Marketing Mix Modeling for Dutch Beauty Brands

The 150K Plateau: Why Your Beauty Brand's Growth Has Stalled

Your Dutch beauty brand is stuck. Every euro you pour into Meta or TikTok ads feels less effective. Your ROAS is declining, and you have no clear path to scaling further. This is not a sign of failure. It is a symptom of a broken system.

For years, the industry has pushed last-touch and multi-touch attribution as the gold standard. These models are simple to understand but flawed. They assign credit based on correlation, not causation. They cannot tell you if a channel is genuinely driving new customers or just taking credit for sales that would have happened anyway. This is the world of cannibalistic channels, where your marketing efforts are actively working against each other.

Before: Trapped by Flawed Attribution

Flawed attribution refers to marketing models that incorrectly assign credit to touchpoints, leading to misinformed budget decisions. Unlike causal analysis, which identifies true drivers of sales, flawed attribution relies on simple correlations, like last-click, often overvaluing channels that are not genuinely creating new customers.

After: Clarity and Scalable Growth with Marketing Mix Modeling

Marketing Mix Modeling (MMM) is a statistical analysis technique that quantifies the impact of marketing and non-marketing activities on sales. Unlike attribution models that track individual user journeys, MMM uses aggregated historical data to provide a top-down, causal view of performance. This allows ecommerce brands to understand the true ROI of each channel and refine their budget for incremental sales, not just correlations.

Unlike traditional attribution, MMM does not rely on user-level tracking. Instead, it uses aggregated data to model the relationship between your marketing inputs and your business outcomes. It is a top-down approach that provides a complete view of your marketing performance, as detailed in seminal research on the topic [1].

By analyzing historical data on sales, ad spend, pricing, and promotions, MMM can isolate the causal impact of each marketing channel. It can tell you how much each channel contributes to your overall sales, how your channels interact with each other, and what the point of diminishing returns is for your ad spend. For Dutch beauty brands, this means a fundamental shift in how they approach growth. Instead of being beholden to the black-box algorithms of ad platforms, they can make data-driven decisions based on a causal understanding of their marketing ecosystem. To see how different attribution models compare, check out our attribution modeling tool.

The Bridge: From Correlation to Causation

Causal inference is the process of determining cause and effect, moving beyond simple correlation to understand the true impact of a variable. Unlike traditional marketing attribution, which often mistakes correlation for causation, causal inference uses scientific methods to identify which marketing activities genuinely drive sales. For ecommerce, this means making budget decisions based on what actually works, not just what appears to be working.

The bridge from the 'before' to the 'after' is causal inference. This is the scientific process of determining cause and effect. While traditional marketing attribution is based on correlation, MMM is a form of causal inference. It allows you to move beyond simply observing that two things are related and start to understand if one is actually causing the other.

One of the key techniques used in modern MMM is Bayesian inference, a topic we explore in depth in our guide to Bayesian Marketing Mix Models. This approach allows you to incorporate prior knowledge and beliefs into your model, resulting in more accurate and stable results. For example, if you have a strong belief that your brand's PR efforts have a significant impact on sales, you can incorporate that belief into your model. This is a far more sophisticated approach than simply letting an algorithm decide what is important, a concept further explored in Google's research on the challenges of media mix modeling [2].

A customer might see a TikTok ad, then a week later see an influencer's post, and then finally make a purchase after seeing a retargeting ad on Meta. Traditional attribution models would give all the credit to the Meta ad. Causal inference is not just a theoretical concept. It is a practical tool that can be used to make better marketing decisions.

Practical Steps to Implement Marketing Mix Modeling

Implementing Marketing Mix Modeling involves a structured process of data collection, model building, validation, and insight generation to sharpen marketing spend. Unlike simply tracking platform metrics, this approach requires aggregating historical sales and marketing data to build a causal model.

So, how can you get started with marketing mix modeling? Here are a few practical steps:

1. Data Collection & Aggregation: The first and most critical step is to gather and aggregate all the necessary data. This includes historical data on your sales (revenue and volume), ad spend per channel (e.g., Meta, Google, TikTok), impressions, clicks, pricing changes, promotions, and any other external factors that might influence your sales, such as seasonality or competitor activities. The data should be aggregated at a consistent frequency, typically weekly or daily. The more granular and complete your data, the more accurate and insightful your model will be.

2. Model Building & Specification: Once you have your data, you can start building your model. This is typically done using statistical software packages like R or Python, with libraries specifically designed for MMM, such as Robyn or LightweightMMM. The model specification involves defining the relationship between the dependent variable (sales) and the independent variables (marketing activities and other factors). This includes specifying the functional form of the relationship (e.g., linear, non-linear), the lag effects of advertising, and the carryover or ad-stock effects.

3. Model Validation & Calibration: After you have built your initial model, it is important to validate it to ensure that it is statistically sound and provides accurate results. This involves checking the model's goodness-of-fit, the statistical significance of the variables, and the plausibility of the estimated coefficients. You should also perform out-of-sample validation to test the model's predictive accuracy on new data. Calibration involves fine-tuning the model parameters to improve its performance.

4. Insight Generation & Decomposition: Once you are confident in your model, you can start using it to generate findings. The primary output of an MMM is the decomposition of sales into its various drivers. This allows you to quantify the contribution of each marketing channel to your overall sales, as well as the contribution of non-marketing factors such as seasonality and price. You can also use the model to calculate the ROI of each marketing channel and to understand the diminishing returns of your ad spend. Our ROAS calculator can provide a starting point for this analysis.

5. Action, Refinement & Iteration: The final step is to take action on your insights to sharpen your marketing mix. This might involve reallocating your budget from less effective channels to more effective ones, adjusting your ad spend levels to avoid diminishing returns, or changing your promotional strategy. It is important to treat MMM as an iterative process. You should continuously monitor your results, update your model with new data, and refine your marketing strategy based on the latest insights.

FAQ

What is the main difference between marketing mix modeling and attribution?

Marketing mix modeling is a top-down, strategic analysis using aggregated data to measure the causal impact of marketing channels on sales. In contrast, attribution is a bottom-up, tactical approach that assigns credit to individual touchpoints along the customer journey. MMM focuses on the why of sales, while attribution focuses on the what.

Is marketing mix modeling only for large companies?

No. While MMM was traditionally a complex and expensive undertaking, modern tools and techniques have made it accessible to businesses of all sizes.

How can I get started with marketing mix modeling?

The first step is to start collecting and consolidating the right data. This includes historical data on your sales, ad spend, pricing, and promotions.

How does marketing mix modeling handle online and offline channels?

Marketing mix modeling is uniquely capable of measuring the impact of both online and offline channels in a single, unified model.