Primary Submission Category: Difference in Differences
Evaluating causal impact of device deprecation and beyond
Authors: Cindy Cui,
Presenting Author: Can Cui*
Older devices present security risk and high maintenance costs as technology evolves and new consumer devices get adopted at faster pace. Business generally approach deprecation their services on certain devices in a planned manner cautiously, balancing potential impact on device manufacturers and consumers. We study the impact of depreciating older devices using observational causal method. We covers different use cases, including simple one-time treatment with valid control group where differences-in-differences or synthetic control works well, and more complex cases where we lack control group or the timing and implementation of treatment is more complicated. In addition to examining performance of these methods, we also build a predicted impact model to support ongoing and future business decisions on which device to retire and when to do so, in order to minimize negative impact on customer and device partners while reducing engineering maintenance costs and security risks. We will share our learnings on performance of various observational causal methods in this setting, and our efforts to setup semi-automatic process to apply most suitable approach for impact evaluation. Furthermore, we will also discuss how we can go beyond measurement and projecting forward potential impact to guide future interventions.