Template
E-Commerce Business Overview
Combine your Shopify, GA4, Google Ads, and Facebook data for insights across your business.
.svg)
Marketing Mix Modeling (MMM) involves analyzing marketing performance data, underlining internal and external factors that influenced the results, and using that information to upgrade a company's marketing strategy or "model."
One of the biggest challenges of running a business is determining how individual marketing channels actually impact your bottom line.
That's why we have techniques like attribution modeling, which identifies the pivotal touchpoints in the customer journey.
But what if you want to analyze your offline marketing initiatives? More importantly, how can you use your analysis to maximize the potential of future marketing plans?
This is where Marketing Mix Modeling comes in.
Marketing Mix Modeling (MMM) involves analyzing marketing performance data, underlining internal and external factors that influenced the results, and using all that information to upgrade a company's marketing strategy or "model."
MMM requires a thorough, intensive look into several variables and marketing channels — be it search, email, social media, advertising, or even offline media. It also aims to measure how those individual channels contributed to specific business outcomes, which may pertain to sales, lead generation, market share growth, and so on.
MMM is sometimes used interchangeably with "Media Mix Modeling." The slight difference is, Media Mix Modeling is used primarily to optimize the allocation of resources across marketing channels.
Before we dig deeper into MMM, let's take a look at the main reasons why it's important:
MMM sounds important, but isn't it just a fancy word for "attribution modeling"?
Yes — both MMM and attribution modeling allow you to identify the individual marketing touchpoints that eventually led to sales. But, unlike attribution modeling, MMM also serves as a predictive model that can guide your future marketing plans.
MMM also has a broader scope, dissecting the combined impact of traditional and digital marketing activities on business outcomes.
To fully grasp how MMM really works, you must first understand the following concepts:
Multiple Linear Regression, or just "Multi-Linear Regression," is a statistical technique that calculates or predicts the "dependent variable" based on "independent variables."
A dependent variable is the predicted outcome, which can represent a marketing Key Performance Indicator (KPI) such as market share or sales.
On the other hand, independent variables are anything and everything that can affect your dependent variable. Some examples are:
The relationship between dependent and independent variables can be described using this formula:
In regression analysis, the constant variable represents an unchanging value. This serves as a benchmark for measuring the effects of independent variables on the predicted variable.
A few examples of constant variables are the baseline expected sales, fixed marketing expenses, customer demographics, and seasonal market trends.
To complete your own marketing mix modeling equation, you need to learn about the different types of independent variables that can impact your marketing performance.
Let's start with internal variables, which are activities and decisions within your company. Some examples are your pricing strategy, product distribution, marketing budgets, and product launches.
External variables are the exact opposite of internal variables. They pertain to events outside your company's control, including natural disasters, unemployment rates, economic situations, weather data, and so on.
While you can't directly influence external variables, you should still track their effect on marketing results.
This allows you to accurately predict how other variables contribute to your business goals. More importantly, they allow you to build contingency plans to cushion the effects of negative external factors.
Just like external variables, you also have no direct control over seasonal or calendar-based variables. However, they're much easier to predict than external events — making it easier to plan adjustments to your marketing campaigns.
Sales holidays like "Cyber Mondays" and "Black Fridays" are a few typical examples of calendar-based variables.
Certain times or seasons of the year also have cyclical effects on business outcomes. For example, scarves, gloves, and pajamas are more in demand during the winter, translating to increased sales from the start of the "ber" months.
Take note that, when dealing with non-numerical variables (i.e., Cyber Monday, natural disaster, location, and job title), you need to create what's called as "dummy variables."
These variables only have two possible values: 0 and 1 (yes and no).
The number of dummy variables you need is always one less than the number of known categories you want (n-1). That's because, if all the created dummy variables are false, the remaining variable is automatically true.
For example, if you need to conduct a regression analysis of how the four seasons affect sales, you need to create three dummy variables.
Creating dummy variables to assign a numerical value to non-numerical data lets you incorporate them into your regression analysis. In turn, you can analyze the performance impact of these variables and visualize them for MMM.
Of course, active marketing and advertising campaigns play a profound role in performance.
Below are some examples of these independent variables:
When using marketing activities in MMM, the costs of these channels are often used as inputs. This allows companies to deduce and predict whether adjusting the budget for a specific channel will lead to better results.
To put things into perspective, let's take a look at a few example scenarios involving MMM.
Let's say your business sells cold refreshments and would like to quantify how the weather impacts sales.
