In an increasingly competitive world, companies are making huge marketing investments to sustain brand loyalty and competitive advantage. However, the pressure on margins is very intense and hence, companies need to maximize sales and profitability on every dollar they spend on the key marketing elements (a mix of advertisement, other promotional measures, trade support, pricing, and distribution).
Allocation of Marketing investments to these elements have largely been based on experience of the marketing managers (what mix has worked historically?), competitor's marketing strategy and a qualitative understanding of customer buying behavior (customer reaction to in-store promotions, advertisement, temporary discounts, etc.). What is lacking is a scientific way of evaluating past marketing decisions on a consistent basis, and using this evaluation to drive future marketing spends, so as to maximize the ROI on marketing spends.
One of the scientific approaches that progressive organizations use today to optimize marketing spend is Market Mix Models. These models uncover customer sensitivity to the marketing decisions undertaken in the past by establishing the link between spends and incremental sales. The effect of competitor marketing activities on sales / market share can also be looked at, depending on data availability.
These models help in:
Market mix models have evolved from simple additive and multiplicative models to Multinomial Logit (MNL) models and Multiplicative Competitive Interaction (MCI) models. The choice of models is problem-specific. For instance, MCI models have proven superior to other models in multiple brand scenarios, when interaction of competitor brands spends also needs to be looked at. Recently, Artificial Neural Networks have also been used. ANNs are very effective when significant inter-relationships exist between spends on marketing mix elements and / or relationships between spend and sales volume is non-linear.
The main challenge in developing good marketing mix models is lack of right data. Consider the problem of measuring effectiveness of advertising on sales. For this, Gross Rating Points (GRPs) are typically collected for the target customer segment, for specific programmes and related to sales. However, GRPs by themselves only represent the number of people watching the program. To understand impact, data on the quality of advertisement also needs to be included.
The main limitation of this class of models is that they assume that overall structure of marketing activities and underlying customer behavior remains same. This is because these models work by uncovering patterns in past customer spend. This can be countered by running these models on a periodic basis, whenever major changes (new competitors, new kinds of marketing activities, new customer segments) happen.
To conclude, well-built marketing mix models help managers prioritize marketing spend, so as to maximize marketing ROI through what-if scenarios. A side benefit is that they help build a culture of tracking data on customer reactions to spend by self and competitors. Over a period of time, they also help organizations make their marketing strategy more effective by aligning it to needs of the market.
2000-10 DecisionCraft Inc.