US-based large media firm focused on commercial broadcasting, television production and publishing.
The client wanted to forecast video-on-demand requirement for a new show that was yet to be telecasted. The client wanted a separate model for each service provider.
SG Analytics undertook the following steps to deliver the desired output to the client:
Data preparation: SG Analytics' team aggregated data at a daily level, and merged the program scheduling information. The team created a training/validation sample and identified proxy data for the new show based on genre, telecast time, etc.
Model building and validation: SG Analytics built time series models to forecast orders and capture the trends, seasonality, and program airing patterns. We also included post-model adjustments based on the business situation and built a separate model for each service provider, as required. To validate the model, SG Analytics' team checked the Mean Absolute Percentage Error (MAPE) i.e. a difference between actual and forecasted orders.
Results: Our model was able to forecast orders for up to one year with an accuracy of 80%. The tool was able to present results in an intuitive dashboard format. The model also had the provision of handling dynamic changes in program schedules.
Advertisement placement: SG Analytics built an algorithm for dynamic ad placement in the content, whereby advertisements could be placed at the beginning, middle or towards the end of the program based on factors such as demographics, historical usage, etc.
Enabled revenue maximization by effective advertisement pricing and placement.
Implemented business changes in the episode placement, taking the viewership pattern into account.