How Faithful Is Attribution for Sales Forecasting? A Counterfactual Study

Glib Kechyn

cs.LGSep 4, 2026

How Faithful Is Attribution for Sales Forecasting? A Counterfactual Study

Authors: Glib Kechyn

Abstract

Deep models for sales forecasting, such as WaveNet-style dilated convolutional networks, are accurate but opaque: when a single model predicts sales for one of many series, it offers no account of why. We add a post-hoc, architecture-agnostic counterfactual interpretability layer to a multi-series WaveNet forecaster trained on the full Corporacion Favorita grocery dataset (174,685 series over 1,688 days). The method decomposes each forecast into contributions that sum exactly to the predicted value, avoiding the allocation artifacts we observed with additive SHAP-style attribution. We evaluate faithfulness with a deletion/insertion protocol and find a statistically significant effect on both tests (deletion gap 0.22, p<0.001; insertion gap 0.27, p<0.01; robust across five background-sampling seeds), establishing that the attributions reflect genuine model behavior rather than plausible-looking artifacts. We then characterize, honestly, where attribution is and is not informative: reliance on the promotion signal is heterogeneous across series (median ratio approximately 1.0, with roughly 20% of series showing a strong effect), and the model captures the shape of the weekly sales cycle (day-of-week r=0.78) while systematically under-predicting its amplitude. Our contribution is not improved accuracy but an interpretability layer with a rigorous faithfulness evaluation and a candid account of its limits.

Figures & tables

Figure 1: Training loss (MSE in $\log(1{+}x)$ space, left) and validation NWRMSLE (right) over $60$ epochs for the multi-series WaveNet with the onpromotion covariate. Validation reaches its best value of $0.6102$ at epoch $43$ ; the subsequent gentle decline in training loss alongside a slight rise in validation indicates mild late overfitting.

Figure 2: Faithfulness of the counterfactual attributions, aggregated over $30$ series. Left: deletion—masking positions in attribution order (solid) degrades the forecast far faster than random order (dashed). Right: insertion—revealing positions in attribution order recovers the forecast far faster than random. Results shown for background seed $0$ ; deletion is significant at $p<0.001$ and insertion at $p<0.01$ , on this and all five seeds tested.

Figure 3: Distribution across $40$ series of the ratio between mean absolute attribution on promotion days and on non-promotion days. The median (green) is $1.02$ ; the distribution has a long right tail, and only about $20\%$ of series exceed a ratio of $2$ .

Figure 4: Actual vs. predicted day-of-week sales profile, aggregated over

300

series and normalized to each series’ own average (

r=0.78

). The model captures the shape of the weekly cycle but under-predicts its amplitude: every predicted bar is closer to

1.0

than the corresponding actual bar.

Explore similar work

Information Bottleneck Learning for Faithful Time Series Forecasting Explanations
Jul 30, 2026Xu Zheng, Wei Cheng, Zhuomin Chen +3Multivariate Time Series Forecasting Explainable Time Series Forecasting
Florida International University, Miami, FL, USA · NEC Laboratories America, Princeton, NJ, USA · University of Houston, Houston, TX, USA +1
NVExplain: Explaining Time Series Forecasting with Latent Trajectory Analysis and Structure-Preserving Surrogates
Aug 25, 2026Muyan Anna Li, Manikandan Ravikiran, Aditi GautamExplainable Time Series Forecasting Interpretable ML
NVIDIA, USA
ConTex: Reformulating Counterfactual Generation For Time Series Forecasting
Jun 16, 2026Jan Voets, Hasan Tercan, Tobias Meisen +1Explainable Time Series Forecasting Counterfactual Explanations
Institute for Technologies and Management of Digital Transformation University of Wuppertal

Jul 30, 2026, cs.LGJ/K move · Enter open · S save

Information Bottleneck Learning for Faithful Time Series Forecasting Explanations

Xu Zheng, Wei Cheng, Zhuomin Chen, Mo Sha+2

Florida International University, Miami, FL, USA · NEC Laboratories America, Princeton, NJ, USA · University of Houston, Houston, TX, USA+1

Multivariate Time Series Forecasting Explainable Time Series Forecasting Interpretable ML Time Series Forecasting Information Bottleneck

Open paper AskPDF arXiv

As forecasts increasingly drive decisions in fields such as energy, transportation, and healthcare, understanding the historical data behind these predictions has become as crucial as the predictions themselves. Although existing interpretable-by-design forecasters reveal their internal structures, they offer no guarantee that these structures faithfully reflect the underlying evidence driving the predictions. In contrast, while faithfulness-oriented methods explicitly verify model behavior, they are almost exclusively designed for post-hoc classification tasks. To bridge this gap, we propose IB-Forecast, an inherently interpretable multivariate time-series forecasting framework. It decomposes forecasting into a learned periodic component and a residual component computed with explainable masks over input tokens. With a budget-constrained information bottleneck, end-to-end optimization enables users to directly control explanation sparsity. With a rigorous faithfulness evaluation protocol, extensive experiments demonstrate that IB-Forecast matches the forecasting error of leading black-box models while providing faithful explanations at no additional inference cost. Furthermore, under a matched sparsity budget, these native explanations consistently surpass gradient-based, occlusion-based, and optimization-based baselines across all evaluated datasets. Ultimately, whereas the native explanations of existing interpretable forecasters exhibit poor faithfulness, IB-Forecast guarantees high explanation fidelity, requiring only 14-20% of the observations to deliver low-error predictions.

Abstract

Explore similar work

Information Bottleneck Learning for Faithful Time Series Forecasting Explanations

NVExplain: Explaining Time Series Forecasting with Latent Trajectory Analysis and Structure-Preserving Surrogates

ConTex: Reformulating Counterfactual Generation For Time Series Forecasting