TV-Maps: TV-Programs Positioning Maps with the SAS System

TV-Maps: TV Programs Positioning Maps with the SAS System Colombo M.*, Di Florio 0.*, Marcon G. ** R.T.I. S.p.A. (Reti Televisive Italiane) Fininvest group. ** Noustat S.r.l. (Italian SAS Quality Partner).

Abstract

Introduction

TV MAPS is a SASIAF application designed to map the competitive position of Prime Time TV programs by means of people meter data. These maps can help the evaluation of alternative program scheduling strategies. Socio-demographic concentration and total viewing data, stratified by sex, age, geographic area, education, and 16 life styles are collected day by day for each program aired on the 6 main Italian networks and for the total viewers. Data collected for other local commercial networks are aggregated and represented as a single netWork. Starting with 105 variables measured in a preselected relevant period, PROC FASTCLUS/STAT and PROC CANDISC/STAT are applied to select the most significant socio-demographic concentration variables in order to obtain the best clustering. First of all the distance matrix between cluster centroids is computed with PROC IML, then this matrix is used by PROC MDS/STAT to perform a Multi Dimensional Scaling. Clusters are represented in a two dimensional space and tatal daily viewers for each program are displayed into the same graph introducing them with the ANNOTATE FACILITY in a fashion similar to a bubble plot. Furthermore, arrows starting from the center of each bubble plot give the dominant demographic target (sex by age) for programs in terms of total viewers. Total viewers of programs belonging to the same cluster are summed over each day of the chosen period to estimate the number of potential viewers who have watched overlapping programs belonging to the same cluster. Finally Proc G3GRID/GRAPH and PROC GCONTOURIGRAPH have been .used to interpolate the estimated number of potential viewers and plot contour maps into the same plane obtained by MDS analysis. . The different background colors provide an intuitive perception of the number of potential viewers for different program types.

The italian television market if formed by two networks: RAI (Radio Televisione Italiana) with three public channels (RAIl, RAI2, RAI3) financed by a TV licence fee, and RTI (Reti Televisive Italiane) with three commercial channels (Canale5, Italial, Rete4). The two networks hold about 90 % of audience share of the market. TV programs competition is very strong especially in Prime Time (between 8:30 and 10:30 p.m.) where programs are scheduled daily and each competitor choose the most suitable programs to maximize audience and minimize program cannibalization. TV-Maps is a creativel example of product positioning system created with SAS System that gives useful insights to optimize program scheduling.

\

\'

1210

The audience data-base People meter data at aggregate· level an stratified by sex, age, geographic area, education, and 16 life styles are collected day by day for each TV program aired on the 6 main Italian networks. Data collected for other local commercial networks are aggregated and represented as a single network. The not-viewers data are collected to elaborate the map position of not-viewers and therefore of total-viewers who are the opposite of not-viewers. Socio-demographic variables, named "targets", are used to identify accurately the tv-program characteristics and peculiarity. . Concentration indexes are computed for each socio-demographic target as in following example: Ex.: .Computing of target concentration for network 1, using target=women with age between 25 and 35 concentration of network 1= (w25 _35[nl ]/aud[nl ])/(w25 _35/aud). where: w25_35[nl]=women between 25 and·35 years watching network 1 in the day of analysis.

TV-Maps: TV-Programs Positioning Maps with the SAS System

aud[nl]=total population watching network 1 in the day of analysis. w25 _35=total of women between 25 and 35 years watching TV in the day of analysis. aud=total population watching TV in the day of analysis.

Statistical methodology Starting with 105 concentration variables measured in two pre-selected relevant periods, PROC FASTCLUS/STAT and PROC CANDISC/STAT are applied to select the most significant socio-demographic concentration variables (or optionally the most significant absolute audience variables) to obtain the best clustering. The number of clusters is parametrical and its average is between 10 and 50. Then the distance matrix between cluster centroids is computed with PROC IML and this matrix is used to perform a Multi Dimensional Scaling analysis by PROC MDS/STAT to position clusters in a two dimensional space. Total viewers of programs belonging to the same cluster are summed over each day in the chosen period to estimate the number of potential viewers who have watched overlapping programs belonging to the same cluster. The same procedure is applied to each target considered in the analysis. Finally PROC G3GRID/~gAPH and PROC GCONTOURIGRAPH haveJ:",een applied to interpolate the total number·of viewers for each cluster and to plot contour maps:

:.:.:.

Fig. 1

1272

TV-Maps: TV-Programs Positioning Maps with the SAS System

Interactive control of Cluster analysis and Multidimensional Scaling

Fig. 2

Parameters: bubble plot option, with cluster dots, with cluster labels, zooming in the center, axis rotation, main targets, circle radius, arrow proportion, absolute audience or variable concentration, parametric variable to explain the quadrant meaning. It's I)Ossible to request a co~lete graphical description ofthe quadrants with all target variables, a description of each cluster profile . and of each program, ....... respectto cluster profile ... :. or relative shares.

, \

Fig. 3

1273

:' ~.~. ,..,~t..'~~Uc •• j@i>.h'htt!i!I3iLjM ; ...{ER,iE!¥#,,", .bi.'

"'Ij:'''£I',F~ __

i.

'~~O:'

[,

~~.'~;I!J~~~"t"'~""'D.'J'T'~~~."'l~"'.\-J'-~'~~'l>~:·n?"'.':!I"")