Marketing Mix Modeling (MMM) Tutorial in Google Spreadsheets

Marketing Mix Modeling (MMM) is a statistical technique that provides insight into the sales contribution of each ad network, helping to optimize advertising spend. It has gained popularity due to changes in digital tracking methods, such as iOS14, which have made traditional methods less reliable. MMM offers a more privacy-friendly approach and is increasingly used by both Fortune 500 companies and smaller startups to understand the incremental performance of each marketing channel.

To use the Recast MMM template in Google Spreadsheets:

1. Make a copy of the Google sheet: Click the provided link and create a copy of the template.
2. Log into the Tenjin dashboard (for Tenjin customers): Export your data using the 'Data Exporter' feature. For non-Tenjin customers, export your daily spend and revenue data from your analytics tool.
3. Copy and paste data into the RAW tab: Ensure your data is in tabular format with one column per variable and one row per week.
4. Build your model: Choose variables, carryover, and saturation rates that provide the best fit with the data. Use the LINEST function and other metrics like MAPE and NRMSE to measure accuracy.
5. Visualize and take action: Use the MODEL page to visualize accuracy and make decisions based on the model's insights. Use the PLAN tab to forecast future spending allocations.

The limitations of using a simple linear regression model for MMM include:

• Assumption of constant performance: The model assumes that marketing performance does not improve or decline over time, which may not be accurate.
• Spurious results: The model may show misleading results, such as some channels driving negative sales.
• Incomplete data: You may not have all the necessary data to build a valid, accurate model, especially if data collection is incomplete or if performance is driven by hard-to-measure variables.
• Risk of incorrect decisions: It's easy to build a wrong model and make decisions that could lose money. Always consult with a statistician or business analyst, or learn more about MMM before making important decisions.

The difference between Marketing Mix Modeling (MMM) and Multi-Touch Attribution (MTA) lies in the data requirements and methodology:

• MMM: Uses aggregate data and matches spikes and dips in sales to marketing events or actions. It estimates the incrementality of each marketing channel without requiring user-level data.
• MTA: Requires user-level data to deterministically assign attribution to users who saw or clicked on ads. It provides a more granular view of user interactions but is less privacy-friendly compared to MMM.