The amount of historical data included should balance between capturing enough historical trends and seasonality for accurate modeling and considering the relevance and quality of the data. Regularly updating and validating the model with the most recent data is also crucial to maintaining its accuracy over time.
Recommended Historical Data Duration
Minimum Requirement: A minimum of two years of historical weekly data is recommended. This duration is generally considered sufficient to capture annual seasonality and any significant market changes.
However including more historical data can be beneficial, especially if your business or industry is subject to multi-year cycles or if there have been significant changes in strategy or external conditions that could influence performance.
Granularity
Weekly: Weekly data is preferred for its ability to more accurately capture fluctuations and the effects of marketing activities while excluding “noise” in the data
More granular data provides a detailed view of performance but requires careful consideration of data quality and consistency.
💡 We strongly suggest gathering data Daily when possible as more granular data can easily be aggregated to weekly for instance, while the opposite is not true.