Media Mix Modeling – Using data to measure & predict media channel performance

How much % is each media channel contributing to the brand’s revenue? Which channel is more effective? More worth investing in and which channels are wasting the business’s capital?…. These are the tough questions that every marketer must answer when building a marketing strategy for the business. With Media Mix Modeling, determining the effectiveness of media channels will be specifically explained under intuitive numbers, instead of being perceived and evaluated emotionally, thereby helping the marketing strategy become more effective and optimal.

With Media Mix Modeling, the effectiveness of the media channels will be interpreted specifically under visual numbers, instead of recognizing and evaluating the emotions, thereby helping the marketing strategy become effective, More optimal.

What is Media Mix Modeling?

Media Mix Modeling is a statistical data analysis method that helps businesses measure the effectiveness of each marketing channel in the overall campaign. Media Mix Modeling helps determine the degree of contribution of each channel (TV, Digital, OOH, Radio, Social, …) to Sales or a specific business goal. In particular, Media Mix Modeling will collect and analyze data in marketing campaigns that have been implemented earlier, use regression methods, statistical algorithms and data analysis, thereby offering efficiency Each channel.

This model will show you each factor in marketing mix such as: Product changes, price strategy, distribution strategy and promotion strategies, … will affect a specific target. The business of the business. Specifically, Media Mix Modeling will help businesses answer questions such as:

• Efficiency of each communication channel: Which channel contributes the biggest revenue?
• Interaction between channels: How does the activity between channels promote each other?
• Cost allocation: If the budget cuts for which channel and increase investment in?

The difference of Media Mix Modeling is that it will show you a panoramic picture of your business marketing, thereby helping you make decisions based on data calculation instead of just relying on emotions.

For example, you can use Media Mix Modeling to measure the impact of television ads on the total revenue of the business, based on indicators such as costs, display levels, conversion efficiency but Advertising. Television brought in previous campaigns. Media Mix Modeling will use statistical models, measurement algorithms and analyze these data to help you understand the effectiveness of this advertising channel. At the same time, through data, it can propose scenarios that will occur when you adjust the budget for ads.

Overall, Media Mix Modeling will bring many outstanding benefits to businesses such as:

• Optimizing marketing spending, cutting channels is not effective.
• Budget allocation into channels that have the most positive impact on the goals and reduce the budget into channels does not affect the target much.
• Forecasting and planning: By measuring the impact of variables on the goals of the business, MMM helps the Brand make forecasts for future marketing strategies and the most optimal planning.
• Understanding customer behavior helps businesses understand how different customer segments will react to each communication channel.

Factors to be measured in MMA

Including some data indicators and factors that play a decisive role in the efficiency of Media Mix Modeling, including some of the most important factors that businesses need to measure the line when using this model is :

Input data

• Advertisement: The amount allocated to advertising channels such as Facebook Ads, Google Ads, TV, Radio, Print and Ooh newspapers, …
• Promotional activities: Expenses for promotions, discounts, coupons, refund.
• Price strategy: Including conventional prices, discounts, promotional prices, packaging and price flexibility.
• Distribution channel: Allocate resources to various distribution channels such as e -commerce, online channel, retail, wholesale, direct sales to consumers and sell to intermediaries.
• External factors: External factors such as economic conditions, seasonal factors, weather and industry trends can affect consumer behavior and market development motivation.

Output data

• Sales: Total revenue collected from selling products or services for a specific period of time.
• Market share: The percentage of the total market sales that the company’s products or services is compared to competitors.
• Attract and retain customers: Data related to attracting new customers, retaining existing customers (Retention Rate) and increasing customer loyalty, increasing customer life cycle value (CLV).

The implementation of Media Mix Modeling will start by identifying KPIs that the brand wants to measure. What are the questions that the brand wants to answer specifically? After clarifying these goals, the brand needs to take the following four steps:

Stage 1: Collecting data

In the first stage, the brand needs to collect all historical data on previous marketing activities. Specifically, all data collected from marketing campaigns should be collected, including data on the level of interaction, display level, demographic of target customers and especially levels spending on each ad channel.

In addition, the brand also needs to clarify external factors affecting previous campaigns, such as economic conditions, seasonal factors, social context, market and competitors.

Currently, the collection of data from third -party cookies is gradually being removed. Therefore, the brand should be more focused on first -party data – data collected from sources such as website, CRM, application, Email or customer survey directly. This data helps the brand to better understand customer behavior, such as website access time, conversion efficiency, purchase history, …

In addition, businesses can exploit the second party data from business partners to enrich their database.

Stage 2: modeling

Once you have collected these data, you need to start creating the Media Mix Modeling model. In it, you need to select the variables:

• Dependent variables are the results that businesses need to explain. For example, you want to explain the sales, downloads of the application, the conversion rate.
• Then, identify independent variables that can affect this dependency. For example, spending on advertising, selecting target audiences, advertising time.

Tell the model that includes controlled variables such as prices and distribution channels, as well as uncontrolled variables such as competition and inflation.

Finally, assign value to both dependent and independent variables, and create a mathematical model that represents the relationship between them.

For example:

Dependent variables (y – results needed): sales

Independent variables (x – factors that affect Y):

• X1: Advertising budget on TV
• X2: Advertising budget on digital
• X3: Advertising budget on OOH (outdoor advertising)
• X4: Advertising budget on social
• X5: Seasonal trend
• X6: Product price or promotion
• X7: Macroeconomic factors (for example, GDP, interest rates, inflation)

Mathematical model shows

A basic linear regression model for Media Mix Modeling has the form:

Revenue = n0+n1x1+n2x2+n3x3+n4x4+n5x5+n6x6+n7x7+e

After assigning the value, running the statistical regression algorithm (also known as the machine learning) we have the following equation:

Revenue = 1000+3.5×1+2.8×2+1.9×3+1.5×4+500×5+700×6+200×7

Stage 3: Data analysis and deep understanding

In phase 3, you will use the measured model from phase 2 to start analyzing the effects of factors and communication channels on your marketing campaign. This model will tell you the degree of contribution of each channel to the business results through the relationship between independent and dependent variables. For example, in the above model, we can see:

Revenue = 1000+3.5×1+2.8×2+1.9×3+1.5×4+500×5+700×6+200×7

This model shows:

• TV ADS is the most impact factor for revenue. Each 1 million dong spending on TV ADS increased sales 3.5 million. Meanwhile, the impact of social channels on revenue is much lower.
• The crop can increase sales by 500 million (factor 500)
• Promotion helps increase sales significantly with a coefficient of 700

Therefore, the brand can also use this model to forecast the level of interaction and revenue of Chien in the future. However, note that the forecast is only exactly when the impacts from the external environment do not change too much compared to the context of the previous campaigns.

Stage 4: Optimization

Optimization is the final stage in Media Mix Modeling. Businesses will use results from phase 3 to adjust their marketing elements to improve performance for future campaigns.

Conclusion:

Above is how to operate a basic Media Mix Modeling process including collection, data filtering, regression calculation and analysis based on regression model. In the context of using a series of media channels today, Media Mix Modeling will help the brand to see the actual effect of each channel, thereby Investing in channels that are really effective instead of the system. Exchange on too many channels. The regression model of Media Mix Modeling can also predict the results that a marketing campaign can achieve in the future, helping Marketer make a marketing plan more effectively.