, Fall 2011
This
note details the data from the new multiplatform audience questions that will
be released in the Fall 2011 report.
In
Wave 65 (March – October 2011 interviewing period), GfK MRI introduced a number
of questions about reading or looking into print brands on computers, tablets,
e-readers and other mobile devices. The
new methodology includes a change in the preamble to the magazine reading
section, additional visual and verbal stimuli to prompt recall and new
qualitative readership questions for electronic devices. (We have already released a white paper
describing these changes and the results from a pilot test conducted last
year. This report is available on GfK
MRI’s website at www.gfkmri.com.) The
comprehensive audience data will be reported on two levels. They are:
- Any
print audience: We will continue to
report the print audience for all magazines that meet release standards on all
available software provider systems and in a “Readers-Per-Copy Card” (or
“Pocketpiece”). These data reflect the
continued recent reading methodology that GfK MRI has employed since its
inception.
- A
total multi-platform audience: We will
report the net audience for print brands across all the platforms and versions
(i.e., websites, digital reproductions, apps and other branded apps, text only)
measured in the study. These data will
only be available for one wave of the Fall 2011 report and will be noted in
that manner. Any user will need to scale
the Wave 65 total multi-platform estimates to the population for proper
assessment. We will also report total
multi-platform audiences for magazine groups.
Users
of these data should note that the software provider’s location of the
readership measures will be changed beginning with the Fall 2011 report. Please consult your software provider and the
Fall 2011 codebook for further information.
In
addition to the changes in the magazine measurement procedures, GfK MRI changed
the wording of the initial newspaper question to include only paper version
readership. Electronic versions of The
New York Times, USA Today and The Wall Street Journal are measured after all
magazine readership questions are asked.
GfK
MRI will not release estimates for individual platform audiences. Our continuous in-wave evaluation of tablet,
e-reader and mobile device audiences suggests that respondent counts are not
yet sufficiently stable for individual platform reporting. We expect these audiences to grow in the
upcoming waves and we will release these data once reporting standards are
met. We will also continue to compare
responses against the availability of brands on individual platforms to ensure
the validity of responses.
Magazine
website levels are more robust than data from the other electronic platforms
beyond print. However, the stability of
these data varies substantially among the magazines and warrants further review
before any individual magazine data will be released. GfK MRI employs a recall method, which by
the very nature of the questioning procedure, can produce results differing
from passively collected data from services such as Nielsen’s and comScore’s
web audience measurement, respectively.
(Users should be aware that GfK MRI has developed a fused database with
Nielsen’s web audience measurement service and is in the process of completing
a fusion with comScore’s web audience measurement service.) As part of our ongoing analysis, we will
explore these potential differences.
One
final note: As earlier stated, GfK MRI
removed the “interest in advertising” and “actions taken” questions from the
National Study. In their place, we plan
on integrating individual magazine topline “ad noting” and “actions taken” data
from Starch AdMeasure with the National study.
These data will not be available with the first release of print
estimates in November. The details of
this integration, including submitting this process for MRC accreditation and
timing, will soon follow.
, Spring 2011
Beginning with Wave 64
(the second wave of the Spring 2011 Report), GfK MRI has added the question
asking “language personally spoken in the home” to its set of variables used in
the sample balancing algorithm.
Respondents are classified into one of five mutually exclusive
classifications. They are:
- Speaks
only English
- Speaks
mostly English, but some Spanish
- Speaks
mostly Spanish, but some English
- Speaks only Spanish
- Speaks
both equally or other language
The weighting is only
applied to Hispanic respondents in the survey.
GfK MRI is using the
Nielsen 2011 universe estimates for these categories in the sample balancing
algorithm. Nielsen is considered to be
the standard for establishing language propensities among Hispanics.
The Media Rating
Council (MRC) Guidelines require accredited companies to communicate the
expected impact of this change on audience ratings and sampling
efficiency. To comply with that
standard, the following table shows the approximate impact on print audience levels
for a single year’s estimates:
|
Number Of Publications
|
|
% of Total Publications
|
|
% Relative Change
|
|
5
|
|
2.31%
|
|
+2.00% or greater
|
|
|
6
|
|
2.78%
|
|
+1.00 to +1.99%
|
|
29
|
|
13.43%
|
|
0.00 to +0.99%
|
|
83
|
|
38.43%
|
|
0.00 to -0.99%
|
|
60
|
|
27.78%
|
|
-1.00 to -1.99%
|
|
|
26
|
|
12.04%
|
|
-2.00 to -2.99%
|
|
7
|
|
3.24%
|
|
-3.00% or greater
|
|
Total
|
216
|
|
100.00%
|
|
|
The expected
relative changes, while generally extremely small, reflect the increased
weights given to Spanish dominant respondents who are generally less likely to
read English language magazines. Additionally, it is reasonable to assume that
any media brand that reaches disproportionately more Spanish dominant Hispanics
than English dominant Hispanics would experience some audience increase with
the introduction of this sample balancing variable. The opposite effect is likely for those media
brands that attract disproportionately more English dominant Hispanics than
they do Spanish dominant Hispanics.
It is also important to note the
relative impact on audiences in the Spring 2011 Report is only half of the
estimated changes in the table, since the Hispanic language weighting variable
was introduced for only one of two waves in the study.
Our analysis
of the effect on effective sample size reveals that there is an approximate
loss of 1% in statistical efficiency.
, Fall 2011
Beginning in March, 2011, corresponding with GfK MRI’s Fall 2011 report, GfK MRI implemented a change in its national survey, in order to measure the readership of printed magazines and newspapers, as well as magazine and national newspaper readership on varying types of digital platforms.
