Open Access

The Interview

PML - Personalised Media List

In 2004/2005, NRS Ltd. conducted a test of the Personalised Media List, or PML. PML was designed by Ipsos MORI as a means of reducing the number of publications shown to any one respondent in the NRS interview.

This arose from the fact that the NRS media list includes over 250 magazines, and every respondent is prompted for every magazine. However the average respondent claims to have read only 8 of these magazines in the past year, and only 4 recently enough (i.e. within the past week, month, etc.) to be counted as an average issue reader.

There was a concern that this procedure might have a detrimental impact on the interview because the majority of respondents discard the majority of the magazines they are shown (and a large amount of repetition in the interview is not considered ideal), and some respondents may feel irritated at being presented with a long list of magazines that they feel are very largely irrelevant to them.

PML was devised by Ipsos MORI as a means of reducing the title-load-per-respondent, and tested to ensure that it worked in the field, and to ascertain its effect on the readership estimates. The PML approach involves beginning the interview with a number of lifestyle questions, the answers to which are instantly processed by the laptop computer to determine which categories of magazine that particular respondent is likely to read.

The magazines in those categories are then used by the laptop to create a media list personalised to that respondent. The questions used to determine the categories of magazines cover:

• 28 topics of interest
• 12 sports topics of interest
• 8 future intentions (getting married, changing jobs, etc)
• presence and age of children in the household
• age, gender and marital status of the respondent.

In a refinement to this approach, it was recognized that the reading behaviour of an individual may not be wholly predictable. In addition to selecting those categories of magazine that the respondent is likely to read, the laptop was therefore programmed to include a sample of those categories the respondent is not likely to read, so that all the different types of reader of a publication have an opportunity of being asked about that title. With PML therefore, not all respondents are asked about all magazines. The 'missing' data (concerning those respondents not asked about a particular publication) are provided by a process of imputation.

PML was tested on half the NRS sample from April 2004 to June 2005. It worked very well in its objective of reducing the title-load-per-respondent; however, it was found that PML had little effect on readership and this, together with the fact that it required a lengthening of the data release schedule and a reduction in the precision of the data, lead to the decision not to adopt PML across the full sample.