In This Issue

Volume I March 2007

Forward This to a Colleague

Q&A with Dr. Julian Baim, EVP & Chief Research Officer, Mediamark Research Inc. 

Over the past several years, MRI’s average-issue audience for all print publications has included all claimed readership, regardless of the source of copy. This has been the consistent reporting standard since MRI’s inception. However, the Internet has prompted almost all publications to create their own websites and/or to develop an Internet deliverable of the printed edition. This change has, in turn, led to respondents’ reporting reading on the web bearing the print media’s name (or brand), although the respondent did not read the hard copy of the magazine. Therefore, MRI has made the decision to exclude web-only (or Internet exclusive) readers beginning with the Spring 2007 release of The Survey of the American Consumer. In this Q&A, Dr. Baim discusses issues surrounding that decision.

Q. Exactly what data will clients receive as a result of the decision to exclude web-only readers from the Spring 2007 data release?

A. MRI will make available in a title’s Total Audience estimate only those respondents who indicated they read the printed copy of the publication. That includes anyone who says they read only the print version and anyone who said they read the printed version and the online version as well. It will exclude those who said they only read online. 

Q. What is the key reason for this decision?

A. Primarily because web-only readers do not always have the same exposure to ads shown in the print edition.

Q. What about sites that display an exact replica of the printed publication?

A. MRI has done considerable research over the past several years to better gauge what it is readers actually read online, and it has been determined that an overwhelming majority of “web readers” are not reading an exact replica of the magazine. As such, the electronic exposure to advertising is extremely likely to be different from the print exposure to advertising.

Q. MRI announced this decision to clients on March 22. How else have you communicated to clients our findings that "web-only" readers were in the Total Audience estimate?

A. MRI issued its first communication on this issue four years ago as we began to make a punch code available to identify those readers who said they read a particular publication online or via the Internet. And, at the same time, we also began to highlight those titles in our “Pocketpiece” that were receiving more than 10% of their readership exclusively online or via the Internet. 

Q. When will this change take place and is it retroactive?

A. This change will affect data from the second wave (Wave 56) of the Spring 2007 release and all future data releases; it will not impact data from earlier releases. We will continue to measure “on the Internet” as a source of copy to isolate the web-only readers.

Q. How do you define the web-only reader (or the Internet exclusive reader)?

A. The web-only reader is identified based on respondents’ response to a two-part question: how did you obtain a copy of the publication; if they answered the “Internet,” the respondent is then asked if there is any other way they obtained the publication, and if “no other way” is noted, that reader is classified as a web-only reader.

Q. How many publications will be significantly impacted by this decision?

A. Currently, there are eleven publications, of the 229 we measure, that have more than 5% of their total audiences estimated as web-only or Internet exclusive. 

Q. Are there other developments in the works to help the industry better gauge the shift in reading habits and help marketers reach readers no matter how they interact with a publication?

A. MRI has been working with Nielsen//NetRatings (which measures online audiences) in an effort to fuse our databases. A key benefit of the fusion will be to provide the total dual-platform audience reach of a publication: the net reach of the Internet and the hard copy print vehicle. We plan to release this database this spring. Stay tuned … 

back to top

Although technology offers more ways than ever to instantly talk, text and stay connected, many consumers are off the grid as far as survey research is concerned. Researchers who rely on telephone interviews are feeling the impact that the increase in cell-only users has on their ability to conduct surveys representative of a certain population.

Linda B. Piekarski, vice president database and research at Survey Sampling International, helped spearhead the formation of a committee of researchers and methodologists from the private and public sectors to evaluate the impact of cell phone only users on telephone survey research. In 2003, the group – including Paul J. Lavrakas and Chuck Shuttles from Nielsen Media Research, Ed Cohen from Arbitron, Jim Lepkowski of The University of Michigan, Clyde Tucker, Bureau of Labor Statistics and Linda Piekarski – organized their first Cell Phone Summit to explore the cell-only trend.

"Historically, we never worried about cell phones; in fact we spent years excluding them in our sample design," Piekarski said. "We figured nobody would want to be bothered on their cells and including cell phones users created probability of selection issues."

When the committee began their investigation, their first challenge was to find out how many people were cell-only users. MRI’s Survey of the American Consumer and NHIS (National Health Interview Survey – conducted by the Center for Disease Control) provided reliable estimates through their large-scale face-to-face surveys. In 2003, MRI reported 3.6% of the population was cell-only and NHIS reported 3.2%. Since both probability samples reported the same findings, these data clarified the levels of cell phone versus landline usage. The committee also looked at MRI and NHIS demographic data to gauge the impact of cell phones when surveying various groups within the U.S. population.

"MRI data comes out six months before the NHIS research, so it gives us a preview of what to expect," said Piekarski. "We know both sources report similar findings."

The group continues to meet, and their work is ongoing. They are instrumental in the cell phone related portions of the upcoming 2007 American Association of Public Opinion Research (AAPOR) Conference in Anaheim, CA. At this conference, sessions devoted to issues relating to cell phones and research, as well as a short course Cell Phones and Survey Research, will be offered. The committee has worked with AAPOR to develop content for these sessions and some of the committee members will teach segments of the short course.