As the popularity of e-readers and tablets, such as the Apple iPad, has increased, the print industry has introduced new and innovative ways to consume print and print-brand related products. Many believe that digital versions of magazines and national newspapers could eventually become a significant source of readership for print brands. Consequently, GfK MRI recognizes the industry need for better understanding of how consumers read digital print.
In order to measure the total audience of print brands across all platforms, GfK MRI had to make a fundamental change to the question wording used to measure print audiences, while preserving a consistent measure of hard-copy audiences. To prepare for the transition to the new measurement, GfK MRI undertook a large-scale, in-person pilot test in 2010 designed to assess the potential impact of the changes on hard-copy audience estimates and to evaluate how survey respondents understood and responded to the chosen descriptions of different print electronic versions. The test involved 500+ people who were shown the newly worded personal interview questionnaire and a control group of 500+ people who were shown the existing GfK MRI questionnaire. We found that the language was understood by respondents and the questionnaire administration was smooth. In addition, the wording we tested had no statistically significant impact on the level of hard-copy readership. (Although, it should be noted that the rapidly changing landscape of print may ultimately affect hard-copy audiences)
Based on the results of the test, we are implementing the revised language and questioning sequence, beginning with the Fall 2011 report. The new questions will capture reading for print websites, as well as other digital reading such as on a computer, an e-reader, a tablet or another mobile device. It will capture reading of digital magazines, as well as magazine-branded “apps”.
The first field period begins in March, 2011 and one wave of data will be completed by November 1, 2011. Release of these additional data will be determined based on in-tab/stability criteria. It is expected that, at a minimum, we will be able to release one wave of a total brand estimate (paper + electronic) for most magazines in Fall 2011. In addition, we will release more detailed information about electronic reading for those magazines with a requisite number of respondents. Only stable estimates will be released, so accordingly, the level of detail released will be dependent upon the results of the study.
For more detailed information about the revised questionnaire language, go to:
GfK MRI Wave 65 Magazine Questioning Language
For additional information about the results of GfK MRI’s pilot test, go to:
Presentation:
GfK MRI Transitioning to Multiplatform Print Brand Measurement
White Paper:
GfK MRI Test of Preamble Wording for Publishing Brands
, Fall 2009
During Waves 60 and 61, with MRC approval, MRI conducted a live test of Broadcast network viewing in the Atlanta DMA.
The intent of the test was to evaluate the addition of questions seeking to measure Broadcast network viewing in a manner comparable to MRI's Cable/Satellite network measurement. For the test, MRI asked the same two usage questions (any watching in the last 30 days, number of hours watched in the last 7 days) that are asked regarding Cable television viewing. Respondents were shown one additional showcard, listing 7 Broadcast network logos. The question was asked of all respondents in the tested sample, not only Cable/Satellite/Fiber Optic subscribers. MRI added these questions to the end of the cable/satellite section in order to minimize change and potential disruption to the questionnaire.
In order to evaluate the test, MRI examined the mean number of cable stations viewed in a 30-day period and compared the results between the control and test groups. Our hypothesis/concern was that, even though the broadcast questions appeared after the cable questions, the addition of these questions would somehow have a negative effect on the levels of cable viewing, generating significantly fewer average mentions of cable stations viewed by a respondent.
To conduct the test, MRI selected cluster samples in the Atlanta DMA and randomly assigned them to the test or control condition. 252 and 244 respondents received the control questionnaires in Waves 60 and 61, respectively. 234 and 240 respondents received the test questionnaires, respectively in Waves 60 and 61.
MRI tracked the results of the test during the two field periods. After each field period closed, MRI conducted a one-tailed t-test of difference between the mean number of cable stations viewed by respondents and determined that the difference in responses between the test group and the control group was not significant in either wave (see attached charts for details.) In the Wave 60 test, the average number of mentioned cable stations was 14.51 in the test group and 15.26 in the control group. A one-tailed t-score of 0.70 was obtained, which is below the significance level of 1.645, at a confidence level of 95% (for a one tailed test). In the Wave 61 test, on the other hand, the difference between the means was not even in the expected direction, with an average number of mentioned cable stations of 16.99 for the test group and 16.75 for the control group. These findings led MRI to conclude that the results of adding the new broadcast network viewing questions had no significant impact on the number of cable stations mentioned by respondents.
Based on the findings from the statistical analyses, MRI decided to roll-out the new questions for all markets starting with Wave 62. Since the test was only conducted in a single market, MRI will not release data from the test in the Spring or Fall 2009 reports.
For additional information, please contact your MRI Representative.
, Fall 2009
To accommodate client requests and in an effort to improve MRI’s reporting of Internet behavior, beginning with the Fall 2009 report, MRI is using a different set of questions to calculate Internet quintiles. The new Internet quintiles will be based on three Time Spent questions (“Number of Hours used the Internet in an average week”). Prior to Fall 2009, Internet quintiles were based on Times Used (“Number of times Used the Internet in a typical month”.)
Clients have requested that MRI change the quintiles since the Time Spent questions are a more commonly used measurement of respondent behavior. In addition, MRI believes that the Time Spent questions (specifically, “Thinking of yesterday/last Saturday/last Sunday, how much time in total did you spend using the Internet?” 10 hours or more, 5 to less than 10 hours, 2 to less than 5 hours, ½ hour to less than 1 hour, less than ½ hour or None.) may be easier for respondents to estimate than Times Used (“On the average, how often do you look at or use the Internet in a typical month?” 5 or more times per day, 2-4 times per day, once a day, 3-6 times per week, 1-2 times per week, and Less than once a week).
Media Usage scores have also been changed from “Number of Times Used in a Month” to “Number of Hours spent in an Average Week.”
For additional information, please contact your MRI Representative.
|