Introduction to Multi-Touch Attribution (MTA) and Marketing Mix Modeling (MMM) #26

What is attribution in marketing?

Attribution is the process of assigning value to various touchpoints (points of contact) in a customer’s purchase path. It allows marketers to understand which marketing activities had the greatest impact on conversion.

Multi-Touch Attribution (MTA)

• What it is:
  • MTA is a model that analyzes all the touchpoints in a customer’s path and assigns them a value based on their impact on conversion.
  • Example: A customer sees a Facebook ad, clicks on an email, and then makes a purchase. The MTA shows how each of these touchpoints contributed to the conversion.
• When to apply:
  • When you have access to detailed customer behavior data (e.g., data from Google Analytics, CRM, campaign tracking).
  • Ideal for short buying cycles (e.g., e-commerce, online sales).

Marketing Mix Modeling (MMM)

• What it is:
  • MMM is a model that analyzes aggregate data (e.g., sales, marketing spending) to understand how various factors (e.g., TV advertising, promotions, pricing) affect business performance.
  • Example: MMM can show that a 10% increase in TV advertising spending leads to a 5% increase in sales.
• When to apply:
  • When you are dealing with long buying cycles (e.g., automotive, real estate).
  • When detailed customer behavior data is lacking (e.g., for offline campaigns).

When to use MTA and when to use MMM?

MTA – Short cycles, detailed data

• Advantages:
  • Precise assignment of values to each touchpoint.
  • Ideal for optimizing digital campaigns.
• Limitations:
  • Requires access to detailed data (e.g., tracking cookies, which is becoming more difficult in the privacy era).
  • It does not take into account offline effects (e.g., TV advertising, billboards).

MMM – Long cycles, aggregated data

• Advantages:
  • It takes into account all factors affecting sales (e.g. advertising, pricing, seasonality).
  • Independent of individual customer data, so no cookie tracking required.
• Limitations:
  • Less precise than the MTA – does not show the impact of individual touchpoints.
  • Requires large historical data sets.

When to choose the MTA, and when to choose the MMM?

• MTA:
  • When you have short buying cycles (e.g., e-commerce).
  • When you want to understand which digital channels are most effective.
  • When you have access to detailed data on customer behavior.
• MMM:
  • When you have long buying cycles (e.g., the automotive industry).
  • When you want to understand the impact of offline factors (e.g., TV advertising, in-store promotions).
  • When detailed customer data is missing.

How do you combine MTA and MMM to get a complete picture of campaign effectiveness?

Why combine MTA and MMM?

• The MTA provides detailed information on digital touchpoints, but does not consider offline effects.
• MMM shows the overall impact of marketing activities, but is not precise in assigning value to individual channels.
• The combination of the two models provides a complete picture of the campaign’s effectiveness – both at the micro level (individual touchpoints) and at the macro level (overall impact on sales).

How to do it? Step by step

1. Collect data from both models:
   • Use MTA to analyze digital touchpoints.
   • Use MMM to analyze aggregate data, including offline effects.
2. Compare results:
   • Check whether the MTA and MMM results are consistent. For example, does the MTA show that Facebook advertising is effective, while the MMM confirms that increased digital ad spending brings increased sales.
3. Fill in the gaps:
   • If the MTA does not account for offline effects, use MMM to estimate them.
   • If MMM does not show the impact of individual channels, use MTA to identify them.
4. Create a hybrid attribution model:
   • Combine the data from MTA and MMM to create a comprehensive model that considers both digital touchpoints and offline factors.
   • Example: Use MTA to assign a value to digital touchpoints, and MMM to estimate the impact of TV advertising and in-store promotions.

Tools for combining MTA and MMM

1. Google Attribution 360:
   • An advanced attribution tool that combines data from different channels.
2. Neustar MarketShare:
   • MMM modeling platform that integrates data from various sources.
3. Adobe Analytics:
   • It allows analyzing data from different touchpoints and creating hybrid attribution models.

An unusual analogy: the MTA and MMM like a puzzle and a map

To surprise you, I will use an unusual analogy:

• The MTA is like putting together a jigsaw puzzle – you see each piece (touchpoint) and know how it affects the whole (conversion).
• MMM is like a map – you see the overall picture (impact on sales), but you don’t see the details (individual touchpoints).
• Combining MTA and MMM is like putting a jigsaw puzzle on a map – you have both the details and the overall picture, allowing you to fully understand the effectiveness of the campaign.

Summary

Multi-Touch Attribution (MTA) and Marketing Mix Modeling (MMM) are two complementary attribution models that allow you to understand the effectiveness of campaigns at different levels. MTA is ideal for analyzing digital touchpoints, while MMM works well for analyzing offline effects and long-term trends. Combining the two models provides a comprehensive picture of campaign effectiveness, which is crucial in the era of multichannel marketing.

The technical aspects of implementing Multi-Touch Attribution (MTA) and Marketing Mix Modeling (MMM) is a fascinating topic that requires an understanding of both analytical tools and data integration processes. Let’s discuss it step by step so you can see how these models can be implemented in practice.

 Implementation of Multi-Touch Attribution (MTA).

Step 1: Collect data

To implement an MTA, you need to collect data from all the touchpoints that the customer interacts with.