According to the MRI Fall 2006 data release, cell-only users have grown to be approximately 12% of U.S. households. The prevalence of cell-only usage has led the group to investigate the inclusion of cell-only respondents in telephone sampling. This, of course, has led to a whole new set of issues: how to randomly generate cell phone numbers, how to establish guidelines for conducting interviews on cell phones and how to weight the cell-only sample. 

Family cell phone plans and the high mobility of young adults present another challenge. For instance, students attending a college away from home might be using phones assigned to their families’ cell phone plan that don’t reflect the area code where they currently live. Screening for current place of residence is an important component of cell phone interviewing.

Another major challenge of incorporating cell-only consumers into a survey is not related to respondents; it’s related to regulations. The Telephone Consumer Protection Act prohibits use of an auto dialer, or any computer telephone equipment, to call a cell phone. "If researchers are going to include cell phones in their samples, they are going to have to hand dial the numbers," Piekarski said.

"Right now if you conduct a landline survey you are not including the approximately 12% of consumers that are cell phone only, so we are back to where we were in the '70s when telephone research was in its infancy and telephone methodologies were just being devised," said Piekarski. "We knew then that telephone households were growing and we knew that non-telephone households were pretty much alike. Now we are finding that the cell phone only population is growing and also has unique characteristics, based on data from MRI and NHIS. Not only are they more likely to be young, single, renters living in roommate situations, but they are also more likely to be under-insured, not get flu shots or other vaccinations, have no regular place for medical care, have been tested for HIV, and to be smokers and drinkers." 

Piekarski and her committee wonder how today’s cell-only users will change as they become older. Is this cohort staying wireless only? Once they are homeowners and there are children in the household, will their cell-only status change, and is that change going to be consistent over time?

This is the kind of information only MRI and NHIS provide – and the closer the data are tracked, the more this committee can say with assurance that the research industry needs to include cell phones in telephone surveys. 

back to top

Teenage girls may be fickle with friends, but they can be counted on when it comes to brand loyalty. According to MRI’s Teenmark study, teenage girls demonstrate consistency in sticking with certain brands and, across the board, they show significantly more brand loyalty than boys. Teenmark reports on the product usage, behavior and mindsets of U.S. teenagers ages 12-19. 

More teenage girls than boys reported they purchased the same brand three consecutive times for 21 out of 22 products. Perfume/fragrance, disposable razors and body moisturizer are the products where differences in brand loyalty are most pronounced between girls and boys. Overall, however, both boys and girls are most loyal to their brand of toothpaste, soda and shampoo. 

back to top

The State of the Art of Fusion: from Theory to Practice 

By MRI Vice President, Research Operations Risa Becker and MRI Senior Vice President for Research Jim Collins

Currency measures of the different major media – television, radio, print, the Internet and outdoor – are founded on different methodologies. Each of these is designed and executed to optimize the measurement of its particular media to the highest quality standards reasonably attainable. To this end, media currency measurements generally focus on measuring a single media within a single population sample – to better reduce respondent burden, increase cooperation and, hence, increase measurement quality.

Media researchers recognize and rely on the quality of these separate currency measures, but at the same time, their separateness complicates their work. Planning a multi-media campaign schedule involves analyzing the vehicles within the individual media types separately and then combining the chosen options across the media types to derive the campaign’s reach, GRP’s, etc. Wouldn’t life be easier if these separate currency measures did not need to be combined; if they were in the same data source to begin with; if interactions among all vehicles across all media could be assessed together? 

What researchers have wanted since the dawn of media research is a currency-quality, single-source multi-media measurement. Too bad that "single-source multi-media" and "currency-quality" conflict with one another – and to avoid compromising the latter, researchers routinely live without the former.

But there are alternatives to this dilemma, all related to what media researchers have been doing more or less formally for decades; integrating the separate media measures using known relationships among them. One such alternative is data fusion-matching respondents from two or more separate media (or related) data sources using measures common to each of these sources to create a quasi-single-source database.

Fusion – From Theory through Practice

Fusion is one of a variety of different data integration techniques, and a good place to begin understanding its foundations is with the Advertising Research Foundation’s (ARF) general definition of data integration:

"A formal process to combine information from two or more separate data sources, 
making use of information in the databases for the purpose of accurately estimating 
certain values that are not available in any single data source." 
                                                                     ARF Guidelines for Data Integration

In general, data integration is the systematic process of combining information from two or more data sources (e.g. two audience currencies) to develop a new data source from which new measures can be obtained (e.g. the net reach of a multi-media print and television schedule). What differentiates fusion from the variety of different data integration techniques is that the separate data sources are combined at the respondent level. Again from the ARF Guidelines, “Fusion attempts to connect respondents [from the different databases] as if they were the same person.”

