Accuracy Is Not Service: A Decision-Aware Benchmark for Intermittent-Demand Forecasting

Demand forecasting is critical in modern industry, offering opportunities to reduce costs and gain competitive advantage through improved inventory management. However, forecasting becomes particularly challenging for products with intermittent demand, where demand occurs infrequently and time series contain many zero observations. Such dynamics are common across diverse sectors, such as industrial organizations, consumer goods, aviation, automotive, and electronics. Motivated by these challenges, this paper explores the potential of gradient boosting models to improve forecasting performance. We evaluate statistical, specialized, machine learning, and ensemble approaches across multiple datasets. The results show that specialized methods achieve the strongest performance among individual models, while gradient boosting on its own tends to underperform. However, combining a machine learning model with a specialized approach improves forecasting accuracy by up to 10%, demonstrating that even simple ensembles can outperform single models. Overall, the findings highlight the value of combining machine learning with domain-specific forecasting techniques for intermittent demand.

Retail demand forecasts are reused across replenishment, capacity, labor, and transportation planning cycles. Point-error objectives do not constrain abrupt movement between adjacent forecasts, while post-hoc smoothing acts only after model fitting. We ask whether a training-time penalty on consecutive within-series movement can improve horizontal forecast-path stability without materially changing point accuracy. The penalty is evaluated in a temporal-structured pipeline combining recent-demand embeddings with calendar, price, hierarchy, item, and store features. On selected M5 demand series at 1000, 3000, and 4000-series scales, the stability-aware hybrid model improves Forecast Stability Score over XGBoost by 6.91%, 6.66%, and 7.68%, respectively, while RMSE changes remain within 0.72% across three random seeds. Post-hoc exponential smoothing attains lower raw movement but incurs a larger RMSE cost; training-time regularization preserves more point accuracy and performs favorably under normalized stability. These findings extend forecast evaluation from point-error minimization toward an accuracy-stability trade-off perspective for operational retail forecasting.

A forecasting benchmark reports which method won. We show that the answer is set by the evaluator's choices before any model is fitted. We benchmark 24 forecasting methods and one textbook reference, including six 2025-era time series foundation models, on a production marketplace panel of 1,887 business customers over 67 months. We hold the data, the horizon and the period fixed, and vary only the evaluation design. Three choices each reverse or dissolve a headline conclusion. Changing the unit of analysis from the market total to the individual customer moves our production baseline from second of nineteen, beaten by nothing, to twenty-third of twenty-five. Nineteen of its twenty-four challengers beat it there. Changing how much error is pooled decides whether a Diebold-Mariano test finds anything at all. Scoring prediction intervals rather than point forecasts reorders the field almost completely, with a rank correlation of 0.02 on intermittent demand. We then measure what the deployed system gets from this. Its selection rule captures 55% of the distance between doing nothing and choosing with hindsight. The reversal is not a quirk of our data. We ran the released protocol, unchanged, on the public M5 retail panel. The same baseline shape places first at the market total and last per series, beaten by everything, and a replayed selection rule closes 64.7% of the same floor-to-ceiling distance there. Adding five zero-shot foundation models to that roster changes who wins at the total, not the shape. The bands' blind spot travels too: conformal bands under-cover most on the spikiest items. Splitting our own panel into ever smaller groups turns the contrast into a curve: the baseline's rank worsens at every level of disaggregation. We release the evaluation protocol and report an error of our own that inverted a result before we caught it.