Benchmark artifact

Large forecasting applications often combine statistical, machine-learning, and neural models. These model families solve the same problem, but they differ in their fitted state, their training procedures, and the way they parallelize work. Forecasting software must therefore either hide these differences behind a single estimator interface, or keep the families in separate packages, which forces users to rewrite data preparation and evaluation for every package. In this paper, we present the Nixtlaverse, an ecosystem of open-source Python libraries for time series forecasting, as a case study of a third design: all libraries share the same long-format panel data and keyed forecast outputs, while every model family keeps its own specialized implementation. We demonstrate the behavior and benefits of this design through three use cases on the public M5 competition data. First, we evaluate statistical, machine-learning, and neural models, as well as an external engine from a separate ecosystem, in a single rolling-origin evaluation. The shared outputs make per-series and hierarchy-weighted accuracy metrics cheap to compute side-by-side. Second, we profile runtime and peak memory from 100 to 30,490 series and locate the bottleneck of each family: statistical fitting scales approximately linearly in the number of series, feature construction dominates machine-learning memory, and neural training time is nearly independent of panel size under a fixed training budget. Third, we reconcile the forecasts of multiple engines, including the external one, over all 42,840 series of the M5 hierarchy, using sparse reconciliation where dense implementations exhausted the memory of our machine. These use cases establish the costs and boundaries of the design and demonstrate the utility of shared data and output contracts for cross-family evaluation and reconciliation. The success of these design choices, and by extension the Nixtlaverse itself, has been validated by the substantial public distribution, scholarly reuse, and adoption through other forecasting frameworks. The Nixtlaverse is released under permissive open-source licenses, with public datasets, reproducible examples, and verifiable benchmark artifacts.