Our proposed solution
To run a time series model incorporating both historical macro economic variables as well as marketing mix variables (of our client and its competitors). Given existing estimates of how the macro economic situation will change in the future, we incorporated this in the predictive models as well.
We collected a lot of data, both from our client as well as from official websites (e.g. local statistical website, the economist, …), covering a huge variety of aspects, possibly related to the sales changes of a product category. The first stage of any (predictive) modelling project involves cleaning and exploring the data to uncover possible trends that we might want to capture in the model. It also tells us which model to use to optimise the usability of the results. It is an iterative process, and we very much like to understand the outcome of the model (so not black box modelling). The models are also tested by using partial data (i.e. not all years) to predict the current situation.
"How can we better anticipate a possible sales stagnation or decline for the coming years and predict future sales for our product categories?"
Results and benefits to our client
As a result, we were able to predict how sales volume would change in the next 5 years. We did produce a few scenarios (both a pessimistic and optimistic forecast) to give a good view of what our client could expect in terms of sales.This was incorporated in their business plan for the next years,especially in terms of budgetplanning.
Photo by JessicaGale at Morguefile.com