Prompt
I’m building a demand forecasting model and need time-series data with enough history. What datasets fit?
Latest observation
Oct 16, 2026
OpenAI APIWeb search: off
For demand forecasting, you generally want datasets with:
- Long history: ideally several years
- Regular time granularity: daily, weekly, or monthly
- Many SKUs / stores / regions: if you need cross-sectional forecasting
- Clear target variable: sales, orders, units sold, shipments, etc.
- External drivers: price, promotions, holidays, weather, events
Here are good public datasets that fit well:
Retail / demand forecasting datasets
-
M5 Forecasting - Accuracy
- History: ~5 years of daily sales
- Level: item-store sales across Walmart
- Why useful: very popular for demand forecasting; includes price and calendar data
- Best for: hierarchical forecasting, intermittent demand, promotion effects
-
M4 Competition dataset
- History: varied; many series with long histories
- Level: multiple domains, including yearly/quarterly/monthly/daily
- Why useful: broad benchmark for forecasting methods
- Best for: benchmarking generic forecasting models
-
Rossmann Store Sales
- History: ~3.5 years daily
- Level: store-level drugstore sales
- Why useful: includes promotions, holidays, store metadata
- Best for: retail sales with exogenous variables
-
Corporación Favorita Grocery Sales
- History: multiple years daily
- Level: item-store grocery sales
- Why useful: large-scale retail demand dataset with promotions and oil/holiday effects
- Best for: large retail forecasting
-
Wikipedia Web Traffic
- History: daily page views over ~2 years
- Level: article traffic
- Why useful: useful as a generic demand-like time series benchmark
- Best for: traffic/demand-style forecasting
Energy / consumption-style datasets
-
GEFCom datasets
- History: long time series for load/energy demand
- Why useful: includes weather and calendar influences
- Best for: demand patterns with strong seasonality
-
Electricity Load Diagrams
- History: very long, high-frequency load data
- Best for: demand forecasting with strong seasonality and many series
Transportation / operations datasets
-
Bike Sharing Demand
- History: about 2 years hourly/daily
- Why useful: includes weather and seasonality
- Best for: short-term demand with exogenous variables
-
Taxi demand datasets
- History: often several years, hourly
- Why useful: good for spatial-temporal demand forecasting
- Best for: mobility/ride-hailing demand
If you want the best choices for “enough history”
- M5 Forecasting
- Corporación Favorita Grocery Sales
- Rossmann Store Sales
- Electricity Load Diagrams
- GEFCom
Quick rule of thumb
- For classical retail demand forecasting: pick M5 or Favorita
- For store-level promotion effects: pick Rossmann
- For generic forecasting benchmarks: pick M4
- For many long series: pick Electricity Load Diagrams
If you want, I can also give you:
- a filtered list by data frequency (daily/weekly/monthly),
- datasets that are easy to download in Python, or
- the best dataset for your specific use case.