Marketing mix modeling data: Building a solid data foundation for your mmm

Marketing mix modeling (MMM) seems like the next big and shiny thing in marketing. Even though the technology has been around for decades, it has become more relevant than ever since the movement toward privacy-first marketing.

Besides that, MMM is a fantastic marketing measurement method that can tell you the true impacts of your marketing campaigns and investments. As a Solution Engineer at Supermetrics, I’ve seen many clients struggling to get their model right, and most of the time, it’s due to the lack of a solid data foundation. So, let’s address that.

Navigate this post:

• What data is required for a good marketing mix model?
• 5 ways to use Supermetrics to build a solid data foundation
• Tips to overcome data constraints

What data is required for a good marketing mix model?

A strong data foundation is crucial to getting accurate and actionable results from your MMM. The quality and completeness of your data directly impact the model’s ability to isolate the true impact of marketing activities, control for external factors, and provide accurate insights for optimization and forecasting. Ever heard of “garbage in, garbage out?”

And since MMM models can vary greatly between different industries, you’ll also want to choose your input accordingly. The data I’m suggesting are more like examples, but when it comes to the actual data collection, always think about what makes sense for your first MMM experiment. For example, do you want to do digital data first and then, later on, incorporate offline data? 

Paid media data

Data on paid media channels—search, social, display, TV, radio, print—quantifies each channel’s investment and reach. This includes impressions, clicks, conversions, and spend, and ideally, granular data like time of day, targeting criteria, and creative versions.

Why it matters: Accurate paid media data allows the MMM to isolate the impact of each channel on sales. Without it, the model can’t accurately attribute sales lift to specific marketing activities, making budget allocation optimization impossible.

Sales data

Sales data is the dependent variable in the MMM—the outcome you’re trying to understand and influence. This data needs to be granular (daily or weekly is preferred), accurate, and complete. A mixed online/offline business should encompass both ecommerce and brick-and-mortar sales.

Why it matters: Inaccurate or incomplete sales data skews the entire model. If sales aren’t correctly recorded, the MMM can’t accurately link marketing activities to business outcomes.

Weather data

Weather significantly influences consumer behavior, impacting both online and in-store sales.

Why it matters: Including weather data allows the MMM model to control for this external factor and prevents misattribution of sales fluctuations caused by weather to marketing activities.

Holidays and seasonal data

Holidays, promotional events, and seasonality drastically impact sales.

Why it matters: Similar to weather data, seasonal campaign data helps the MMM avoid misattribution and isolate the true impact of marketing efforts.

Competitor ad spend data

Competitor ad spend data provides valuable context, as increased competitor activity can impact your sales regardless of your own marketing efforts.

Why it matters: Including this data, though often difficult to obtain, allows the MMM to account for competitive pressure and gain more accurate insights into your marketing effectiveness.

Census/demographic data

Understanding the demographics of your target market and its changes—population growth, income levels, etc.— provides crucial context for channel strategy, especially in diverse markets. For example, OOH advertising may be more effective in densely populated urban areas, while TV advertising might be better suited for rural areas.

Why it matters: Demographic data helps explain underlying sales trends, avoid misattribution, and improve long-term forecasting. It also allows for optimizing channel strategies and tailoring campaigns to specific demographic segments, maximizing reach and impact.

Economic indicators

Macroeconomic factors—inflation, interest rates, unemployment, and consumer confidence—significantly influence purchasing behavior.

Why it matters: Incorporating economic indicators helps separate the impact of economic forces from marketing activities, leading to more informed decisions about budget and campaign timing during different economic cycles.

5 ways to use Supermetrics to build a data foundation for your marketing mix model

When your data is scattered, it’ll be hard to bring them under one roof and make sure they’re in the right format for your mmm model. That’s where Supermetrics comes in. I’ll discuss several ways you can use Supermetrics to build your data foundation.

1. Native Supermetrics integrations

Supermetrics offers pre-built connectors to the major ad platforms—Google Ads, Facebook Ads, TikTok Ads, web analytics tools—Google Analytics 4, Adobe Analytics, and sales platforms—Salesforce and HubSpot. These native marketing data integrations make moving data from point A to point B easy and eliminate manual work. This ensures a consistent and reliable data flow directly into your chosen destination.

2. Custom Data Import

Custom Data Import is a feature that allows you to import offline data from various sources, including CRM systems—Salesforce for sales data and customer interactions, ERP systems—SAP and Oracle for product information, pricing, inventory, and other relevant databases. You can combine your first-party and zero-party data with other external factors that influence marketing performance. You can upload this data via CSV uploads or integrations with databases and cloud storage solutions.

In this tutorial, my teammate Jack did a fantastic job explaining how Custom Data Import works.

3. Data blending and transformation

Once you manage to collect all your data, you need to transform and enrich it. For example, you can implement naming conventions, do data blending, create custom calculations and metrics, and handle currency conversions. This step makes sure your MMM dataset is clean, consistent, and ready for analysis. You can apply these transformations either within Supermetrics before the data is exported or within the data warehouse itself after export.

Check out this article, where our Product Marketer, Tea, showcases 7 data transformation use cases you can do with Supermetrics.

4. Export to data warehouses

When you’re done enriching your data, you can bring it to your data warehouses or lakes— Google BigQuery, Snowflake, Amazon Redshift, and Azure Synapse Analytics. Supermetrics’ direct export capability eliminates intermediary steps, ensuring data integrity and simplifying the data pipeline. You can use the computing power of data warehouses to store and process your data, especially larger datasets.

5. Automation and scheduling

Finally, since Supermetrics automates the entire data pipeline. You can schedule data transfers to run automatically at defined intervals. This makes sure your MMM dataset is always fresh and up-to-date.

What to do when you have limited data

In an ideal world, of course, it’d be great if you could feed more data to your model. However, in case your business hasn’t yet accumulated enough data, here are some tips to overcome the data limitation:

• Extend the data collection period: Gather more historical data to increase the number of data points. You need at least 2 years of data to run the model.
• Aggregate data at a higher level: For example, if you have daily data but not enough data points, consider aggregating it to weekly or monthly data. This reduces the number of parameters the model needs to estimate. However, be cautious, as aggregation can mask important short-term effects.
• Simplify the model: Reduce the number of variables included in the model. Focus on the most important marketing channels and control variables.
• Consider Bayesian modeling: Bayesian methods can incorporate prior knowledge and assumptions to improve model stability even with limited data. However, these techniques require careful selection of priors.
• Focus on directional insights: With limited data, the goal might shift from precise quantification to understanding the directional impact of marketing activities.

Ultimately, the amount of data required for a reliable MMM depends on the complexity of your marketing mix, the industry, and the specific business objectives. I recommend consulting with experienced data scientists or MMM specialists to determine the appropriate approach given your specific data constraints.

Better data means better outcomes

Building a solid data foundation is going to determine the results of your MMM. By controlling the data pipeline, from sourcing each input—sales, media spend, external factors—to the final MMM calculations, you can gain full visibility into the model’s mechanics. This transparency improves your confidence in the model and helps identify potential biases, errors, or omissions that could skew results.

And ultimately, you can use these results to guide effective budget allocation and marketing strategies.