• Data sources:
  • Website: tools such as Google Analytics, Adobe Analytics.
  • Emails: Marketing systems like Mailchimp, HubSpot.
  • Social Media: Facebook Ads, LinkedIn Ads.
  • Display ads: Google Display Network, programmatic advertising.
  • Call center: data from telephone calls.
  • Mobile apps: SDK for tracking user behavior.
• Data collection tools:
  • Google Tag Manager: To manage tracking tags on a website.
  • Segment: A platform for integrating data from various sources.

Step 2: Data Integration

Data from different sources must be integrated into a single system so that it can be analyzed.

• Integration tools:
  • Customer Data Platform (CDP): Salesforce CDP, Adobe Experience Platform.
  • ETL (Extract, Transform, Load): Tools like Talend, Informatica.
• Integration process:
  1. Extract: Get data from various sources.
  2. Transform: Transform data to a common format.
  3. Load: Load the data into the analytics system.

Step 3: Select an attribution model

There are several attribution models that can be applied to the MTA:

• Last Click: The entire value is assigned to the last touchpoint.
• First Click: The entire value is assigned to the first touchpoint.
• Linear: Value equally distributed among all touchpoints.
• Time Decay: Higher value assigned to touchpoints closer to conversion.
• Position-Based (U-Shaped): 40% of the value assigned to the first and last touchpoint, 20% distributed among the others.
• Attribution tools:
  • Google Attribution 360: An advanced multi-touch attribution tool.
  • Adobe Analytics: offers various attribution models.

Step 4: Analysis and optimization

Once the MTA is implemented, regularly analyze the results and optimize the campaigns.

• Analysis tools:
  • Tableau, Power BI: For data visualization.
  • Google Data Studio: For creating reports.

Implementation of Marketing Mix Modeling (MMM).

Step 1: Collect data

MMM requires aggregated historical data.

• Data sources:
  • Sales: Sales data from various channels.
  • Marketing spending: TV, digital, outdoor advertising budgets.
  • External factors: Seasonality, price changes, competitive actions.
• Data collection tools:
  • ERP systems: SAP, Oracle.
  • Financial systems: QuickBooks, Xero.

Step 2: Prepare the data

Data must be prepared for statistical analysis.

• Data preparation process:
  1. Data cleaning: Remove errors and incomplete records.
  2. Data aggregation: Combine the data into appropriate categories (e.g., monthly ad spending).
  3. Data normalization: Adjust data to a common scale.
• Data preparation tools:
  • Python, R: For cleaning and processing data.
  • Excel: For simpler analyses.

Step 3: Build the model

MMM is based on regression analysis to understand how various factors affect sales.

• Methods of analysis:
  • Linear regression: A simple method to analyze the influence of various factors.
  • Multiple regression: Takes into account multiple variables simultaneously.
  • Machine learning: advanced methods such as random forest, gradient boosting.
• Model building tools:
  • Python: libraries like scikit-learn, statsmodels.
  • R: lm package for linear regression.
  • SAS, SPSS: Professional tools for statistical analysis.

Step 4: Validate the model

Once the model is built, test it to make sure it is accurate.

• Validation methods:
  • Split into training and test collection: 70% of data for training, 30% for testing.
  • Cross-validation: Multiple testing of the model on different subsets of the data.
• Validation tools:
  • Python: scikit-learn library offers functions for cross-validation.
  • R: caret package for model validation.

Step 5: Interpret the results

After validating the model, interpret the results and make business decisions.

• What to analyze:
  • The impact of different marketing channels on sales.
  • Optimal marketing mix (e.g., how much to spend on TV advertising versus digital).
• Tools for interpretation:
  • Tableau, Power BI: For visualization of results.
  • Excel: For simpler analyses.

Merging MTA and MMM – Technical aspects

Step 1: Integrate data with MTA and MMM

To combine the two models, you need to integrate the data from the MTA and MMM into one system.

• Integration tools:
  • Customer Data Platform (CDP): Salesforce CDP, Adobe Experience Platform.
  • ETL (Extract, Transform, Load): Talend, Informatica.
• Integration process:
  1. Extract: Download data from the MTA and MMM.
  2. Transform: Transform data to a common format.
  3. Load: Load the data into the analytics system.

Step 2: Create a hybrid attribution model

Once the data is integrated, you can create a hybrid model that combines the advantages of MTA and MMM.

• Joining methods:
  • Weighted Average: Combine MTA and MMM results, assigning appropriate weights.
  • Machine Learning: use ML algorithms to combine data from both models.
• Tools for creating a hybrid model:
  • Python: libraries like scikit-learn, TensorFlow.
  • R: caret package for advanced analysis.

Step 3: Analysis and optimization

Once you have created a hybrid model, regularly analyze the results and optimize the campaigns.

• Analysis tools:
  • Tableau, Power BI: For data visualization.
  • Google Data Studio: For creating reports.

Summary

Implementing Multi-Touch Attribution (MTA) and Marketing Mix Modeling (MMM) requires an understanding of the technical aspects of data collection, integration and analysis. MTA is ideal for analyzing digital touchpoints, while MMM works well for analyzing offline effects and long-term trends. Combining the two models provides a comprehensive picture of campaign effectiveness, which is crucial in the era of multichannel marketing.