GB2489841A

GB2489841A - Analysing audiences for channels delivered to a population of households

Info

Publication number: GB2489841A
Application number: GB1209541.0A
Authority: GB
Inventors: Andrew Roberts
Original assignee: Kantar Media UK Ltd
Current assignee: Kantar Media UK Ltd
Priority date: 2012-05-29
Filing date: 2012-05-29
Publication date: 2012-10-10
Anticipated expiration: 2032-05-29
Also published as: GB2489841B; GB201209541D0

Abstract

The present invention includes a method of analyzing audiences for channels delivered to a population of households, each household having one or more presentation devices operable to present a channel for viewing by one or more members of the household, the channel being selected from a signal provided to the household by a service provider and fed to a presentation device via a signal source coupled to the presentation device. The population of households includes a first sample of households, the members of which households compile viewing information including data indicating which channels were viewed, who by, when, and the service provider of the signal from which the channel was selected and a second sample of households, each having a presentation device coupled to a channel selector device receiving a signal from one of a predetermined one or more service providers. The channel selector device is both a signal source configured to select a channel from the signal provided to the household by the service provider and feed it to the presentation device and a means to transmit channel selection data indicating which channels were selected and when via a return path data. An audience analysis unit obtains the viewing information from the first and second samples of households and allocates to each household from the second sample a household from the first sample from which is copied a portion of the viewing information from the first sample household for which the channel was selected from a signal provided by a service provider other than the predetermined one or more service providers e.g. a terrestrial aerial with no return path. This copied portion is then stored for use as data representing the viewing of channels in the household from the second sample. Embodiments of the present invention may be used to integrate STB and Peoplemeter (people meter) panels by fusing households to establish donor and recipient relations for sharing viewing records and hence to establish a large sample size for audience analysis.

Description

Method, Apparatus, and Program for Analysing Broadcast Channel Audience The present invention is in the field of analysis of audiences of broadcast channels. In particular, the invention relates to the analysis of audiences viewing a television channel at home.

Existing TV audience measurement/analysis systems make use of so-called "people meters" installed in a panel (sample) of households, chosen such that their occupants are demographically representative of the population as a whole, or at least such that their demographic characteristics are known so that weightings can be applied in extrapolating the sample to correct for demographic misrepresentation of the population. This technique is robust when analysing the audience to major channels by major demographic groupings (all men or all housewives for example). However, the widespread introduction of multi-channel program services, for reception via digital cable and/or satellite for example, and thus the availability of many channels, including minority-interest channels, has created a need for improved audience measurement/analysis system which can reliably analyse the audience even to minority-interest channels. Using a "people meter" analysis system for this would require, to provide reliable minority channel/specific demographics, a very large "people meter" panel size, which in practice would become unaffordable.

As is well known, digital program receivers such as digital satellite receivers and cable boxes (hereafter Set Top Boxes or STB5, which function as channel selection devices) can usually transmit data back to the system operator or broadcaster, typically either via the "back channel" of the cable system or via a telephone line connection. This feature is primarily provided to allow interactive TV services to be implemented, including for example pay-per-view. Such STBs can also run software applications, pre-loaded in the STBs or downloaded over the cable or over the air to the STBs.

Typically these would be games or "infomercials" . However, it has been appreciated by the inventors that it is also possible to pre-load, for example as firmware, or download software to STBs that can record, and then transmit back to the system operator, details of the channels selected from time to time by the viewer. It has further been appreciated by the inventors that since the incremental cost of doing this is very low, it can be done for a very large number of STB5 -if necessary for the whole subscriber base of households using the STBs of a system operator or service provider.

People meter type systems provide a reliable set of data for audience analysis in terms of being an accurate representation of the viewing activity of the sample of households participating in the panel. However, due to cost and inconvenience, it is only feasible to have a relatively small sample size.

Return path data from STBs representing channel selections is incomplete in terms of being an accurate reflection of the viewing activity of the sample it represents, but it has the benefit of having the potential to represent a larger sample size at a lower cost than would be feasible with a people meter system.

Embodiments of the present invention include a method of analysing audiences for channels delivered to a population of households, each household having one or more presentation devices operable to present a channel for viewing by one or more members of the household, the channel being selected from a signal provided to the household by a service provider and fed to a presentation device via a signal source coupled to the presentation device; the population of households including: a first sample of households, the members of which households compile viewing information including data indicating which channels were viewed, who by, when, and the service provider of the signal from which the channel was selected; and a second sample of households, each having a presentation device coupled to a channel selector device receiving a signal from one of a predetermined one or more service providers, the channel selector device being a signal source configured to select a channel from the signal provided to the household by the service provider and to feed the selected channel to the coupled presentation device, and to transmit channel selection data indicating which channels were selected and when as return path data. The method comprises, at an audience analysis unit: obtaining the viewing information from the first sample of households; obtaining the channel selection data from the second sample of households; and allocating to each household from the second sample a household from the first sample, and copying the portion of the viewing information from the household from the first sample for which the channel was selected from a signal provided by a service provider other than the predetermined one or more service providers, and storing said copied portion for use as data representing the viewing of channels in the household from the second sample.

Advantageously, embodiments of the present invention provide a method or mechanism for using viewing information from a first sample of households to form a more complete picture of the audiences from a second sample of households. A bigger sample with more complete viewing information is achieved. The audience statistics derived therefrom are more stable than for smaller or less complete samples, and give a better read of minority-interest channels.

In invention embodiments, viewing information for the first sample of households may be compiled using a peoplemeter scheme. Peoplemeter is used herein as exemplary of schemes for gathering viewing information indicating which broadcast channels were viewed, who by, when, and the service provider of the signal from which the channel was viewed. Viewing information such as that obtained via a peoplemeter scheme may be compiled on an individual basis by individual members of a household. The values of certain demographic characteristics of those individuals may be known, so that the number and demographic characteristics of members of a sample of households participating in the scheme watching a particular channel at a particular time (i.e. participating in the audience of a program a channel) can be ascertained.

