View
3
Download
0
Category
Preview:
Citation preview
Optimum Blending of Panels and Social Network Respondents
Adam Portner - Research NowSteven Gittelman Ph.D. - Mktg, Inc.
CASRO Online Research Conference, Las VegasMarch 2011
Social media video
2© 2011 Research Now
The online population and social media
3© 2011 Research Now
Extend our sampling universe
Engage social media
Include to provide a richer, more comprehensive and inclusive sample
All sources are different!Modal differences in buying behavior segmentations
4© 2011 Research Now
What are the differences?
5© 2011 Research Now
Blending social media sample
6© 2011 Research Now
Different sample sources possess different characteristics
Important to ensure consistency – results must be repeatable
To avoid variability social media respondents must be blended by design with panel sample
A scientific blend
7© 2011 Research Now
A scientific blend seeks to achieve a standardIt must be transparent, documented and repeatable
Blending is different from “mixing” which lacks precision Blending is a robust process
Minimum measurable difference
8© 2011 Research Now
The threshold at which we begin to detect statistical difference at a level so low that it represents a conservative measure of similarity
Requirements A system of measurement – metrics
Tools to measure with – segmentations
Precision - minimum measureable difference
Standard - have a plan for what you are building, and what the sample will represent
Model based sampling
Methodology
9© 2011 Research Now
Sample sources: Valued Opinions Panel and Peanut Labs in the United StatesSample size: VOP - 4009 respondents
Peanut Labs - 3887 respondents TOTAL of 7896 Respondents
Fieldwork: 9/14/2010-12/19/2010Survey length: 17 minutesQuotas: Gender, age, household income & ethnicityGenerates ten segmentations:
• Buying behavior, 37 variables • Socio-graphics, 31 variables • Media, 31 variables• Plus seven other market segmentations
Structural segmentations
10© 2011 Research Now
By identifying groups of respondents who answer similarly across a broad array of behavioral items, we divide them into ‘segments’
The segmentations allow for behavioral standards by which to compare different sample sources. We call it a behavioral fingerprint; it is an important measurement tool
11© 2011 Research Now
What are the differences between social network and panel respondents?
Marital status was similar across sources
12© 2011 Research Now
Panel respondents are better educated
13© 2011 Research Now
Social network respondents use the internet more broadly and own more hi-tech devices
14© 2011 Research Now
Social network respondents use the internet more often for connecting with others
15© 2011 Research Now
Social network respondents use the internet for entertainment and news gathering more often
16© 2011 Research Now
Yes, they are different...
17© 2011 Research Now
18© 2011 Research Now
An absence of detectable difference implies similarity
Establishing a threshold
19© 2011 Research Now
A change in the assumed sample size alters the number of social network respondents that can be blended
20© 2011 Research Now
α=0.32
The greater the similarity found within demographic groups the more liberally we can blend them
21© 2011 Research Now
Maximum blending ratioAfter making conservative assumptions on income and segments these are the final blending ratios for each sex by age group
22© 2011 Research NowNote – this is subject to change pending future blending analysis
The final sample at maximum blending percentage has barely changed the observed behavior of the VOP sample
23© 2011 Research Now
The blend and VOP also remain similar on socio-graphics
24© 2011 Research Now
Media preferences remain similar at the maximum blending percentage
25© 2011 Research Now
Respondent engagement remains the same at the maximum blending percentage
26© 2011 Research Now
Panel tenure, an important diagnostic, tends to decrease after blending
27© 2011 Research Now
Conclusion
28© 2011 Research Now
Social media respondents represent a considerable and growingproportion of the global population
Social media respondents are different to panel respondents –opportunity to be more inclusive but must be able to deliver consistent data
Through comparative research we established the minimum measurable difference, in order to determine the maximum blending ratio
Transparency is essential – researchers must be confident that changes are real, not due to sample source
Blending analysis must be an ongoing process, blending ratios are not static and will change as sources evolve
Summary video
29© 2011 Research Now
Thank you!
30© 2011 Research Now
Adam PortnerSenior Vice President, Client Development220 Montgomery Street, Suite 1058San Francisco, CA 94104Direct: (415) 948-2230aportner@researchnow.comwww.researchnow.com
Steven Gittelman, Ph.D. President200 Carleton Avenue, East Islip, NY 11730Phone (631) 277-7000Cell (631) 466-6604 Steve@mktginc.com www.MktgInc.com
Recommended