, you didn't qualify for this
survey’.
 “I have not been able to take several [most] surveys because I work in
the grocery industry. I don’t think it is fair. I AM A CONSUMER TOO …
It is not like I work for a company that makes or sells one brand.”
Respondents Speak Out
 “…ability of the human eyes to see fonts of microscopic size. What’s
with this survey? The font size was like .002.”
 “I said yes before; why are you going back and asking me the same
question.”
 “Right to the point”…“if they say it’s 5 minutes, it’s 5 minutes.”
 “By and large ask sensible questions in a straightforward way.”
What Needs To Happen?
 Partnership…common terms and definitions
 Avoid red herrings
 We are not in the business of supplying bad panelists
 Survey design is critical
 We can’t do it alone