Home Blog Marketing Mix Modeli ...

Marketing Mix Modeling and Incrementality: how to analyze performances in a privacy-first world

Filippo Trocca

“Half the money I spend on Advertising is wasted, the trouble is I don’t know which half”.


This famous quote from John Wanamaker, a successful American trader and marketing pioneer, was spoken more than 100 years ago. 

Unfortunately, for many companies this hasn’t changed much.

On the contrary, new Privacy and Data Transfer Regulations and ever-increasing competition are making the situation even worse.

In force since 11 July 2023, a new agreement between the EU and the US on the subject of data transfer overseas effectively puts in place a third privacy shield, resolving a thorny issue that lasted over a year.

Until digital rights activists challenge the compliance of this agreement, this privacy shield guarantees a peaceful scenario for marketers and digital advertisers with active campaigns in international markets, even though the cookieless web and user rights remain central.

With this in mind, how should you manage and measure your advertising activities in a privacy-first world?

On the one hand, users demand greater control and transparency over the collected data, on the other we are witnessing a major change when it comes to Regulations, since GDPR, CCPA, e-privacy and other regulatory frameworks are introducing various solutions to protect users and better manage data sharing.

Regulation of browser updates is also changing for Safari, FireFox Edge, Opera, with blocked cookies and consequently limited measurement protocols.

If attribution tools can no longer work as accurately, in terms of business impact it means that

  • performances are less clear
  • CPA is on the rise
  • ROI is more difficult to analyze

As we have seen, measuring and attributing conversions by channel is complicated, and marketers still fail to answer two fundamental questions:

  • How many conversions would have happened without doing anything?
  • How many conversions belong to each channel?

Once you can answer these kinds of questions, you truly have full control over your marketing activities.

What is Marketing Mix Modeling and how it works

Marketing Mix Modeling represents a turning point in this scenario, because it analyzes the impact of marketing and non-marketing activities on incremental sales, with a privacy-friendly and data-driven approach.

As a matter of fact, this model is able to calculate the baseline – i.e. all the sales that would have occurred even without marketing activities – and which conversions can be attributed to each individual channel.

The algorithm is fed only by aggregated data: impressions, sessions, conversions, cost information and results in terms of revenue. This proprietary data can be collected without identifying and profiling the user: they are always available, because they will never be influenced or limited by new privacy laws or institutional decisions.

To evaluate the macro-economic trend of the context, these data are integrated with those relating to competition or demand for the specific product/service. 

In this sense, the model collects general data – such as the average cost of products on the market, inflation data or the employment rate – but also more specific information – for example, people’s mobility data obtained from Google Maps, very useful during the pandemic.

Robyn’s Analysts guide to MMM – Meta’s Open Source MMM package –  helps marketers to fully understand the potential of this tool: “Think of the Marketing Mix Model as “unbaking” a cake. Imagine going to a restaurant and ordering dessert, the cake is delicious. You want to find out what ingredients were used to make it so you can make it yourself. Discover the recipe and you’re good to go! Now you can successfully recreate this dessert anytime you want, and you might want to experiment and see if it’s just as delicious with other ingredients, or what amount of ingredients will turn into even better, more flavorful results.”

In fact, MMM models allow you to better understand the factors (ingredients) that influence the business KPI (the cake) over time, allowing more flexibility throughout the strategy.

This basically results in a contribution on the turnover and more budget control, to better understand current performances and which channel/medium need less or more investments.

An end-to-end solution to apply Marketing Mix Modeling 

The main advantages of Bytek MMM solution are the speed of execution, the strong first-party data base and the ability to forecast future scenarios.

Even though it adapts to each specific case and each company’s specific needs, it consists of 6 main steps.

  1. Business Understanding

In the first phase Bytek and your team will discuss with your team to define:

  • Business Goals
  • KPIs to monitor
  • Active channels
  • Variables impacting the business such as seasonality, promotions, etc.

This process highlights which input data to use to develop the most accurate statistical and descriptive analyses.

 2. Automated data pipelines

The second phase is dedicated to the collection of data from all available sources, in order to centralize them in a single Data Warehouse.

This operation unlocks essential information, such as the contribution margin, profit analysis and the real attribution of the last click, etc.

3. 100% data integrity

In this phase, samples of raw data are analyzed and compared with information directly from the source.

Once you make sure there are no discrepancies in the raw data, the process is repeated for all of the aggregated data.

4. Data exploration and modeling

In this phase of exploration and modeling of the available data, the Bytek platform explores and learns from the specific data of each company, creating and testing a custom model.

5. Budget allocation

Through sophisticated optimization algorithms, the model develops the best budget distribution between campaign types in order to maximize conversions or sales.

These algorithms are optimized to consider many variables at once: from upcoming holidays, to the impact of seasonality, to channel saturation points, and so on.

6. All data in one place

All data is cleaned, organized and aggregated and then displayed in a custom dashboard, where all information on marketing and growth performance are available and easy to understand.

Thanks to shared insights, the following tactics and actions will be based on strategic insights, so as to implement real data activation.


In fact, Bytek’s Marketing Mix Modeling approach can be included in an even more complete Data Empowerment framework: a broader strategy of collecting, enriching and activating your company data.