Analysing the challenges and opportunities in Media Mix Modelling using encrypted sales and media spending time series data.
Media Mix Modelling is a statistical analysis technique widely used by many businesses for the past two decades for evaluating the effectiveness of different media spends on sales volume or sales revenue. They use the inferences obtained in driving optimal budget allocations. Having been commonly used for so many years, advertisers and modellers ought to know the various challenges and opportunities involved in the technique, facilitated by a clear understanding of the fundamental concepts and methodologies. This analysis uses encrypted time-series data of media spending and sales information and shows how linear regression analysis on such data can extract valuable insights for efficient decision-making. Besides pointing out the favourable aspects during the process, our analysis will show the difference between data-driven model results and those based on modellers’ choices in the model building with the uncertainties involved in such results.
This analysis was done as part of the Business Intelligence course held at Dalarna University for the master in the data science program. If you are someone starting to get the hang around media mix modelling, do have a look at this mini-project of mine.
The analysis is presented in MMMAnalysisReport.pdf and the code for analysis in media_mix_modelling.ipynb. In addition, the images used in the report are present separately for easy access to the information.
This project is licensed under the MIT License - see the LICENSE file for details.