The television industry is growing at an exponential rate. The television stations try to gain more and more viewers in order to increase their market share. In this stage comes the need for a system that helps them forecast their viewership as well as plan their program. In this paper we are proposing a system for forecasting viewership. The modeling approach simulates the viewer’s decision process, which is composed by two different components. The first component is the total viewership, which employs time series, and the second component is the Competitive Environment, which employ the market response models
The exploratory analysis of the time series data with graphical representations, qualitative pattern matching, statistical descriptives and other methods was done in order to identify seasonality, trends, outliers and patterns of the time series. During the data analysis process the historical data were also included for related descriptive statistics. Such exploratory analysis helped in the choice of the class of quantitative model that was used.
In the proposed forecasting system, many of the factors that affect viewership and the competitive environment and were identified during the analysis process were incorporated even indirectly. Up to now the forecasts for the television viewership was done manually depending entirely on the experience of the person that made the forecasts. To produce these results was very long and based on individual subjective assessments. The results were not accurate since there was no quest for supporting evidence and the person that did the judgmental forecasting had all the types of bias and limitations that usually appear in the judgmental forecasting.
The design of the system was based on a general schema that was created. The schema simulates the decision process of the individual to watch a television program. The decision process can be seen as a two step process. The first step is the decision of the individual to watch television. The second step is the viewers’ decision to choose which channel to watch.
The decision of all individuals leads to the total viewership and the decision of the viewer of which channel to watch leads to the market shares. This approach presented above is general enough and all it needs to work is the existence of competitive environment, which is found in all the developed countries television environment.
The total viewership, is affected from the season of the year, the day of the week, the time of the day and the viewers life style.
The second module of the schema is the market share of each channel separately for each minute of the day. In this module we simulate the decision of the viewer of the channel he wants to watch. An innovation in this module is the usage of market response models. This way the program can incorporate the competition environment that exists among the television stations.
The third module of the schema composes the first two modules. The combination gives the viewership of each television station for every minute of the day.
The fourth module then produces program’s viewership. It takes as input the viewership that is produced for each channel every minute. Then the average viewership of the program is calculated for a day. This module gives as output the program's viewership, which is very important information for the television stations.
The fifth module produces the viewership of the commercials. The advertising messages are separated into two categories, commercials between programs (commercials between) and commercial within programs.
The system’s performance is better in comparison with the conventional method used. This conclusion has come out after experiments that were conducted for the system’s efficiency. The accuracy of the results is measured with the simple formula that the error in forecasts corresponds to the actual viewership. Error can be either an overpriced viewership or under-priced. The bigger the error, the bigger is the deviation of the actual viewership from the forecasted.
We have presented a decision support system, which incorporates qualitative factors for the forecasting of television viewership. The system can be applied to many countries environment for the forecasting of viewership. The system that is proposed here gives better forecasts than the forecasting method that the advertising companies use. It gives to the user a minimum forecast indication to be based on whether he is experienced or not. He can also judgmentally intervene in the forecasting process and include his experience in the system. It minimizes the time needed for the forecasting as it can run throughout the night and the user can have the results ready in the morning.
Easy access to the results, as the software allows networking of many users and “at the same time” file reading.
The software can exchange data with all the well-known packages that support the advertising companies’ tasks. That way the results from the system are passed to the packages easily and fast.
It allows the user to see the competitive relations that exist between programs and to follow the viewers’ preferences.
There is an important decrease in the time needed as the time that needed for the forecasting of the programs viewership for one month was 3 days. The software does that in only 2,5 hours.
περισσότερα
