Marketing mix modeling : case of an iGaming company
Anwer, Mujtaba (2023)
Pro gradu -tutkielma
2023
School of Business and Management, Kauppatieteet
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2023080793257
https://urn.fi/URN:NBN:fi-fe2023080793257
Tiivistelmä
This thesis explores cutting-edge marketing mix modelling (MMM) methods and their application to a case-company's data from the iGaming industry. The objective was to quantify the impact of media channels on sales and develop an accurate sales forecasting model using marketing variables. Three methods were implemented: log-differenced OLS regression, Hamiltonian Markov Chain Monte Carlo simulation, and Adaptive Neuro Fuzzy Inference System trained using Particle Swarm Optimization.
OLS regression showed strong goodness of fit but moderate performance with test data, while exhibiting some multicollinearity and deviations from normality. In contrast, MCMC simulations fully converged and estimated the contribution of posterior distributions of media variables, to the target distribution of sales. The ANFIS model provided high forecasting accuracy and validated the rankings of media channels' contributions to sales provided by the other two methods. The research revealed that TV, Online Display, and Affiliates were the primary contributors to the case-company's sales, whereas the saturation points for TV advertising spend were far above other media channels. This confirms what is already known from literature (Fareniuk & Chornous, 2023) that TV tends to have a higher contribution than digital media, due to its wider reach, and a higher saturation point due to the various programmes during which advertisements can be placed to reach different age groups & segments of audience.
Overall, this thesis demonstrates the significance of using multiple MMM methods in conjunction to gain deeper insights and support decision-making in the marketing industry. The findings provide practical implications for allocating capital to media channels and highlight the potential of explainable Ai methods such as ANFIS with PSO for forecasting in MMM.
OLS regression showed strong goodness of fit but moderate performance with test data, while exhibiting some multicollinearity and deviations from normality. In contrast, MCMC simulations fully converged and estimated the contribution of posterior distributions of media variables, to the target distribution of sales. The ANFIS model provided high forecasting accuracy and validated the rankings of media channels' contributions to sales provided by the other two methods. The research revealed that TV, Online Display, and Affiliates were the primary contributors to the case-company's sales, whereas the saturation points for TV advertising spend were far above other media channels. This confirms what is already known from literature (Fareniuk & Chornous, 2023) that TV tends to have a higher contribution than digital media, due to its wider reach, and a higher saturation point due to the various programmes during which advertisements can be placed to reach different age groups & segments of audience.
Overall, this thesis demonstrates the significance of using multiple MMM methods in conjunction to gain deeper insights and support decision-making in the marketing industry. The findings provide practical implications for allocating capital to media channels and highlight the potential of explainable Ai methods such as ANFIS with PSO for forecasting in MMM.