The number and demographic characteristics so obtained can be extrapolated over a population to ascertain TVRs (television ratings). However, sample sizes tend to be small due to the cost of maintaining data regarding the demographics of households in the sample, and the inconvenience to individual members of households in the sample.

Therefore, the TVRs obtained via extrapolation of peoplemeter data tend to be quite lumpy in terms of the minimum size of audience which can be detected. This can lead to problems in detecting audience sizes for minority-interest programs or channels, or in detecting small differences in audience sizes. Figure 3 illustrates the percentage of individuals watching a particular channel throughout a day, for a particular demographic grouping. In the case of Figure 3, it represents the percentage of males aged between 16 and 34 who are watching Sky Sports News over the course of approximately a day (timeslots indicated on the horizontal axis) as a percentage of total males aged between 16 and 34 watching television. There are two estimates illustrated on the graph, the first, denoted "BARB" (dashed line), is based on audience statistics derived from peoplemeter data, the second, senoted "fusion" (solid line), is based on audience statistics derived from channel selection data from households that have been fused with peoplemeter households, according to a method embodying the present invention. Whilst the two estimates are similar in terms of the total amount of viewing that they represent, it can be seen that the "fusion" line, derived based on a method embodying the present invention, is much more stable, which is more likely to represent the true viewing activity.

Return path data providing channel selection data are easier to gather since the process is effectively automated, although it is desirable to survey the demographic characteristics. Therefore, larger sample sizes tend to be used from which to gather such channel selection data. Again, once the channel selection data have been subjected to certain processing they can be extrapolated over a population of households to obtain TVRs. However, there are inaccuracies (including determining the actual length of time for which a selected channel was being viewed, and determining which individual members of a household were viewing a presentation device presenting a channel selected by the channel selection device) associated with the processing, so that for certain purposes, peoplemeter-derived TVRs are preferred.

Furthermore, the channel selection data only represent viewing on a particular presentation device coupled to a channel selection device. It may be that the households in the second sample have further presentation devices not coupled to the channel selection device and which present channels from a signal provided by a service provider other than that which provides a signal to the channel selector device.

For example, a signal provided from a service provider via an aerial as a signal source to the presentation device from which direct tuning by the presentation device itself enables channel selection.

Advantageously, by providing a mechanism for combining the two data sources, the viewing information of the peoplemeter and the channel selection data of the RPD, embodiments of the present invention provide an audience analysis method which enables a large sample size to be achieved with a high level of accuracy, and which accounts for viewing in the second sample of households from all service providers.

The viewing information may be a set of viewing records, wherein each individual viewing record details an individual member of a household watching a channel between a start time and an end time, the individual viewing record may also indicate, or it may be derivable from the individual viewing record, the service provider for the signal from which the channel was selected. Similarly, the channel selection data may be a set of channel selection records, wherein each individual viewing record details a channel being selected on a channel selection device between a start time and an end time.

In particular, the inventors of the present invention have come to the realisation that an inaccuracy in channel selection data obtained via RPD is that, in some households providing channel selection data via RPD, some channels are presented on presentation devices using a signal source other than a channel selector device, and channels are selected from a signal provided by a service provider other than that which provides the signal to the channel selection device. For example, it may be that a terrestrial signal is used for which an aerial is the signal source, which provides a signal for decoding at the presentation device from which a channel is selected by the presentation device itself, and not by a channel selector device. Household members viewing channels in this way would contribute to an audience of a channel, however the channel selection would not be recorded in the channel selection records and hence TVRs extrapolated from such records would be missing a portion of the audience. Effectively, data representing viewing of channels via signal sources other than SIBs from which RPD is gathered in the second sample of households is "missing". Therefore, TVRs or other audience statistics derived from such channel selection records suffer from an inaccuracy.

A service provider may be, for example, a legal entity responsible for distributing television signals to households, or a particular signal distribution network. For example, a service provider may provide a paid subscription service over a cable or satellite network, or a service provider may transmit free-to-air signals in a radio signal.

Embodiments of the present invention provide a method or mechanism by which the "missing" viewing can be simulated and the simulated viewing attributed to the household from the second sample for audience analysis purposes. The copying of a portion of a viewing information from a household of the first sample, for storage and hence processing as data representing the channels viewed in the household of the second sample from which channel selection data are obtained, enables the "missing" viewing to be simulated. Effectively, viewing information is donated from a household in the first sample.

Embodiments may be implemented so that the presentation device is a television and/or the channel is a television channel and/or the channel selector device is a set top box and/or the predetermined one or more service providers are one or more paid subscription service providers. Furthermore, members of households who were recorded as viewing a channel may be identified by a set of values of demographic characteristics.

Optionally, the households from the second sample are allocated to households from the first sample by means of one or more data hooks, data hooks being characteristics of the households which can be used as a basis for comparison between the household of the second sample and each of the households of the first sample.

Such a method, wherein the allocating and copying data on the basis of data hooks can also be referred to as fusing data, and the data hooks as fusion hooks, provides an exemplary technique for matching households from the second sample (as recipients), to households of the first sample (as donors). Performing the allocation on the basis of identified characteristics already known and stored regarding the households (such as the demographic characteristics of the household or its members) or the data (channel selection information or viewing information) obtained from the households provides a technique for improving the accuracy of the copied information in terms of the viewing it is supposed to represent. Wherein the accuracy of the copied data is the similarity of the data copied and attributed to a household from the second sample to the viewing which the copied data is intended to represent, namely the viewing of channels which involves direct channel selection via the presentation device itself or viewing of another presentation device not coupled to the channel selector device, or in other words, for which the channel was selected from a signal provided by a service provider other than the predetermined one or more service providers.

