Seasonal adjustment in forecasts for retail — solving the Christmas spike problem

Started by Priya Ramaswamy · 1 year, 1 month ago · 3 replies · 2 views

Priya Ramaswamy Participant

1 year, 1 month ago

Our business has strong seasonality (retail). When I initialise a forecast using 'average of last 3 months', it doesn't capture the Christmas spike. What's the best way to handle seasonality in Fabric Plan forecasts?

3 Replies

Sort

Sofia Martínez

Budget & Forecast Analyst · IberiaFin · 1 year, 2 months ago

We use the corresponding prior year approach. The one thing to watch is extraordinary prior year events (a one-off promotion) that would distort the seasonal baseline. We exclude those programmatically in the semantic model.

Up

0

Raj Subramaniam

Microsoft Fabric Consultant · CloudBridge Solutions · 1 year, 2 months ago

For seasonal businesses, use 'corresponding prior year period' as your forecast initialisation method, not trailing average. This seeds December 2025 from December 2024 actuals, capturing the seasonal pattern. You can then apply a growth rate on top.

Up

0

Marcus Delacroix

Senior BI Architect · Peak Logistics Europe · 1 year, 2 months ago

Alternatively, build a seasonal index measure in your semantic model (each month's % of annual total, based on 3-year average) and use it as the distribution weight when allocating annual forecasts to months.

Up

0