Double Machine Learning for Downstream Causal Measurement: A Hybrid Econometric-ML Framework for Feature Valuation
International Journal of Computational and Experimental Science and Engineering · 14.12.2024 · 10 (4)
DOI: 10.22399/ijcesen.5432 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Computational and Experimental Science and Engineering, online erschienen am 14.12.2024. Verfasst von Dharmateja Priyadarshi Uddandarao und 2 weiteren Autor:innen. Die Arbeit nennt 46 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Digital platforms require defensible evidence on which product features causally improve downstream customer value, yet feature-prioritization systems often rely on predictive importance, usage frequency, or descriptive engagement indicators. These measures can mislead product teams because customers who adopt advanced features may already be more engaged, better onboarded, or less likely to churn. This study develops and tests a hybrid econometric-machine learning framework for causal feature valuation using customer-level B2B SaaS product analytics data. The dataset covers 24 months, 18,742 customer accounts, 126,815 active users, 12 product features, 78,450 treatment observations, and 91,220 control observations across 27 countries. Double Machine Learning with cross-fitting was used to estimate causal effects of product-feature adoption on retention, renewal probability, and expansion revenue while controlling for high-dimensional behavioral, subscription, and customer-level confounders. The results show that the AI Recommendation Engine produced the largest causal effect on retention, renewal, and incremental annual recurring revenue, while predictive feature rankings based on SHAP differed substantially from DML-based causal rankings. The study contributes a causal feature valuation construct and a reproducible framework for separating predictive association from downstream causal value in digital product analytics.
Zitieren
APA
Uddandarao, D. P., Mahajan, A., & Krishnamsetty, P. (2024). Double Machine Learning for Downstream Causal Measurement: A Hybrid Econometric-ML Framework for Feature Valuation. International Journal of Computational and Experimental Science and Engineering, 10(4). https://doi.org/10.22399/ijcesen.5432
Vancouver
Uddandarao DP, Mahajan A, Krishnamsetty P. Double Machine Learning for Downstream Causal Measurement: A Hybrid Econometric-ML Framework for Feature Valuation. International Journal of Computational and Experimental Science and Engineering. 2024;10(4). doi:10.22399/ijcesen.5432
BibTeX
@article{Uddandarao2024,
author = {Uddandarao, Dharmateja Priyadarshi and Mahajan, Ankush and Krishnamsetty, Prashanth},
title = {Double Machine Learning for Downstream Causal Measurement: A Hybrid Econometric-ML Framework for Feature Valuation},
journal = {International Journal of Computational and Experimental Science and Engineering},
year = {2024},
volume = {10},
number = {4},
doi = {10.22399/ijcesen.5432}
}
EUVIMED