In addition, the household from the first sample may be selected from a subset of households in the first sample which have a presentation device coupled to a channel selector device receiving a signal provided by one of the predetermined one or more service providers.

Optionally, a comparison metric is used to determine the household from the first sample that is the most similar to the household of the second sample based on the data hooks, and the most similar household from the first sample is allocated to the household from the second sample.

In more sophisticated implementations of invention embodiments, it may be that certain of the data hooks are deemed to be more important in matching donor households (from the first sample) to recipient households (from the second sample) in an accurate fashion, so a comparison metric may be derived to enable a value to be attributed to a comparison between households. For example, it may be that particular data hooks are weighted so that a difference in that characteristic has a lesser or greater bearing on the overall assessment of similarity. It will be appreciated that the precise design of such a metric very much depends on the implementation of the invention embodiment, and may be designed at a centralised or global level for the system. For example, the allocating may be performed by a fusing algorithm, and the one or more data hooks may be fusion hooks.

The one or more data hooks may include one or more of the following characteristics: -demographic data representing members of the households; -data representing the presentation devices used in the households; -data representing the propensity of the households to select particular channels; and/or -data representing the propensity of the households to select particular channels from particular service providers.

The inventors of the present invention have had the insight that viewing habits and behaviours can be correlated to one or more of the above characteristics. Therefore, in improving the accuracy of the process of matching donor households to recipient households, it is desirable that one or more of the above be taken into consideration.

Whilst embodiments of the present invention are concerned with incompleteness of viewing selection data from households in the second sample in terms of representing the complete viewing activity of the household, rather just on a single presentation device of channels from a single service provider, the inventors have had the insight that combination of techniques for overcoming said inaccuracy with procedures for overcoming other sources of inaccuracy may yield particularly detailed audience analysis. For example, one issue previously identified by the inventors is that the STB data (exemplary of channel selection data) does not reveal when the TV (exemplary of a presentation device) to which the SIB (exemplary of a channel selector device) is coupled was switched on and off. SIBs are very typically left on permanently, so at the end of a viewing session the IV may be switched off but the SIB continues to be tuned to the last channel watched. This means that channel selection data from SIBs provides, or almost certainly provides, a false indication of actual viewing of the TV.

The integrity or value of the 5Th viewing data, even as household viewing information, is impaired.

Optionally, the method further comprises subjecting the channel selection data from the second sample of households to a capping procedure, whereby the indicated length of time for which a channel remains selected is reduced to a capped maximum permitted length of time if the indicated time exceeds the maximum permitted length, so that the channel selection data include capped data In accordance with a method embodying the present invention, the "raw" STB data is effectively edited in such a way that very long STB sessions are capped after a given interval. In other words, the method of the present invention provides that the TV will be assumed to have been turned off if the STB has been tuned to the same channel uninterrupted for more than the capped duration limit. Surprisingly, this procedure provides an effective way of dealing, with the problem outlined above, removing a substantial part of the distortions caused by the problem and thus providing refined data which allows audience analysis to be more accurately carried out on the basis of STB data, rendering the product of the analysis more useful and commercially valuable.

Another issue previously identified by the inventors of the present invention is that channel selection data does not reveal which individual members of a household were watching at the relevant times.

Optionally, the method further comprises employing the capped data to select, for each of the indications of the channel selected and time at which the channel is selected within the capped data, an array of probabilities of individuals' exposure (PIVs) to the presentation of the channel on a presentation device, for individuals of different demographies, whereby for each member of the household from which the capped data originates a probability can be selected from the array of probabilities and a monte carlo simulation employed to predict whether or not that member viewed the presentation device at the indicated time, whereby the capped data can be resolved into an estimation of whether an individual member of the household, and an estimated viewing information obtained predicting which channels were viewed, who by, when, and the service provider of the signal from which the channel was selected.

Channel selection data from a channel selector device, even when capped, fundamentally provides an indication of what was being presented by a presentation device. In accordance with embodiments of a further aspect of the present invention, however, it is provided that information relating to individuals' viewing can be derived from the channel selection data, in accordance with stored data relating to the household, specifically information representing who the members of the household are (in terms of demographic characteristics, and the number of members). Channel selection data has preferably been subjected to capping in accordance with the present invention as indicated above. It has been realised that the viewing of individual members of a household can be attributed by employing Monte Carlo sampling techniques using a "probability of an individual's viewing" (Ply). In this context, a probability of an individual's viewing (PIV) is a probability based on for example, daypart, channel and the demographic characterisation of the individual, derived on the basis of reference information obtained from a random selection of households, and thus individuals in the households, of known demographic character. This reference information is non-STB-based, or may be based in part in STB data enhanced with additional information, obtained for example using individual viewing tags, which enables PIVs as reference data to be determined per daypart, channel and individual's demography. Thus, this probability can be derived from knowledge of individuals' viewing behaviours analysed by daypart or other time interval (e.g. for every quarter hour), channel (s) viewed, and the demographic characterisation of each individual concerned, for a representative sample of all households (preferably of households having channel selector devices belonging to the service providers' system). This acquired knowledge is available to embodiments of the present invention, for example in the form of a PIV matrix dimensioned by demographic group, time of day, day of week, and channel (wherein the channels may be grouped). Given this reference information, channel selection data subjected to capping in accordance with the present invention as indicated above, can be resolved into estimates of viewing data for individuals or different demographic classes, effectively on the basis that the sub-set represented by the representative sample of households, and hence the individuals in making up those households, is an indicator of individuals' behaviour in the set of all households providing the channel selection data. Effectively, the PIV obtained from the PIV matrix in accordance with characteristics of the capped channel selection data and known information about members of the household, is subjected to a Monte Carlo process to determine, for each member of the household, whether or not they should be included as an audience member for the times and channels indicated by the relevant part of the channel selection data (which relevant part may be an individual channel selection record). A PIV may also be considered to be a conditional probability of a member of a household viewing, assuming that the presentation device was switched on and presenting a channel, and based on the above-discussed characteristics of both the viewing record (the individual viewing event -channel and on/off time) and the demographics of the individual.

