By Bill Seota, Data Scientist

Marketing is a field that I find particularly interesting. I recently delved into Marketing Mix Modelling (MMM) and was sceptical when I came across it being referred to as a Money-Making Machine. Can it be? Let’s unpack this technique that is being used extensively by company’s marketing professionals and advertising agencies.

MMM is a technique used to analyse how different marketing activities impact sales. Using historical data, it involves creating a model that includes sales as the dependent variable and various marketing activities (like television advertising, online advertising, outdoor advertising, discounts, and the like) as independent variables. By analysing this data, the model can show which activities contributed most to sales and how effective and efficient each activity was.

Specifically, a MMM analysis can help organisations:

• Analyse how changes in spending affect media channel performance.
• Examine the impact of each media channel on their key performance indicators (KPIs), likely sales/revenue.
• Determine the best budget distribution across different media channels as a result of the above.

MMM can be used to:

• Analyse media and advertising: Determine the impact of different media types (like television, print, and online advertisements) on sales.
• Evaluate trade promotions: Understand how sales promotions (like discounts and special displays) affect sales in different retail outlets.
• Assess pricing: Measure how changes in price affect sales and determine the price elasticity (how sensitive sales are to price changes).
• Optimise distribution: Identify which distribution channels (like supermarkets or local shops) are most effective.
• Launch new products: Measure the impact of marketing efforts on the sales of new products.

To conduct this analysis, you need the following data:

• Media: Metrics for each channel over a specified time span (for example, impressions per time period).
• Target (to optimise): The KPI the model aims to predict (for example, revenue amount and number of app installations). This is the target that your MMM will seek to optimise.
• Costs: The total cost per media unit for each channel.
• Other: Additional features you may want to include.

With these inputs, a MMM analysis can effectively determine the most efficient allocation of marketing resources, predict the impact of various marketing activities on key performance indicators, and optimise your marketing strategy.

An example of MMM applied:

A good example of a problem that can be solved using MMM, is when launching a new smartphone with a mix of television and digital advertisements to determine the most effective medium.

The data required:

• Media Data: Weekly impressions per time period for television and digital advertisements.
• Target: The number of smartphones sold, which the model aims to predict and optimise.
• Costs: The cost per impression for television and digital advertisements.
• Other Features: Competitor advertising spend.

As output, you could find the following results:

1. Analyse how changes in spending affect media channel performance

The model shows that increasing the digital advertising spend by 20% results in a 15% increase in impressions and a 10% boost in sales. Conversely, increasing television advertising spend by the same amount leads to a 5% increase in impressions but only a 2% rise in sales, indicating diminishing returns on additional television advertising spend.

2. Examine the impact of each media channel on your key performance indicators (KPIs), such as sales

The regression analysis identifies that every R10 000 spent on digital advertisements generates an additional 500 units sold, while every R10 000 spent on television advertisements results in an additional 200 units sold. This highlights the higher efficiency and ROI of digital advertisements in driving smartphone sales.

3. Determine the best budget distribution across different media channels

The analysis reveals that allocating 60% of the smartphone marketing budget to digital advertisements and 40% to television advertisements maximises overall sales. This distribution effectively uses the higher engagement rates of digital platforms and the reach of television advertisements over the period.

While the above is a crude example that shows digital spend is better for this case, there are other factors that this model could have included such as seasonal trends.

Limitations of MMM

MMM often focuses on short-term sales impacts, which can undervalue long-term brand-building activities. I personally believe the kind of marketing recommendations that MMM provides should not undermine what is achieved through good, consistent branding. MMM can provide valuable insights and guide resource allocation, but it is essential to integrate it with a broader marketing strategy that takes into account the importance of long-term brand-building efforts. In so doing, a company’s marketing activities could potentially translate into a consistent Money-Making Machine.
—-

If you would like to join our community and contribute articles such as the one above, please follow this link

Find out more about Bill: