Your Best Volume Month Could Be Destroying Your Forecast: The Hidden Danger of Demand Signal Pollution
Your best volume month could be your most dangerous data point for your forecast model…
#Volume_spikes_clearance_sales
You just sold off your slow moving and dated inventory in a liquidation sale to a dollar store. At a deep discount – 40% off. Your historical volume spikes. The model sees the pattern and creates a spiky forecast.
The sales team sees momentum. Operations sees growth. Production starts planning for more volume.
Alternate scenario. The retailer clears the inventory and creates a massive load-in. Your sell in spikes.
But this is not #true_demand! It was an exit – the timing is unpredictable in the future. Just like a massive volume drop due to a hurricane.
What happens next is where value gets destroyed.
The spike either inflates the baseline forecast, or completely removed as an outlier, eliminating the context behind it. In both cases, leadership is left making production and inventory commitments based on a demand signal that is compromised.
The warning signs are usually there:
Drop in weighted average price in those months (Note this) 
Channel inventory being liquidated 
Short-term volume with no repeat pattern
So weighted average price is an important indicator of demand signal quality. Viewed in isolation and modeled can cause volatile forecasts that may not come to fruition.
Your model needs to harmonize for these volume effects if your ML model has to produce a holistic and realistic forecast.
If compromised demand signals are distorting your baseline forecast, contact the Valtitude team to explore how we engineer cleaner, context-aware demand planning process.
Read our related posts: