Media mix modeling is the tried-and-true practice of analyzing sales and marketing data to make sure you’re spending your advertising dollars wisely. While much has changed since media mix model first emerged, it’s as important as ever to carefully consider the mix of media you use to reach your audience efficiently — particularly as the number of channels available to advertisers keeps growing.
Media mix modeling offers a way to view and analyze data from a variety of channels and understand where prospects are consuming your content. It will also tell you what information they’re looking for on which channels, so you can make changes to your plan accordingly.
Advertisers and marketers using media mix modeling can get a high-level look at channel effectiveness, both current and historical, to make predictions and decisions about future strategy. Media mix modeling can also incorporate big-picture trends like economic changes, channel popularity growth or decline, and seasonal shifts. Finally, the media mix model allows marketers to make connections between their efforts across channels.
What Is Media Mix Modeling?
Media mix modeling is a statistical analysis used by marketers, which is why it’s also referred to as marketing mix modeling. It uses aggregate data to show how different elements of the marketing strategy — digital and traditional — are contributing to the goals or KPIs that marketing is working toward. These might be leads, conversions, or other common marketing metrics, with the ultimate goal of contributing to the company’s sales numbers.
The media mix marketing ratio is calculated by taking into account the marketing channels in use, the money being spent on each channel, and campaign results and insights (more on that below). Marketers use media mix modeling alongside other tools to see trends or patterns. But though it’s useful to employ media mix marketing as part of your toolset, it’s not usually the primary way to gauge and update your marketing strategy, since it doesn’t always reflect factors like seasonality, competitors, and the overlap between channels.
For businesses that have been spending their budget on multiple advertising channels, such as paid social, paid search, or print ads, media mix marketing can shine a light on what’s really succeeding.
How Media Mix Modeling Works
The data science technique behind media mix modeling is a multi-linear regression model, meaning it gauges the relationship between a dependent variable like sales and an independent variable like ad spend across channels.
There are different options to use based on what you’re measuring: If you’re using or getting started with media mix modeling, an important first step is knowing which data you want to measure and then figuring out how you can access that data, whether it’s gathered internally or found in outside sources. The data has to be high quality, too, and ideally you’ll have a few years’ worth. This data might include points like:
- Marketing spend from the past few years, by quarter or year.
- Conversions.
- Site traffic.
- Macroeconomic trends.
- Competitor activity.
You should also define your goals to ensure good output from the model, whether you’re focused on increased sales or brand awareness. Then, you perform media mix modeling — whether you’ve built the model in-house or purchased a tool — by assigning a numerical value to the impact of marketing activities across channels. The value you assign should reflect how those marketing activities are contributing to their ultimate goal, whether it’s top-of-funnel awareness or down-funnel sales leads. Media mix modeling analysis can show how sales trends over time have correlated to marketing activities.
Media mix modeling does not measure an individual customer’s journey, since it doesn’t incorporate specific user data like clicks or impressions. It’s a more holistic way to understand channel and ad spend effectiveness. You’ll likely need to do some testing and study the results carefully to see if they make sense, before employing media mix marketing broadly or making recommendations or decisions based on it.
Key Metrics Analyzed in Media Mix Modeling
Modern marketers typically have a wide range of channels and metrics to look at when performing media mix marketing. What’s most insightful will depend on your industry, company, and short- and long-term marketing and sales goals. These are some common metrics to analyze with the media mix model.
Return on Investment (ROI)
ROI might be considered the primary measurement of media mix modeling, since its holistic nature shows where you might be spending too much or too little money across channels. The model can help you gauge the ROI of each channel you’re using (ROI is calculated by subtracting how much you’ve spent from the revenue generated, then dividing by what you spent). That percentage value is the return on investment. It can often reveal surprising details about which channels are performing best based on ad spending.
Sales Uplift
An increase in sales is the gold standard for determining whether a new technique or campaign is working for any business. To understand sales uplift when using media mix model, you’ll want to compare sales during a period of activity, such as a campaign, against a baseline period without that activity. Media mix marketing can then isolate the impact of various channels within that data.
Customer Acquisition
This metric is useful for media mix model because it shows how various channels are working together to influence customer behavior. You might see that one channel is attracting more net-new customers, while another is bringing in existing customers or converting prospects into customers more readily. That’s an opportunity to reconsider your channel spend for a more impactful customer acquisition strategy.
Importance of Optimizing Ad Spend
Optimizing your marketing spending these days is essential. With so many channel options and ways to spend budget, constant monitoring and measurement will help you optimize ad spend and get the most return on investment. Making decisions based on those measurements, without having to track individual user actions, can lead to better outcomes.
Media mix marketing’s role in optimizing ad spend is salient in a digital era. It can reveal where spending makes an impact — or doesn’t. There are situations where spending more on a particular channel won’t boost your sales, such as if you’ve tapped out your sponsored search terms or already reached your niche target audience. It can be difficult to know when you’ve saturated a channel. The point of diminishing returns isn’t one a marketer likes to reach, but knowing that information can be transformative for future planning.
