Get Started
Home
Topics
Search
Library
Research questionHow can we identify which inputs genuinely drive an opaque sales forecast?Accurate deep sales forecasters can conceal which signals drive individual predictions. Post-hoc attribution methods may produce plausible explanations that do not reflect the model’s actual behavior.
AI
Business
Evaluation & Benchmarks
Machine Learning
Research Paper
Statistical Machine Learning
Technology
Latest papersRecent research connected to this question, newest first.How Faithful Is Attribution for Sales Forecasting? A Counterfactual StudyThe evidence concerns a multi-series WaveNet-style forecaster trained on the Corporación Favorita grocery dataset, with 174,685 series over 1,688 days. It studies an architecture-agnostic counterfactual attribution layer using exact decomposition and deletion/insertion faithfulness tests, then examines promotion effects and weekly sales-cycle patterns. The findings show statistically supported faithfulness, heterogeneous promotion reliance, and systematic under-prediction of weekly-cycle amplitude; they do not establish improved forecast accuracy or universal interpretability.research paper · Sep 4, 2026
Related questions
How can we detect internally incoherent language-model forecasts before relying on them for consequential decisions?How can we tell whether internal estimates guide effective actions, rather than merely predict effects accurately?How can evaluators distinguish missing knowledge from miscalibrated outputs in language models?How can historical base-model residuals robustly correct multivariate forecasts without overcorrection?