The fundamental challenge in fusing two or more databases is how best to "connect" different respondents from the separate databases such that this combination best approximates the values that would have been obtained had all the information been collected from a single set of respondents. Critical to this connection process is that the separate databases have discriminating or predictive measures in common. These common measures are used to connect respondents from the separate databases into the single fused database – the quasi-single-source. For example, if two media currency databases each contain information about their respondents’ Sex, Age, Marital Status, Education and Household Income, then respondents from these two databases can be matched on these measures. For instance, MRI and Nielsen//NetRatings are fusing databases in order to offer the industry total audience reach of magazine brands across print and online platforms.

Given the nature of survey research, whether media related or not, it is highly unlikely that two or more separate sets of respondents will have exactly the same composition on the common measures used for connection. For example, one database may have more young, well-educated, high-income females than the other ... such that it is impossible to exactly match all respondents from each data source. Consequently, fusions routinely entail the development of a hierarchy of matching variables such that it is more important for respondents to match on some measures than on others.

It is critical to emphasize the importance of a robust hierarchy of common variables for matching; the discriminating or predictive quality of these common measures with respect to the unique data being fused assures the relationships between these unique data elements are accurately embodied in the fused database. In fact, while many media fusions have been performed using only demographics as common matching variables, since demographics were the only commonly available measures, using a richer and more eloquent and predictive set of variables for matching measures reflecting the actual data to be fused yields improved fusions. For example, in a fusion involving separate television and print currencies (such as the English BARB/TGI fusion) the inclusion in each of some general measures of the media most accurately measured by the other greatly improves the quality of the fusion results beyond what is obtainable using demographics alone.

Fusion – Where We are Today

Fusion has been used fairly extensively throughout the European media and market research communities. In fact, with fusion as a known objective, many separate media currencies in Europe include as part of their designs measures of multiple media exactly to facilitate fusing currencies one with another.

In the United States, a few fusions of syndicated databases have been developed, perhaps most prominently the one by Kantar MARS/Nielsen NTI. With the increasing importance of the Internet as a media vehicle, several new initiatives are underway to fuse Internet currency measures with those of other, longer-standing media (e.g., Nielsen//NetRatings' NetView with Nielsen's NTI and MRI's national study with Nielsen//NetRatings' NetView).

Commensurate with the need to integrate currency media (and other) data sources, a number of refined fusion techniques are being developed. One such trend is towards fusions having greater levels of flexibility and subtlety with respect to the hierarchy of matching variables employed. Another trend is toward insuring that the fusion retains the exact levels of the currencies being fused.

For media research, the future of fusion as well as for other data integration strategies rests on its ability to develop robust multi-media currency-related databases supporting the increasingly complex task of developing efficient multi-media campaign schedules. 

(A similar version of this article has been published in the MRCC newsletter.)

back to top

Clients Beta Testing MediaDay Delivery Software

Fieldwork for MediaDay is complete and the delivery software is in beta testing with a group of MRI clients. 

MediaDay adds a new dimension to media analysis: time. By complementing the product usage and attitudinal data collected in MRI’s Survey of the American Consumer with time-based information, MediaDay reveals when specific targets use media, where they are when they use them, what else they are doing at the time and how engaged they are with particular media.

Marketers can use MediaDay data to craft more efficient media plans, build improved cross platform programs based on targets’ simultaneous use of media and develop communication strategies with an eye towards how targets spend their days.

MediaDay’s sample size is 8,305; it is conducted via a telephone re-contact of participants in the Survey of the American Consumer. Because fieldwork is continuous, seasonal and weekday versus weekend differences in media consumption can be explored. 

Please contact your MRI representative to learn more about MediaDay.

back to top

MRI Embarks on Outdoor Measurement

MRI is participating in the Traffic Audit Bureau’s (TAB) development of a new measurement system that, for the first time, will quantify consumer viewing of billboards and other outdoor advertising. The data will provide advertisers key insights they can use to plan and evaluate the success of outdoor campaigns.

Until now, outdoor advertising has been purchased based on traffic counts, which are used as a proxy for the opportunity to see a given ad. The new measurements will provide estimates of the number of people within demographic groups likely to see individual outdoor ads, whether they are walking or driving and how often they are likely to see them.

MRI’s sister company, GfK Custom Research North America, as well as the Telmar Group and Transearch will also be contributing to this system.

MRI will conduct two surveys for this project: 

1. A travel survey in five markets: Detailed trip information and demographic data will be gathered from 4,200 consumers in Atlanta, Chicago, Dallas, Philadelphia and San Francisco using face-to-face, computer-assisted personal interviews. 
   
2. A survey of major travel to destinations: Using a mail survey conducted among 45,000 consumers in 15 markets, MRI will gather data about frequently visited local destinations.

Both of these studies will be conducted independently from MRI’s Survey of the American Consumer. 

back to top

The Source
Mediamark Research Inc.
A member of the Gfk Group
75 Ninth Ave 5th Floor, NY, NY 10011
www.mediamark.com           www.mriplus.com

New York (212) 884-9200 • (800) 310-3305 • Chicago (312) 329-0901 • Los Angeles (323) 882-6325

Comments? Want to be removed from The Source mailing list? E-mail and let us know.

^©2007 MRI All Rights Reserved