Conversely, media mix model can show the untapped areas of ad spend in a marketing budget. The new social channel you tested might be an opportunity as it grows and your audience there increases well. Or, a traditional medium like print advertising or a local radio ad might be more worth your time than you expected, which you can see easily using media mix marketing.
Media mix model also helps you see a deeper view than just last-click attribution, so that channels that were visited further up the funnel can be given their proper due in the broader marketing strategy.
5 Key Elements Within Media Mix Modeling
There isn’t a one-size-fits-all approach to media mix modeling when you consider the busy digital marketing landscape, range of channels, and demand on users’ attention today. But there are some key elements you should include as you’re creating your business’ media mix model.
Base Sales
This number is a company’s sales without marketing efforts added on top. Base sales can come from factors like market trends and brand reputation.
Incremental Sales
This figure reflects sales that are attributed to marketing activities, and can be more easily isolated when using media mix modeling.
Marketing Channels
This piece of the puzzle captures all the platforms that a marketing team uses in its activities, and may include search, social, email, earned media, website, blog, and more.
Product or Pricing Changes
Add your company’s product releases or updates and any pricing changes in the media mix model to understand how they affected sales.
Outside Influences
You know your company’s vertical or industry best. Any specifics should be reflected in your media mix modeling. These might include seasonal events, special offers, weather or other events that affected shopping patterns, and data related to the larger economy, such as unemployment rates, inflation, and more. Competitor promotions, pricing, products, and other information can also be included here.
Calculating the Media Mix Model Ratio
The media mix model can be built with a language like Python, using multiple linear regression to create an equation that assigns ratios to each marketing channel. There isn’t one model or template for media mix model — rather, a data scientist creates variables out of different marketing efforts to factor into the model. This might include how much the business is spending on ads, how often promotions are running, or how often pricing or products are changing.
After this, the modeler will go through several iterations to create a model that explains sales trends. This will involve deep knowledge of the business, expertise in modeling methodologies, and reliable, clean data in the correct format. A business can either ask their data science team to build a media mix model, or hire a data scientist to do so, or pay for a platform or tool that includes media mix model as part of its offering.
Each database or spreadsheet row indicates a set period of time (day, week, month) and each column is one of the model’s independent variables, like ad spend. The dependent variables are things like sales volume or value.
Briefly, these are the four steps to create a media mix model:
- Define the scope. What’s the goal of creating this model? What would the business like to do better, or achieve more of? Start with the purpose to get to the right model.
- Gather the data. A media mix model will not succeed if it doesn’t include the right data. Work with colleagues to gather all the relevant data and make sure it’s accurate.
- Model the data. This step requires algorithm refinement to connect data points, and can take time and patience through multiple iterations to get to a reliable model.
- Analyze and take action. Take a deep dive into initial metrics, outputs, and recommendations so you can start making adjustments based on your goals in step 1.
Media mix modeling in its modern form might use automation rather than traditional manual models built by hand by econometricians. Fans of automated tools prefer them because they eliminate human bias and deliver results faster from more data, while manual modelers say human expertise is still an essential part of a good model, and that automation can miss key elements or make incorrect correlations. Ultimately, which you choose will be up to you.
The Difference Between Media Mix Model and Data-Driven Attribution
Media mix modeling measures how marketing efforts are meeting business objectives. As mentioned before, media mix model doesn’t use any customer journey or user-level data in its calculations. Instead, it offers high-level insights into marketing tactics over a long period of time, perhaps several years, to capture trends.
Media mix model can’t be used by itself when measuring and planning marketing strategies, though. Data-driven attribution is a type of modeling that tracks engagements through a customer journey, so marketers can understand the tactics that work best from the top to the bottom of the funnel. Data-driven attribution models typically gauge performance a few months after a marketing activity concludes.
Neither of these models is used alone. Data-driven attribution doesn’t always account for offline activity and is mostly focused on digital marketing. Media mix model measures both of these, but its lack of user-level insight isn’t helpful when marketers are customizing campaigns or content.
Emerging Trends in Media Mix Modeling
Media mix modeling has been around for 50 years or more, and like most practices in marketing, it’s always evolving. With the influx of more accessible data science and data analytics technology, there are a few important trends to be aware of in media mix model.
Machine Learning and Advanced Data Analytics
Media mix modelers are incorporating machine learning models and statistical techniques to get better, more nuanced insights and make more accurate predictions for marketing strategies.
Model Accuracy
Media mix model has always included model refinement, but it’s easier now than ever to use technology wisely, refining and updating models on an ongoing basis to keep up with market conditions and buyer behavior.
Platform Integration
It’s possible for marketers to incorporate media mix model with other marketing performance platforms to easily grasp trends and make smart decisions. Marketers are also using incremental lift tests — which gauge the impact of marketing efforts on sales — to validate the results of a model. This helps combine the correlative nature of media mix model with causal impacts of media on sales.
Advantages of Media Mix Modeling
Privacy
One key advantage of media mix modeling in an era of privacy concerns and post-cookie websites is that it doesn’t track individuals across the web. Media mix model uses statistics on aggregate data instead.
Speed and Data-Reliance
It’s easy for marketers to go on gut instinct, but tools like media mix model and data science in general are able to perform analysis very quickly, whether you have an in-house data science team or use a media mix model tool, and serve as unbiased sources for marketing predictions and suggestions.
Future-Simulation
Media mix marketing offers a way to perform simulations of ways the market might change, or how your marketing efforts may lead to varying results. Take advantage of media mix model to plan different future-state scenarios, particularly if budgets are tightening and you need to spend even more carefully.
Holistic Approach
Unlike many marketing performance tools, media mix model lets you evaluate multi-channel campaigns to understand how channels work together. It’s a big-picture view, rather than a list of separate dashboards. It also includes hard-to-measure numbers like diminishing returns and adstocks, which measures the longer-term results from an ad.
Media Mix Modeling Limitations
Media mix modeling is generally used in conjunction with other marketing measurement options. With limited consumer attention spans, media mix modeling isn’t built to customize messaging to target individual consumers — a major way to reach audiences today. It’s important to recognize a few common limitations when you’re employing media mix model:
- Reporting results requires several years of historical marketing data.
- Data isn’t as granular as other marketing metrics.
- Brand/messaging insight isn’t available.
- Customer experience not reflected.
- Speed isn’t adequate for quick campaign adjustments or market disruptions.
The Challenge of Correlation in Media Mix Model
Media mix marketing is not always the best performer if there is a lack of variation in ad spend, or too much correlation between factors: If two ads run at the same time, for example, or two ads run on similar social channels concurrently, the model might not be able to tell which one is driving sales. That leads to unstable estimates in the model, which leads to unreliable recommendations for marketers.
In addition, models can’t always account for long-term effects, such as a TV ad that brings returns for many years beyond its airing. It’s important to account for these, but they can be difficult for media mix models to capture.
Best Practices for Implementing media Mix Model
Media mix modeling in the digital era should happen in tandem with other marketing metrics and data gathering. Ideally, marketers can see both the big picture with modeling as well as user-level numbers in a unified measurement platform.
To develop an accurate media mix model (or find the right provider), remember first that you have to trust it. Quality data and quality modeling are essential for you to get the right insights and then use them wisely. Done incorrectly, media mix model can show incorrect marketing attribution information, which will then skew your ad spending decisions.
Make sure you pay attention to these best practices when implementing media mix model:
- Understand the diagnostics and metrics that are part of the model to make sure it’s the right one for your data.
- Check the model’s outputs against actual outcomes. You might compare a forecast against what actually happened to make sure the model is accurate. Double-check any predictions that seem improbable.
- Aim for simplicity and practicality. Especially if you’re new to using media mix model, look for a model that offers accuracy but is also easy to understand.
- Test and learn. As with marketing more broadly, see if you can make changes to your marketing mix in small ways to then compare to the model’s predictions. From there, you can recalibrate if needed.
Case Studies and Real-World Examples
Banco Sabadell Mortgage (JellyFish
Jellyfish was looking to increase the volume of high-quality traffic to Banco Sabadell landing pages to increase the number of leads generated. The campaign as a whole included a robust media mix. With the inclusion of Taboola, Jellyfish could serve landing pages in new environments with guaranteed viewability, and drove 17% more conversions for Banco Dabadell.
Renault Australia
Renault wanted to drive potential car buyers to Renault’s website at scale and retarget engaged audiences to increase conversions. Using Taboola in their marketing mix, the car manufacturer achieved their marketing goal with a 51% lower CPA and a 16% lower CPC.
Look After My Bills
Look After My Bills was looking to reach more United Kingdom (UK) residents missing out on savings for their utility bills. They wanted to diversify their media mix beyond Facebook to the open web, in order to reach high-quality audiences on premium publisher sites. Using Taboola’s automated bidding feature designed to maximize conversions, Look After My Bills finds 10,000 potential customers on the Open Web, while reducing their CPA by 60%.
Key Takeaways
Media mix modeling is a strong part of a marketer’s arsenal when considering ad spend in a holistic environment. It’s especially useful in a fragmented, largely digital media landscape, where consumers are bombarded by information, and where budgets are tight. With media mix model, you can analyze data across channels to understand what’s working and what isn’t, then recalibrate your ad spend accordingly.
Frequently Asked Questions (FAQs)
What is the cost of media mix modeling?
Your media mix modeling cost will vary widely depending on a few things: Model complexity, data quality and availability, and internal vs. consultant expertise. A lower-end media mix model, built in Excel, might cost $500 to $5,000. A mid-range model built with dedicated software might run from $5,000 to $20,000. An advanced media mix model with complex data analysis, custom-built, could cost a business $20,000 or more.
What is the difference between MTA and MSA?
In the world of media mix modeling, MTA is multi-touch attribution, a way to gauge a customer’s journey by the various marketing touchpoints they encounter. Each touchpoint gets some amount of credit so marketers can understand which touchpoints are most valuable along the way to a conversion or other desired action. MSA is measurement system analysis, which isn’t a common term within media mix modeling — it’s a statistical way to gauge how accurate data collection is within a specific measurement system.
