Forecasting sales trends based
on previous sales using sophisticated methodologies
New Product forecast using
several proxy variables
Achieved Forecast accuracy
of over 90%
Methods: Double Exponential
Smoothing, ARIMA, etc.
Multiple individual statistical
model simulations run
Used to select the most
promising and stable models for forecasting
10 forecasting models
were developed in this process
How is an appropriate
forecasting model selected ?
Start with the set of
the 10 forecasting models
are run for each of the 10 models per ink
An algorithm picks the
model that results in the minimal forecasting error
Based on actual transaction
data in recent past
An MNC found that as it introduced
new SKUs it did not have a way of accurately forecasting
sales of its existing and new SKUs. Moreover, the
distributors and wholesalers did not have clear guidelines
for buying new SKUs.
We looked at past trends and
other variables to work out a system of forecasting
sales with 90% accuracy.
We also developed a
way of forecasting optimal shipments by taking into
account the inventory with trade.