Learning to Predict Email Open Rates Using Subject and Sender

Daniel Vitor de Oliveira Santos, Wladmir Brandão

2024

Abstract

The burgeoning daily volume of emails has metamorphosed user inboxes into a battleground where marketers vie for attention. This paper investigates the pivotal role of email subject lines in influencing open rates, a critical metric in email marketing effectiveness. We employ text mining and advanced machine learning methodologies to predict email open rates, utilizing subject lines and sender information. Our comparative analysis spans eight regression models, leveraging diverse strategies such as morphological text attributes, operational business factors, and semantic embeddings derived from TF-IDF, Word2Vec, and OpenAI’s language models. The dataset comprises historical email campaign data, enabling the development and validation of our predictive models. Notably, the CatBoost model, augmented with operational features and dimensionally reduced embeddings, demonstrates superior performance, achieving a Root Mean Squared Error (RMSE) of 5.16, Mean Absolute Error (MAE) of 3.60, a Coefficient of Determination (R2) of 77.53%, and Mean Absolute Percentage Error (MAPE) of 14.73%. These results provide actionable insights for improving subject lines and email marketing strategies, offering practical tools for practitioners and researchers.

Paper Citation

in Harvard Style

Vitor de Oliveira Santos D. and Brandão W. (2024). Learning to Predict Email Open Rates Using Subject and Sender. In Proceedings of the 20th International Conference on Web Information Systems and Technologies - Volume 1: WEBIST; ISBN 978-989-758-718-4, SciTePress, pages 59-70. DOI: 10.5220/0012945700003825

in Bibtex Style

@conference{webist24,
author={Daniel Vitor de Oliveira Santos and Wladmir Brandão},
title={Learning to Predict Email Open Rates Using Subject and Sender},
booktitle={Proceedings of the 20th International Conference on Web Information Systems and Technologies - Volume 1: WEBIST},
year={2024},
pages={59-70},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012945700003825},
isbn={978-989-758-718-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 20th International Conference on Web Information Systems and Technologies - Volume 1: WEBIST
TI - Learning to Predict Email Open Rates Using Subject and Sender
SN - 978-989-758-718-4
AU - Vitor de Oliveira Santos D.
AU - Brandão W.
PY - 2024
SP - 59
EP - 70
DO - 10.5220/0012945700003825
PB - SciTePress