← All Posts /

# Modelling the 4P Mix

How far should I hike the price of my product? Is my distribution strategy working as it should? Should I spend more on traditional media or social media? Am I getting the right ROI for my marketing expenses?

Marketing Mix Modelling is one of the best methods available to value and analyze various marketing tactics on sales and/or market share, and then forecast the future impact of the similar activities. This basically means quantifying the information available with the firm to analyze their impact on sales and market share.

It uses both the internal information (Pricing, Quality Metrics, Inventory Levels, Spending on Print Media) as well as the external information (Macroeconomic Situation, Seasonality and Competitors). The key objective is to maximize the return on investment while minimizing the budget required for the same.

Digital Transformation of various business processes have helped the collection of better data for modelling purposes. This is because in order to create an effective model, dynamic metrics, such as social media and customer engagement metrics, help. GRP (gross rating point) in TVs and Click-through rates in Online Advertising are some of the popular metrics often used to understand the impact of different promotion methods.

We will be illustrating two important concepts of Marketing Mix Modelling- Linear Regression and Incremental Drivers.

Using Linear Regression for Marketing Mix Modelling

Marketing Mix Modelling uses the concept of Multiple Regression. The dependent variable can be Sales while the independent variables commonly used are Price, Advertising spends, Location, product discounts, and so on.

Before diving into the exacts of Linear Regression, let’s understand an important component of Marketing Mix Modelling- Adstock. Adstock refers to the prolonged effect of advertising on consumers. This tool can help us quantify the impact of advertising.

The basic Adstock model is as follows

At = Tt + λAt-1

Where, At is Adstock at time t, Tt is value of advertising variable at time t, and λ refers to the decay parameter. Thus, when t approaches infinity, the Adstock value approaches 0. We will use Adstock as a component of the Linear Regression equation that we will create.

We will also include one component for seasonal trend in our equation, as well as one for the number of distribution points. To illustrate, imagine that there is a Clothing store which wants to know the relative importance of parts of its Marketing Mix. The store usually experiences higher sales in Summers than at other times of the year (Seasonal Trend). It has around 100 outlets across India (Number of Distribution Points) and advertises only through Online Media (for simplicity purposes).

Here is an example of how the regression equation would look like-

Sales = β0 + β1 (Seasonal Trend) + β2(Number of Distribution Points) + β3(Adstock of at time t)

The betas generated from Regression analysis will help in quantifying the impact of each of the inputs. Basically, the beta depicts that one unit increase in the input value would increase the sales by Beta times those units, keeping the other inputs constant.

Many other attributes can be added to this equation; even polynomial features can be used in order to show non-linear changes. Actual models are much more complicated than the one shown above.

Distinguishing Base Sales from Incremental Sales

Marketing Mix Modelling breaks down the Business metrics into two major contributors- Base Drivers and Incremental Drivers.

Incremental Drivers refer to the business results generated by using marketing tactics different from those used in the normal course of business, while Base Drivers refer to the business results generated in the normal course of business, mainly due to the goodwill and brand equity developed over the years.

Using the same example as before of a clothing store, the base drivers would be that of the number of outlets, its advertisement stock, and the average price level. A temporary reduction in the price would be an example of an incremental driver. The effect of this incremental driver can be separated and analysed to measure its impact.

There are innumerable other concepts that can be applied to model the 4 Ps of Marketing than Linear Regression and Incremental Drivers, such as Budget Optimization or Deep-dive analysis. Usually these concepts are not used in isolation, but rather a collection of all these concepts is used. Marketing Mix Modelling has a much wider scope and shows us how Analytics can be useful for business.

A customer’s journey of ‘from thinking to buying’ takes him/her through multiple touch points before deciding the final product to buy. As marketing is becoming more customer centric, the need of identification of right channels to target potential customers has become critical for companies. This helps companies to utilise their marketing funds in a better way and target the right customers in right places. So, we hope that you must have got the gist of the concept. Hit the like button, if you liked our post; and subscribe our blog to be connected to us, as many more wonderful topics are underway-)

Recommended » Benford's Law