Optionally, the method may further comprise obtaining estimated viewing information for each household in the second sample; and based on a comparison of values of characteristics attributed to the first and second samples with values of the same characteristics attributed to the population, performing RIM weighting to obtain RIM weighted estimated viewing information for a combination of the first and second samples of households; and producing audience statistics for the population based on the RIM weighted estimated viewing information for the combination of the first and second samples of households.

Advantageously, applying RIM weightings (or iterative proportional fitting) to a combination of the samples of households enables audience statistics to be generated which are adjusted to correct for characteristics which are not represented proportionately in the sample (for example, on a per capita basis). RIM weighting processing can be applied to characteristics including equipment owned, and subscription to pay-TV packages, and/or other characteristics such as demographic characteristics. Effectively, RIM weighting corrects imbalances in measurable characteristics of a sample with respect to the known characteristic of the population the sample is intended to represent (for example, proportion of people aged under-5).

The RIM-weighted estimated viewing information can then be considered to fairly represent the population in terms of the characteristics for which RIM weighting was applied. The RIM-weighted combination of the estimated viewing information from households in the second sample and viewing information from households in the first sample can then be further processed, for example, extrapolated, to obtain audience statistics for the entire population. The audience statistics may include, for example, TVRs and optionally TVRs broken down by demographic grouping.

In an embodiment of the present invention, a household from the second sample may be allocated to a household from the first sample having known demographic characteristics, and those known demographic characteristics are copied and stored for use as data representing the household of the second sample; wherein the households from the second sample are allocated to households from the first sample by means of one or more data hooks, data hooks being characteristics of the households which can be used as a basis for comparison between the household of the second sample and each of the households of the first sample, and said data hooks may include data indicating presentation equipment in the households, and propensity to select particular channels from particular service providers.

This particular embodiment provides for an anonymous second sample of households, or for a second sample of households about which little or no demographic information is known. Advantageously, significant cost savings can be made by obviating the need to gather demographic information about the second sample of households. This enables a larger sample size to be used, which results in an increase in accuracy in certain aspects of audience statistics based on the sample.

in embodiments of another aspect, there is provided an audience analysis unit for analysing audiences for channels delivered to a population of households, each household having one or more presentation devices operable to present a channel for viewing by one or more members of the household, the channel being selected from a signal provided to the household by a service provider and fed to a presentation device via a signal source coupled to the presentation device; the population of households including: a first sample of households, the members of which households compile viewing information including data indicating which channels were viewed, who by, when, and the service provider of the signal from which the channel was selected; and a second sample of households, each having a presentation device coupled to a channel selector device receiving a signal from one of a predetermined one or more service providers, the channel selector device being a signal source configured to select a channel from the signal provided to the household by the service provider and to feed the selected channel to the coupled presentation device, and to transmit channel selection data indicating which channels were selected and when as return path data. The audience analysis unit comprises: a viewing information acquisition module operable to obtain and store the viewing information from the first sample of households; a channel selection data acquisition module operable to obtain and store the channel selection data from the second sample of households; and a viewing record donor module operable to allocate to each household from the second sample a household from the first sample, to copy the portion of the viewing information from the household from the first sample for which the channel was selected from a signal provided by a service provider other than the predetermined one or more service providers, and to store said copied portion for use as data representing the viewing of channels in the household from the second sample. The audience analysis unit may also be configured to carry out the method steps of invenUon embodiments described elsewhere in this document as occurring at an audience analysis unit.

ln embodiments of a further aspect of the present invention, there is provided an audience analysis system for analysing audiences for channels delivered to a population of households, each household having one or more presentation devices operable to present a channel for viewing by one or more members of the household, the channel being selected from a signal provided to the household by a service provider and fed to a presentation device via a signal source coupled to the presentation device, the population including a first sample of households, the members of which households compile viewing information including data indicating which channels were viewed, who by, when, and the service provider of the signal from which the channel was selected. The audience analysis system includes: a plurality of channel selector devices situated within households forming a second sample of households from the population of households, each having a presentation device coupled to a channel selector device receiving a signal from one of a predetermined one or more service providers, the channel selector device being a signal source configured to select a channel from the signal provided to the household by the service provider and to feed the selected channel to the coupled presentation device, and to transmit channel selection data indicating which channels were selected and when as return path data; and an audience analysis unit as described above.

A computer program may also be provided, which, when executed by a computer, causes the computer to carry out a method according to embodiments of the present invention. In particular, a computer program may be provided which, when executed by a computer, causes the computer to function as an audience analysis unit embodying the present invention.

Embodiments of the present invention will now be described, with reference to the accompanying drawings, in which: Figure 1 illustrates an implementation of an embodiment of the present invention; Figure 2 illustrates an audience analysis unit embodying the present invention; Figure 3 illustrates the data stability resulting from embodiments of the present invention; Figure 4 illustrates the problem of "missing" data from channel selection records provided via RPD; Figure 5 illustrates the effect of simulating the missing data using an embodiment of the present invention; Figure 6 illustrates the dependence of viewing activity on equipment owned; Figure 7 illustrates the dependence of viewing activity on subscription package.

Figure 1 schematically illustrates a system in which multi-channel program services, for reception via digital cable and/or satellite and/or aerial for example, are delivered to households 10 & 20 which provide the audience which is to be analysed. The programs or channels available, in this case TV channels, are provided by one of two service providers or system operators each having a delivery network.

