About Us 
Careers Students
Pointlogic has developed several tools for the evaluation and optimization of advertising effects.
One of these tools, Chorus, supports strategic multimedia planning. It helps people to design an optimal media mix, taking into account factors such as sales, awareness and brand image. The evaluation and optimization of potential plans is based on econometric, non-linear models which are incorporated into the system. These models describe media performance i.e. the non-linear relationship between media spend (in terms of budget) and media effect (in terms of marketing variables such as sales). The models used by Pointlogic are client specific and are derived from historical data, such as time series of media spend and sales effects.
When the client is working with multiple brands or products, the econometric model must also relate brand and campaign characteristics to media performance. This requires the construction of a ‘meta model’, which is a complex process.
Approach
Because of the complexity of this hierarchic, non-linear meta model, it must be estimated in a number of consecutive steps, for instance, firstly constructing a model for each individual brand or campaign. This is a time consuming and suboptimal process, from a theoretical point of view, and so the goal of this thesis was to improve the model estimation process.
Results
Pointlogic developed an estimation technique based on Bayesian theory in which the complete hierarchic model was estimated in one step. The procedure proved to be theoretically sound and efficient to use.
Careers Students
Marketing-mix Estimation
Pointlogic has developed several tools for the evaluation and optimization of advertising effects.
One of these tools, Chorus, supports strategic multimedia planning. It helps people to design an optimal media mix, taking into account factors such as sales, awareness and brand image. The evaluation and optimization of potential plans is based on econometric, non-linear models which are incorporated into the system. These models describe media performance i.e. the non-linear relationship between media spend (in terms of budget) and media effect (in terms of marketing variables such as sales). The models used by Pointlogic are client specific and are derived from historical data, such as time series of media spend and sales effects.
When the client is working with multiple brands or products, the econometric model must also relate brand and campaign characteristics to media performance. This requires the construction of a ‘meta model’, which is a complex process.
Approach
Because of the complexity of this hierarchic, non-linear meta model, it must be estimated in a number of consecutive steps, for instance, firstly constructing a model for each individual brand or campaign. This is a time consuming and suboptimal process, from a theoretical point of view, and so the goal of this thesis was to improve the model estimation process.
Results
Pointlogic developed an estimation technique based on Bayesian theory in which the complete hierarchic model was estimated in one step. The procedure proved to be theoretically sound and efficient to use.