Media Mix Modeling (MMM) Simplified!

With the ever-expanding digital landscape and increasing customer-centric data, are you making the best use of it? Most companies are not. While many are still relying on outdated or more simplified methods to determine their next marketing strategy, it has become increasingly difficult to determine a cost-efficient marketing channel. Media Mix Modeling helps businesses not just identify the best channel to acquire customers but helps keep an eye on the overall ROI or ROMI.

What is Media Mix Modeling (MMM)?

Media Mix Modeling or sometimes referred to as Marketing Mix Modeling (MMM) is an analysis technique to measure and compare the impact of your marketing campaigns to your goals, which is often to drive cost-effective conversions.

Still not clear! Let’s understand with an example:

Let’s assume, you’re operating a shoe manufacturing and distribution business operating both physical outlets and online website sales. In the past, you have used Google Ads for display and search advertisements, Facebook, and Instagram ads. Now how will you determine which channel is the best fit for your brand and when to invest where? This is where MMM comes into play.

Media Mix Modeling takes into account a wider range of channels, both traditional and digital while allowing marketers to factor in external variables like seasonality, promotions, market condition, etc. using statistical models to get the most robust and accurate visibility into impact analysis.

How does Media Mix Modeling (MMM) work?

MMM uses historical data and a mix of statistical models to provide you with the most accurate visibility into how changes to marketing spending will impact your conversions while considering various external variables that are correlated.

Multi-linear regression analysis is used in this model to determine the relationship between the dependent variable such as sales and the independent variables such as ad spending across channels

Media Mix Modeling only measures the impact of marketing efforts on marketing objectives without factoring in the user engagement touchpoints at various stages in the marketing funnel.

Data-driven attribution or Multi-touch attribution models are commonly used to measure impact at a user level and determine key touchpoints that are more likely to get the buyer closer to the purchase stage.

There are various stages of Media Mix Modeling which can be better explained using a simple workflow.

1. Business Understanding

Knowing where to begin helps solve 30% of any problem. Identify who are the key stakeholders, what are the data sources, scope, and objective of the study. Organizing a kick-off meeting to discuss these in detail will help you plan for the next stages.

2. Data Understanding

• Do you have the required data for the analysis?
• Where is the data stored and how can you or the team access the data?
• Is your data sanitized?
• What help would you require to capture any additional data or access the data?
• Has there been any initial market analysis?
• Setup project folder and calendar
• Identify the teams/parties involved in the process

3. Data Preparation

• Identify different variables – Dependent and Independent variables
• Independent variables:
  • Media: Offline and Online
  • Price
  • Promotions
  • Display & features
  • Marketing activities
  • Logistics and distribution
  • Others: Competitors, weather, seasonal, trends, etc.
• Processing of Data
  • Split variables
  • Create decays
  • Identify lags
• Confirm the accuracy of the data, hypotheses, and objectives.

4. Modeling and Evaluation

• Build models automatically and then refine them manually
• Evaluate the models, both statistically and commercially
• Finalize the model
• Calculate contribution and ROI
• Create Aggregated Contribution and waterfall (YoY) charts
• Extract results
• Setup progress meeting to show initial results and refine the hypothesis

5. Deployment

• Setup final results and recommendation meeting
• Provide Optimization and Forecasting insights
• Perform all the necessary follow-ups
• Explain how to use the results
• Prepare a detailed summary report for a clear understanding

Conclusion

While Media Mix Marketing (MMM) uses historical data (preferably more than 3 years) and uses probability to determine which channel will work best with predicted changes in variables, its success is completely dependent on the quality of data. An alternate approach to understanding marketing mix impact and contribution is continuous incrementality testing.