Household 10 has an aerial ha which receives a multi-channel program service from delivery network a, and provides a signal to the presentation device I 2a. The signal in this example is a multi-channel signal, and the presentation device itself is operable to select a channel to present. No separate STB is utilised between the delivery network and the presentation device I 2a. There is no return path from signal source 11 a to the delivery network a or to any centralised data gathering or analysis facility.

Household 10 also has a channel selector device 11 b, for example a so-called Set Top Box or STB which acts as a signal source for a coupled presentation device 12b. The channel selector device 11 b is operable to receive a multi-channel program signal from a delivery network b including digital cable and/or satellite, for example, and is distinct from signal source 11 a by having a return path by which it is configured to return data to the delivery network b, or to some sort of centralised data gathering or analysis facility (either via the delivery network or service provider, or directly). The channel selector device hIb provides a signal (which may include only a single, selected, channel) to the coupled presentation device 12b for presentation to individual members of the household 10.

The channel selector device 11 b in household 10, although having a return path, is not configured to return channel selection data as return path data (although it may be configured to return other data as return path data, for example, to utilise interactive services). In fact, household 10 is a member of a first sample of households which are members of a panel whose individual members provide viewing information (which may be a set of viewing records) to an audience analysis unit 30. The viewing information details which members of the household were watching which channels, when (in terms of both time and duration), and the configuration of the presentation device. The configuration of the presentation device may be represented by data indicating whether the presentation device is coupled to a signal source operable to provide RPD, or whether it is coupled to a signal source which is not configured to provide RPD, or the configuration of the presentation device may be represented by data indicating the delivery network providing the signal to from which the presented channel was selected. The manner in which such records can be compiled is known to practitioners in this field of technology, and may be based on, for example, a combination of prior knowledge about which presentation devices are in which room in a building attributed to the household, the configuration of those devices, the demographics of the individual members of the household, and live data recorded by apparatus present in the household indicating which presentation devices were on and when, and which members of the household were present in which rooms and when.

Household 20 has a signal source 21a and coupled presentation device 22a which are configured in the same way as the signal source ha, and presentation device 12a in household 10. That is to say, the presentation device 22a is coupled to a signal source 21a which is not operable to provide return path data.

Household 20 also has a signal source 21 b and coupled presentation device 22b which are configured in the same way as the signal source lib and presentation device 12b in household 10. However, the signal source 21b in household 20, which is a channel selector device, is configured to provide channel selection data to an audience analysis unit 30, either via a direct data connection to the audience analysis unit 30, or via one or more intermediate destinations, such as a service provider of the delivery network.

The channel selection data (which may be a set of individual channel selection records) indicate which channels were selected on the channel selection device 21 b and when.

The channel selection data may be stored locally on the channel selection device 21 b and transmitted intermittently to the audience analysis unit 30. Alternatively, the return path data indicating channel selecUon my be provided to the audience analysis unit continuously while the channel selection device 21 b is in operation.

The schematic illustration of Figure 1 shows a single household in the first sample, household 10, and a single household in the second sample, household 20. However, in embodiments of the present invention the sample sizes may be much greater, and there may be a plurality of houses in each sample, along with a further plurality of houses belonging to the population for which audience statistics are being established but which do not belong to either sample. For example, there may of the order of a thousand households in the first sample, of the order of ten thousand households in the second sample, and of the order of between one million and one hundred million households in the populaflon.

The audience analysis unit 30 includes a storage unit 31 for storing information and data both before, during, and after processing, and a processor 32 for carrying out data processing operations on the gathered data in order to generate audience statistics such as TVRs. Hence, the audience analysis unit 30 may be a computing device or a virtual computing resource, having stored thereon or having access to a selection of algorithms and processing instructions for generating audience statistics based on the gathered records. Along with other procedures which may be carried out by the audience analysis unit 30, the audience analysis unit 30 is at least operable to perform a function of allocating to each household from the second sample a household from the first sample, and copying the portion of the viewing information from the household from the first sample for which the channel was selected from a signal provided by a service provider other than the predetermined one or more service providers, and storing said copied portion for use as data representing the viewing of channels in the household from the second sample.

Figure 2 is a schematic illustration of an audience analysis unit 30. Figure 2 illustrates the audience analysis unit in terms of its functional modules. The hardware configuration required to realise such functionality may be, for example, a computing device having a network interface card, a processor, a memory, and a storage unit.

For example, it may that the functional modules of Figure 2 are each programs, routines, algorithms, or sub-routines, stored on a computing device as a series of instructions which can be executed by a processor in order to perform their respective functions. In that sense, a computing device executing a computer program according to embodiments of the present invention is caused to function as an audience analysis unit embodying the present invention, such as that illustrated in Figure 2.

The channel selection data acquisition module 301 is configured to acquire the channel selection data compiled at, or compiled on the basis of RPD from, STBs in households from a sample of households allocated (for example, by selection and registration) by a service provider to provide channel selection data. The channel selection data include data indicating which channels were selected and when at the STB from which they are acquired. The channel selection data, or data enabling channel selection data to be compiled, are transmitted from the participating STBs as return path data. The return path data may be received directly by the channel selection record acquisition module 301, or may be transmitted via the service provider. Some mapping may be required in order to allocate channel selection data to channels for analysis, for example, channel selection data from the STBs may indicate that the SD and HD versions of a particular channel are different channels, whereas for audience analysis purposes it may be required to combine the two as a single channel (or vice versa). In certain embodiments, the channel selection record acquisition module 301 may be configured to compile data indicating which channels were selected by an STB and when into a channel selection record. The channel selection records may also include data indicating an identity of the STB from which they were obtained, or the service provider of the signal received by the STB.

The viewing information acquisition module 302 is configured to acquire the viewing information compiled at, or compiled on the basis of data received from, households in a sample of households participating in a viewing information reporting scheme, for example, a people meter scheme. The viewing information acquired in this way may be considered to be "reported viewing records". The reported viewing information (which may comprise a plurality of individual viewing records) include data indicating which channels were viewed, who by (in terms of identity of the viewer or demographic characteristics of the viewer), when, the service provider and/or the signal source of the presentation device on which the channel was viewed (via a satellite signal received by an STB coupled to a presentation device, via a cable signal received by an STB coupled to a presentation device, or via an aerial). The reported viewing information may have been acquired via a third party responsible for compiling viewing information, or directly from the households participating in the viewing recordal scheme.

The channel selection data acquisition module 301 is configured to provide data in the form of channel selection data to the channel selection data capping module 303. The channel selection data capping module 303 is configured to apply a capping algorithm to the channel selection data to correct for inaccuracies such as those caused by STBs being operational and supplying a selected channel to a TV in spite of the TV being turned off or otherwise not being set to display the channel. That is to say, an STB could be supplying a data signal to a coupled TV, but the TV is set not to display the channel. Channel selection data from STBs supplying selected channels to TVs which are not being displayed can be described by certain characteristics, and hence they can be removed from channel selection data for analysis, the channel selection data from that STB can be capped' in the sample (for example, by capping individual channel selection records). Capping algorithms are now established for this purpose.

More detail on capping algorithms can be found in, for example, international patent publication W02006/037631. The channel selection data capping module 303 is configured to transform channel selection data into capped channel selection data by applying a capping algorithm.

The channel selection data capping module 303 is configured to provide data in the form of capped channel selection data to the viewing likelihood determination module 304. The viewer determination module 304 is configured to apply a probabilistic algorithm to the capped channel selection records to determine the likelihood that individuals of different demographies were watching a channel that was being presented, and then to apply monte carlo analysis to turn that probability into a yes or no (or a I or 0) for each member of the household from which the data was provided.

It should be noted that the identity of household members, in terms of number of members and their demographic characteristics, may be known from survey or other means, or may be derived by data fusing). Effectively, a probabilistic (PIV) algorithm is applied to the capped channel selection data to determine a conditional probability of a member of a household viewing, assuming that the presentation device was switched on and presenting a channel, and based on the above-discussed characteristics of both the capped selection data (the individual viewing event -channel and on/off time) and the demographics of the individual. That is to say, the capped channel selection record indicates the SIB channel selected at each point in time, but not who was watching it. Wherein the "who" may be characterised by demographic parameters such as any of age, gender, household income, marital status, employment status etc. To this end, the viewing likelihood determination module may be configured to refer to stored data representing the demographic characteristics of the household from which the relevant channel selection records were obtained, in the form of a PIV matrix dimensioned by demographic group, time of day, day of week, and channel (wherein the channels may be grouped). The capped channel selection data may be employed to select, in dependence upon one or more factors such as the channel selected on the STB and the time at which the channel is selected, an array of probabilities of individuals' exposure (P1Vs) to the presentation of the channel on the presentation unit, for individuals of different demographies from the matrix. The capped data can then be resolved into data estimating whether or not the individual members of the household watched the channel at the times indicated by the portion of channel selection data being considered, such as an individual channel selection record, based on the demographic characteristics attributed to household members. More detail on the probabilistic algorithm can be found in, for example, international patent publication W02006/037631. Hence, the viewer determination module is configured to transform the capped channel selection data into an incomplete estimated viewing information.

Which is incomplete insofar as only the viewing of a particular presentation device of channels provided by a particular service provider is represented.

It may be that, in households selected to provide channel selection data, all STBs having the capability to provide channel selection data are configured to do so, and the channel selection data from each STB in a single household can be combined at the audience analysis unit. On the other hand, it may be that households selected to provide channel selection data via RPD only provide channel selection data relating to a single presentation device when receiving a signal from a particular STB. However, it is possible that a household having one or more STBs configured to provide channel selection data to an audience analysis unit also has one or more presentation devices configured to receive a signal from a signal source other than the STB(s). The channel selection data would carry no indication of channel selections or viewing of such presentation devices or their STBs. Therefore, the estimated viewing information generated by the viewer determinaUon module 304 may be considered to be incomplete estimated viewing information. The incomplete estimated viewing information is transmitted from the viewer determination module 304 to the viewing information donor module 305. The viewing information donor module 305 is configured to complete the incomplete estimated viewing record by simulating the missing viewing information. For example, the viewing information donor module 305 may use a data fusing algorithm to "fuse" the household 20 from which the channel selection record being processed was obtained, to a household 10 participating in a viewing recordal scheme. Both the household 20 for which the missing viewing information is being simulated, and the donor household to which it has been matched by fusing, may have a presentation device coupled to an SIB receiving a signal from a particular service provider. In the recipient' household 20, it is the SIB receiving a signal from the particular service provider that transmits channel selection data via RPD. Ihus, it is the portion of the viewing information from the donor household relating to viewing of channels from signals provided by other service providers, for example, signals received via a television aerial, that is copied and attributed to the recipient household 20, thus completing the incomplete estimated viewing information and resulting in complete estimated viewing information for the household in question.

The process by which the household 20 (from which the channel selection data being processed was obtained) is matched to a donor household participating an the viewing recordal scheme is known as fusing.

Ihe fusing algorithm uses fusion hooks (or data hooks) which are variables, characteristics, or parameters of the households that can be measured or values obtained for the household 20 and for each household 10 participating in the viewing recordal scheme (potential donor households). Based on comparisons between the measurements or values of the fusion hooks between the household 20 and each potential donor household, a match can be established, for example, by choosing the potential donor household having values of the fusion hooks that most closely match those of the household 20. In exemplary embodiments, the fusion hooks are the presentation devices owned by the households (which may be characterised according to size, manufacturer, model number, or value), demographic data representing the members of the households, and the propensity to watch particular channels (for example, free-to-air channels) via an SIB.

