Interpretable Deep Learning for Forecasting Online Advertising Costs: Insights from the Competitive Bidding Landscape

Fynn Oldenburg,Qiwei Han,Maximilian Kaiser

2024-08-21

Abstract:As advertisers increasingly shift their budgets toward digital advertising, accurately forecasting advertising costs becomes essential for optimizing marketing campaign returns. This paper presents a comprehensive study that employs various time-series forecasting methods to predict daily average CPC in the online advertising market. We evaluate the performance of statistical models, machine learning techniques, and deep learning approaches, including the Temporal Fusion Transformer (TFT). Our findings reveal that incorporating multivariate models, enriched with covariates derived from competitors' CPC patterns through time-series clustering, significantly improves forecasting accuracy. We interpret the results by analyzing feature importance and temporal attention, demonstrating how the models leverage both the advertiser's data and insights from the competitive landscape. Additionally, our method proves robust during major market shifts, such as the COVID-19 pandemic, consistently outperforming models that rely solely on individual advertisers' data. This study introduces a scalable technique for selecting relevant covariates from a broad pool of advertisers, offering more accurate long-term forecasts and strategic insights into budget allocation and competitive dynamics in digital advertising.

Machine Learning,Artificial Intelligence,Social and Information Networks

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to accurately predict the daily average cost - per - click (CPC) in the digital advertising market. As advertisers increasingly shift their budgets to digital advertising, accurate prediction of advertising costs has become crucial for optimizing the returns of marketing campaigns. This paper aims to solve this problem by using various time - series forecasting methods to predict the daily average CPC in the online advertising market.

Specifically, the researchers evaluated the performance of statistical models, machine - learning techniques, and deep - learning methods (including Temporal Fusion Transformer (TFT)). They found that extracting covariates from competitors' CPC patterns through time - series clustering and incorporating them into multivariate models can significantly improve prediction accuracy. In addition, the study also demonstrated how these models can utilize advertisers' data and insights from the competitive landscape, especially their robustness during significant market changes (such as the COVID - 19 pandemic).

The main contribution of the paper lies in introducing a scalable technique for selecting relevant covariates from a large number of advertisers, thereby providing more accurate long - term predictions and strategic insights regarding budget allocation and the competitive dynamics of the digital advertising market. This goes beyond the capabilities of existing tools (such as Google's Keyword Planner), which can only provide estimates of the general weekly average maximum CPC based on different bidding strategies, but cannot adjust to generate actual CPCs or gain in - depth understanding of the cost structure in the online advertising market.