Marketing Mix Modeling (MMM): Background, Scenarios, Solutions, and Case Applications

Analysis Background

Media types and sales channels are constantly evolving, making the customer journey increasingly complex. Optimizing at the level of a single activity can no longer meet client demands. Brands need a complete marketing view at both strategic and tactical levels to optimize efficiency across channels and achieve the highest ROI. For example, offline advertising, promotional events, private‑traffic operations, pricing, and macro‑economic conditions all affect overall sales. Because marketing channels are becoming more diversified, a method is needed to explain the composition of sales/marketing, quantify each influencing factor, and determine how much marketing contributes to sales at a macro level. This includes quantifying each factor’s contribution and allocating budgets rationally—key strategic problems.

Analysis Scenarios

The Marketing Mix Model (MMM) helps brands understand the marketing mix at a macro level and supports budget allocation. Its analysis scenarios include three aspects:

Help the business understand driving factors and resulting outcomes. For example, which marketing channel brings the most sales and by what percentage? Driving factors include marketing, pricing, competitor activities, etc. MMM can allocate sales to these factors.

Optimize marketing spend allocation. As a macro‑analysis tool covering both online and offline channels, MMM can guide conditional decisions on well‑performing or under‑performing channels.

Simulate marketing performance. MMM enables strategic simulation by adjusting factor values (e.g., lowering price or increasing ad spend) to see the impact on sales.

Solution

Solution 1: Marketing Mix Model

What is a Marketing Mix Model?

Marketing Mix Modeling (MMM) is a statistical modeling technique (e.g., regression) that measures and predicts the impact of different marketing investments on sales, helping understand overall marketing effectiveness and decide optimal budget allocation across channels.

MMM originated from econometrics, first applied in fast‑moving consumer goods, and later spread from academia to the market. The “Mix” originally referred to the 4Ps (Product, Price, Place, Promotion). The goal was to find the optimal 4P combination and predict each activity’s sales impact.

With digital marketing, inputs now include many more variables beyond the traditional 4Ps, such as digital ads, TV, radio, email, etc.

Product data : basic attributes of the brand’s products.

Competing data : basic attributes of competitor products.

Economic data : macro‑economic indicators representing the product’s environment.

Marketing data : metrics of various advertising channels.

Conversion data : sales amount, number of purchases, etc.

What can MMM do?

Measure marketing factors : better understand the relationship between channels and sales.

Budget allocation : identify high‑ROI and low‑ROI channels to optimize spend.

Simulation & prediction : forecast future conversions given channel inputs.

Basic Steps of MMM

The basic steps are as follows:

Step 1: Data Preparation

MMM requires various data types, granularity, and organization.

The model uses five major data categories: product, competitor, macro‑economic, marketing, and conversion data. An example brand’s data includes:

Data granularity is usually weekly or monthly.

The model needs a long time horizon (over one year) to ensure enough training samples.

Step 2: Data Processing

Missing‑value handling

Imputation: fill missing values with mean, median, etc.

Prediction: use time‑series forecasting to estimate missing points.

Zero‑fill: set to zero when no transaction occurred on a day.

Deletion: remove rows/columns with excessive missingness.

Data Transformation

Some factors (e.g., digital ads, TV ads) have non‑linear relationships with sales. After a certain saturation point, additional spend yields diminishing returns. Therefore, transformations are applied to these factors:

Adstock (lag) effect: the impact at time t combines current spend and the decayed effect from previous periods.

Saturation / diminishing‑return effect: modeled with an S‑curve to capture the initial low impact, rapid increase after a threshold, and eventual plateau.

Step 3: Model Training

To keep factor contributions additive, MMM typically uses multivariate linear regression (or transformed linear models). Independent variables include price, ad spend, macro‑economics, promotions, etc.; the dependent variable is sales (or market share). After applying transformations, the model quantifies the marginal contribution of each input.

Step 4: Analysis & Decision

MMM results support several decisions:

Explain overall sales composition : quantify each factor’s share (e.g., digital ads 40%, discounts 20%).

Simulate & predict sales changes : estimate sales impact of increasing ad spend by 10% (e.g., sales +8%). Simulations must consider realistic joint factor changes.

Solution 2: Fusion of MMM and Attribution Modeling

Differences between MMM and Attribution Models

Attribution focuses on optimal digital‑channel mix; MMM covers digital, traditional media, economic factors, competitors, etc.

MMM is a macro model using aggregated (weekly/monthly) data; attribution is a micro model tracking user‑level (second‑by‑second) behavior.

How to choose?

If a brand invests in both online and offline channels for >60 days and wants a full purchase‑journey view, use MMM.

If the brand only uses digital channels, use an attribution model.

MMM + MTA Fusion Model

Because ad data can be sparse, a classic approach is to first run a traditional MMM with digital ads aggregated as one factor, then apply Multi‑Touch Attribution (MTA) to split that factor’s contribution to individual channels or audiences.

Case Applications

Explain Sales Composition

MMM can reveal each factor’s contribution to total sales (e.g., baseline product sales 30%, digital ads 50%, discounts 20%, competitor discounts negative).

Marketing Simulation & Recommendations

MMM can simulate outcomes: a 10% increase in digital ad spend yields a 10% sales lift; a 20% increase yields a 16% lift, showing saturation. Negative factors can also be simulated (e.g., a 10% rise in competitor discount reduces sales by 2%).

About Us

Alibaba Mama SDS (Strategic Data Solutions) team uses data to make growth strategies for merchants and platforms more scientific and effective. We provide marketing insights, strategies, value quantification, and attribution services for all Alibaba Mama advertising clients.

Resume submission email: [email protected]