From bids to bliss

: achieving campaign excellence with ML and data insights

Abstract

Telecommunication companies encounter challenges, such as regulatory requirements, fast-paced technological changes, high competition, and customer expectations. One of their biggest challenges is the efficient allocation of their marketing budget. This study delves into three modeling approaches for optimizing budget allocation across four digital marketing channels to increase total sales. Our goal is to increase sales and transform digital marketing practices in the telecommunications sector by implementing AI-powered budget allocation methods. The three modeling approaches are a baseline (Random Forest) machine learning model leveraging the automatic creation of multivariate time-series data, as well as two Marketing Mix Models (MMM): one pipeline that implements the ideas from Meta’s Robyn framework and the other using Google’s Lightweight-MMM framework. We aim to gain maximum data insights from models offering diverse perspectives and approaches. We evaluate the models based on their ability to predict total sales, analyzing different performance metrics such as RMSE and R 2. Additionally, we assess the models for their practical relevance regarding their utilization for the marketing budget allocation. One important discovery is the significant impact of Channel 0 on total sales. However, the models only capture some of the dynamics and interactions between the marketing channels. The LightweightMMM model achieved the best performance, with a R2 of 0.666 on unseen data. Therefore, we recommend a further development of the MMM approaches in this study. Also, because a MMM provides detailed data insights for long-term budget allocation and captures the crucial carryover and saturation effect evident in digital marketing campaigns.

Date of Award	8 May 2024
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Nuno Filipe Loureiro Paiva (Supervisor)