Once the household 20 has been matched to a donor household in this way, the portion of the viewing information relating to the viewing of channels from signals from service providers other than that providing the signal to the SIB of the donor household is replicated and added to the incomplete estimated viewing information already obtained for the household 20. Ihus, complete estimated viewing information is obtained for the household 20. Once this process has been performed for each of the households selected to be part of the sample (the second sample of households) from which channel selection data are obtained, the viewing data for that sample of households is complete.

The "missing" viewing which is simulated by the viewing information donor module 305 is illustrated in Figures 4 & 5. In Figures 4 & 5, the numbers represent an average TVR over the course of a day for a channel. In Figure 4, the vertical axis represents the TVR extrapolated from a sample of households providing RPD representing channel selected by a channel selector device receiving a signal from a particular service provider, whereas the horizontal axis represents the TVR extrapolated from a sample of households participating in a peoplemeter scheme recording all viewing of channels in the household, regardless of service provider. The dashed diagonal line represents equal TVRs from both samples. However, it can be seen that the TVRs derived from the data provided by the sample of households providing RPD only representing channels selected from a signal from a particular service provider tend to be lower than those derived from the households in the peoplemeter scheme -as illustrated by the solid line of best fit. In Figure 5, however, in accordance with a method embodying the present invention, a portion of the viewing information from the peoplemeter sample has been attributed to the sample providing RPD representing channels selected from a signal from a particular service provider. In this case, the line of best fit for the derived TVRs coincides with the line representing equal TVRs from each sample, hence Figures 4 and 5 illustrate that methods according to the present invention are effective in simulating the "missing" viewing data.

The complete estimated viewing information is transmitted by the viewing information donor module 305 to the population statistics generation module 306 for each house belonging to the second sample of households. Television rating (TVR) figures are statistical estimates of the number of people watching (the number of audience members) a particular channel in a particular timeslot, or for a predetermined portion of that timeslot. TVRs are normally attributed to a particular program being broadcast on channel over a particular period of time. TVRs may also be used by advertisers to assess the desirability of broadcasting a piece of commercial programming (an advertisement) at a particular time. When the demography of a sample of potential viewers is known (for example, the members of households participating a peoplemeter monitoring scheme, or members of households registered to provide RPIJ such as the second sample of the present example), it is possible to apply RIM weighting to the sample of data in order that it can represent a population more effectively. A combination of the viewing information from households in the first sample and complete estimated viewing information from households in the second sample is demographically weighted using RIM weighting in order to correct for skewed characteristics in the sample, that is, to correct for ways in which it is known that the combined sample is disproportionate with respect to the population as a whole. An estimate of the number of people from a population that were watching a particular channel during a particular timeslot, or that participated in the audience of a particular program, can then be obtained via extrapolation.

For example, prior to extrapolating RPD from a sample over a population, RIMS may be applied for equipment type (manufacturer, display size, display resoultion, value etc) of the presentation units coupled to the STBs, and in the case of STBs with pay tv functionality, RIMS may be applied for subscription package, in addition to demographic characteristics such as age, gender, and/or marital status. For example, Figure 6 represents the amount of viewing in a household of the top 25 channels available free-to-air (left hand bar of each pair) relative to the amount of viewing in a household of the top 25 channels available via a subscription package (right hand side of each bar). The same metric is used in each case, which represents an amount of viewing, and hence the sizes of the bars can be compared. The pair of bars on the left hand side represent the relative viewing of top 25 free-to-air v top 25 subscription channels in all homes in a sample. The next pair represents the relative viewing in homes having an HD (high definition) 5Th for viewing channels. The next pair represents the relative viewing in homes having an STB with a recording function, but standard definition, and the next pair represents relative viewing on a standard STB without HD or recording functions. The final bar represents viewing by the same metric in households which do not have access to the subscription package. It can be seen from Figure 6, that viewing habits are dependent upon the equipment via which channels are presented. Hence it is appropriate in embodiments of the present invention to apply RIM weighting with respect to equipment owned. Figure 7 uses the same metric, again representing the amount of viewing in a household of the top 25 channels available free-to-air (left hand bar of each pair) relative to the amount of viewing in a household of the top 25 channels available via a subscription package (right hand side of each bar). Again, the left hand pair of bars are averaged over all households in the sample, and the right hand bar is for households not having a subscription package. The remaining pairs of bars each represent relative amounts of viewing of free-to-air channels and subscription channels for a particular subscription package. it can be seen from Figure 7 that viewing habits are dependent upon the subscription package, if any, to which a household subscribes for provision of a signal containing television channels. Hence it is appropriate in embodiments of the present invention to apply RIM weighting with respect to the package to which a household subscribes in order to better represent the population of households.

The population statistics generation module 306 is configured to apply a RIM weighting procedure and extrapolation to the estimated complete viewing information obtained from each household of the second sample combined with the viewing information obtained from households in the first sample to obtain audience statistics for the whole population from which the samples were taken. Statistics include, for example, TVRs indicating how many members of a population watched a particular channel between particular times, both as a total viewer figure, and broken down by particular demographic characteristics.