In this scenario, your dependent variable would be sales, while a must-have external independent variable would be the weather data. Just remember that the timeframe for both variables should be aligned (if you're tracking sales over 120 days, you need 120 days' worth of weather data).
Understanding the correlation between calendar-based variables and sales performance lets you optimize your marketing budget and schedule marketing activities accordingly.
MMM can establish the correlation between your dependent variable and all the tiny things you can change within your marketing strategy.
Using regression analysis, you can measure the linear relationship of factors like advertising spend and email marketing budget with sales performance.
You can also identify variables with diminishing returns. For example, using a line chart, you can measure incremental sales alongside your cumulative TV ad spend — identifying the "point of maximum yield" (where additional investment stops generating increasing returns).
Measuring the effects of independent variables on sales while identifying channels with diminishing returns will help you allocate your budget more efficiently. For example, if you're overspending on TV ads (without increasing returns), you can shift some of that budget into another channel that has a linear relationship with sales.
Pricing changes not only have a direct effect on your profit per transaction. It can also change your target consumer group and the perceived value of your product.
MMM can unveil the relationship between product pricing adjustments and the number of transactions you close. It can help you estimate or predict the maximum price point you can use without affecting transaction volume.
Inspired by the Marketing Mix Modeling example above?
Before you work on your first model, here are five crucial tips to make sure you get it right:
You can't do MMM without defining a clear marketing goal first. This allows you to decide the KPIs and metrics to incorporate into your model, starting of course with the dependent variable.
Let's consider something basic, like increasing sales. Your KPI for the dependent variable could be your base sales (transactions generated through brand equity — or sales you make without spending money on marketing) and incremental sales (transactions generated through marketing activities).
You'll then be able to set your constant variable and independent variables, including your marketing and customer acquisition channels like social media marketing, paid advertising, and email marketing. Track the budget you allocated in these channels over the past 12 months.
Don't forget about external and seasonal variables for a more robust model. When it comes to sales, some external factors to consider are sales holidays, seasons, and even weather data.
Once you identify the information you need for MMM, it's time to implement data collection methods for your dependent and independent variables. Depending on your marketing objective, this may require the use of tools like ecommerce analytics and advertising platform reporting.
If possible, target the most data possible by maximizing the date range.
For example, if you have three years' worth of marketing data, don't settle for the reports from only the last one or two years. Remember, the more data you use, the more reliable and accurate your model will be.
Before creating data visualizations, remember to clean and transform your data for uniformity. Keep the fundamentals of data analysis in mind, like cleaning up duplicate data, using data imputation (for missing data), and standardizing data formats.
When ready, you can start your regression analysis and visualize the results. This will help you determine the linear relationship between your marketing objective and independent variables.
There are several tools you can use to visualize correlation charts. Polymer, for example, lets you build interactive correlation charts or scatter plots with trend lines with just a few clicks.
By analyzing the relationship between objectives and independent variables, you can use regression analysis to develop your marketing mix model.
Using multi-linear regression analysis, you can determine the required values for independent variables to achieve a target goal or dependent variable.
To simplify, let's say you make 1,000 units of sales if you have a value of 200 for the independent variable "ad spend." Using your statistical model, you can estimate the required ad spend to achieve the target of 2,000 or 3,000 sales.
The final step is to implement your model and track its real-world performance.
Keep on measuring your goal KPIs, like return on ad spend or number of customers, to validate the accuracy of your model. A Business Intelligence (BI) solution like Polymer can come in handy in consolidating your datasets and visualizing the results for analysis.
Polymer is a versatile BI and dashboarding platform that can help MMM in multiple ways.
For one, it's equipped with all the data management tools you need to clean and standardize your marketing data.
You can build your data repository easily through data integrations with sources like Google Sheets, Facebook Ads, Stripe, and Shopify. Then, use the data manager to standardize your data's headers or create custom metrics to automate KPI calculations.
To instantly visualize the relationship between your variables, use the drag-and-drop editor to create scatter plots or correlation charts.
Just select the metric you want for the Y-axis (sales, revenue, etc.) and the independent variable (marketing channel spend).
Ready to give MMM a shot?
With Polymer, you'll have all the tools you need to skip most of the complexities and jump straight to data analysis.
Click here to try all of Polymer's data visualization and dashboarding features for free.
See for yourself how fast and easy it is to uncover profitable insights hidden in your data. Get started today, free for 7 days.
Try Polymer For Free
.svg)
