Data Modelling

business intelligence architectures

Setting up business intelligence architectures is our passion!

Predicting sales on the basis of various factors like price, advertising spends (TV, print, radio), sponsorship spends etc. is one of the specialties at Tetrahedron. It is an iterative process, and once we are within the 90% confidence level mapping the predicted sales vs actual sales, then starts the art of optimizing the spends, thus saving millions of dollars in spends.

In the automotive industry, sometimes we get lost with the spends that we make!

There is the corporate regional spend, and the corporate local spend per market. Within the local spend per market; there is also the dealer kitty. Then there is the dealer spend on outdoor, cinema, newspaper, magazine and TV – which is different than the manufacturer spend on magazine! All this is part of the marketing mix, and different than the sales promotions and PR budget!

The objective of all this is to increase retail sales. So if we start with the end in mind – take the retail sales volume per month, and try to isolate the impact of each marketing mix variable, we should end up with the ones that are making the most impact. It is an iterative process, and/which broadly follows the logic that:

  • From a total of ‘n’ variables (spends), which ones impact retail sales in the order of importance,
  • For that extent of impact, how much resources (spends) are being allocated; are we generating a positive ROI for that activity?
  • For all activities that are generating a positive ROI, we then arrive at the OPTIMAL spend levels that help us achieve the same sales volume. This is an iterative exercise of varying the spend levels among high ROI activities and assessing their impact.

The modelling is run through multiple regression or the Markov process. Sometimes, we also use quadratic optimization, based on the Markowitz model.


Email to someoneShare on LinkedInGoogle+Tweet about this on TwitterShare on FacebookPrint this page