Claims

CLAIMS1. A method of analysing audiences for channels delivered to a population of households1 each household having one or more presentation devices operable to present a channel for viewing by one or more members of the household, the channel being selected from a signal provided to the household by a service provider and fed to a presentation device via a signal source coupled to the presentation device; the population of households including: a first sample of households, the members of which households compile viewing information including data indicating which channels were viewed, who by, when, and the service provider of the signal from which the channel was selected; and a second sample of households, each having a presentation device coupled to a channel selector device receiving a signal from one of a predetermined one or more service providers, the channel selector device being a signal source configured to select a channel from the signal provided to the household by the service provider and to feed the selected channel to the coupled presentation device, and to transmit channel selection data indicating which channels were selected and when as return path data; the method comprising: at an audience analysis unit: obtaining the viewing information from the first sample of households; obtaining the channel selection data from the second sample of households; and allocating to each household from the second sample a household from the first sample, and copying the portion of the viewing information from the household from the first sample for which the channel was selected from a signal provided by a service provider other than the predetermined one or more service providers, and storing said copied portion for use as data representing the viewing of channels in the household from the second sample.
2. A method according to claim 1, wherein the households from the second sample are allocated to households from the first sample by means of one or more data hooks, data hooks being characteristics of the households which can be used as a basis for comparison between the household of the second sample and each of the households of the first sample.
3. A method according to claim 2, wherein a comparison metric is used to determine the household from the first sample that is the most similar to the household of the second sample based on the data hooks, and the most similar household from the first sample is allocated to the household from the second sample.
4. A method according to claim 2 or 3, wherein the allocating is performed by a fusing algorithm, and the one or more data hooks are fusion hooks.
5. A method according to any of claims 2 to 4, wherein the one or more data hooks include one or more of the following characteristics: -demographic data representing members of the households; -data representing the presentation devices used in the households; -data representing the propensity of the households to select particular channels; and/or -data representing the propensity of the households to select particular channels from particular service providers.
6. A method according to any of the preceding claims, wherein the method further comprises: subjecting the channel selection data from the second sample of households to a capping procedure, whereby the indicated length of time for which a channel remains selected is reduced to a capped maximum permitted length of time if the indicated time exceeds the maximum permitted length, so that the channel selection data include capped data.
7. A method according to claim 6, wherein the method further comprises: employing the capped data to select, for each of the indications of the channel selected and time at which the channel is selected within the capped data, an array of probabilities of individuals' exposure (P1V5) to the presentation of the channel on a presentation device, for individuals of different demographies, whereby for each member of the household from which the capped data originates a probability can be selected from the array of probabilities and a monte carlo simulation employed to predict whether or not that member viewed the presentation device at the indicated time, whereby the capped data can be resolved into an estimation of whether an individual member of the household, and an estimated viewing information obtained predicting which channels were viewed, who by, when, and the service provider of the signal from which the channel was selected.
8. A method according to claim 7, wherein the method further comprises: obtaining estimated viewing information for each household in the second sample; and based on a comparison of values of characteristics attributed to the first and second samples with values of the same characteristics attributed to the population, performing RIM weighting to obtain RIM weighted estimated viewing information for a combination of the first and second samples of households; and producing audience statistics for the population based on the RIM weighted estimated viewing information for the combination of the first and second samples of households.
9. A method according to any of the preceding claims, wherein members of households who were recorded as viewing a channel are identified by a set of values of demographic characteristics.
10. A method according to any of the preceding claims, wherein the presentation device is a television and/or the channel is a television channel and/or the channel selector device is a set top box and/or the predetermined one or more service providers are one or more paid subscription service providers.
11. A method according to any of the preceding claims, wherein the household from the second sample is allocated to a household from the first sample having known demographic characteristics, and those known demographic characteristics are copied and stored for use as data representing the household of the second sample; wherein the households from the second sample are allocated to households from the first sample by means of one or more data hooks, data hooks being characteristics of the households which can be used as a basis for comparison between the household of the second sample and each of the households of the first sample, and said data hooks may include data indicating presentation equipment in the households, and propensity to select particular channels from particular service providers.
12. An audience analysis unit for analysing audiences for channels delivered to a population of households, each household having one or more presentation devices operable to present a channel for viewing by one or more members of the household, the channel being selected from a signal provided to the household by a service provider and fed to a presentation device via a signal source coupled to the presentation device; the population of households including: a first sample of households, the members of which households compile viewing information including data indicating which channels were viewed, who by, when, and the service provider of the signal from which the channel was selected; and a second sample of households, each having a presentation device coupled to a channel selector device receiving a signal from one of a predetermined one or more service providers, the channel selector device being a signal source configured to select a channel from the signal provided to the household by the service provider and to feed the selected channel to the coupled presentation device, and to transmit channel selection data indicating which channels were selected and when as return path data; the audience analysis unit comprising: a viewing informaUon acquisition module operable to obtain and store the viewing information from the first sample of households; a channel selection data acquisition module operable to obtain and store the channel selection data from the second sample of households; and a viewing record donor module operable to allocate to each household from the second sample a household from the first sample, to copy the portion of the viewing information from the household from the first sample for which the channel was selected from a signal provided by a service provider other than the predetermined one or more service providers, and to store said copied portion for use as data representing the viewing of channels in the household from the second sample.
13. An audience analysis system for analysing audiences for channels delivered to a population of households, each household having one or more presentation devices operable to present a channel for viewing by one or more members of the household, the channel being selected from a signal provided to the household by a service provider and fed to a presentation device via a signal source coupled to the presentation device, the population including a first sample of households, the members of which households compile viewing information including data indicating which channels were viewed, who by, when, and the service provider of the signal from which the channel was selected; the system including: a plurality of channel selector devices situated within households forming a second sample of households from the population of households, each having a presentation device coupled to a channel selector device receiving a signal from one of a predetermined one or more service providers, the channel selector device being a signal source configured to select a channel from the signal provided to the household by the service provider and to feed the selected channel to the coupled presentation device, and to transmit channel selection data indicating which channels were selected and when as return path data; and an audience analysis unit according to claim 12.
14. A computer program which, when executed by a computer, causes the computer to carry out a method according to any of claims 1 to 11.
15. A computer program which, when executed by a computer, causes the computer to function as an audience analysis unit according to claim 12.