Written by Joleen Bothma

“If your site uses third-party cookies, it’s time to take action as we approach their deprecation”

— Google

Attribution modeling enables marketers to answer this critical question: “How do I best allocate my marketing spend to maximize ROI?” In other words, marketing attribution models allow marketers to credit the marketing channels and touchpoints throughout the buyer's journey that made a lead convert.

Traditionally, methods like first-click and last-click attribution have dominated. They rely heavily on data from third-party cookies to track user interactions across the web.

However, popular web browsers have already started saying goodbye to third-party cookies. The latest company, Google, finally also announcing plans to phase out third-party cookies in its Chrome browser. In their words, “If your site uses third-party cookies, it's time to take action as we approach their deprecation”.

Google also released a timeline summarising the upcoming phase-out. As of January 4th, 2024, they had already restricted cookies for around 1% of Chrome users. In 2025, they will ramp up the full phase-out.

This move is largely positive because it greatly enhances user privacy. However, it poses a challenge: traditional attribution models will lose their effectiveness without third-party cookie data.

This article discusses attribution models and their alternatives. We also present a case study of how we successfully implemented marketing mix models for a leading travel agency.

The challenge of traditional attribution models

Traditional attribution models assign credit to different touchpoints in the consumer journey, each with its own method of assigning value.

Here are a few commonly used traditional models:

First-click: Credits the first touchpoint that a customer interacts with before converting. It emphasizes the importance of initial engagement.
Last-click: Assigns all the credit to the final touchpoint before conversion, prioritizing the action that directly led to the sale.
Linear: Distributes credit equally across all touchpoints, recognizing each interaction's role throughout the customer journey.
Time decay: Attributes more credit to touchpoints that occur closer in time to the conversion, assuming that more recent actions are more influential.
Multi-Touch: a dynamic and probabilistic approach to understanding and attributing conversions, accommodating the multifaceted and complex nature of modern consumer journeys.

These models rely heavily on tracking user interactions, predominantly through third-party cookies, across websites and platforms. These cookies track users' digital footprints across the internet, providing the data needed for attribution models and helping marketers optimize their strategies.

However, these traditional models face significant reliability issues with the phasing out of third-party cookies.

Without third-party cookies, tracking user behavior across multiple sites becomes challenging. It disrupts the data flow needed to assign conversion credit across various marketing touchpoints accurately. As a result, marketers might see incomplete or skewed data, leading to less informed decisions about where to allocate marketing spend.

Why Consider Alternative Attribution Models Now?

“Cookies are only meant to be eaten!”

As third-party cookies become obsolete, companies must reevaluate their attribution models to maintain the accuracy and effectiveness of their marketing activities.

Adopting new models sooner rather than later is not just a response to technological changes but a proactive step towards aligning with global trends toward greater data privacy.

This transition away from third-party cookies is part of a broader movement toward enhancing consumer privacy. Regulations such as the GDPR in Europe and the CCPA in California have set precedents emphasizing the importance of protecting user data. These changes indicate a clear direction away from invasive tracking practices towards more privacy-conscious methods.

Companies that fail to comply with these regulations could face legal repercussions and damage to their reputation and customer trust.

Marketing mix modeling is a robust alternative, allowing us to analyze aggregated data rather than relying on user-level tracking.

Understanding Marketing Mix Models

Marketing mix models (MMMs) are made up of a group of statistical techniques used to quantify the impact of various marketing activities on sales outcomes.

These models analyze historical data to understand how different elements of the marketing mix, such as advertising, promotions, pricing, and distribution, contribute to these sales outcomes.

While traditional attribution models focus on user-level data to track individual consumer journeys across touchpoints, MMMs utilize aggregated data at a higher level.

Traditional models attribute sales directly to specific interactions or touchpoints. In contrast, MMMs employ statistical analysis to infer the effectiveness of various marketing activities over time. These models can often incorporate external factors like economic conditions and competitive actions.

Advantages of marketing mix models

Marketing mix models do not depend on individual user data, making them inherently compliant with privacy regulations.

By analyzing broader market trends and external factors like economic conditions and competitor actions, MMMs provide deeper insights that help in strategic decision-making.

These models can also adapt to changes in marketing strategies and consumer behavior without needing granular tracking.

Google published a research paper highlighting the challenges and opportunities in marketing mix modeling. It’s a must-read if you’re interested in diving into the technical details behind MMMs.

Python libraries for marketing mix models

The Python programming language is the number one choice for data analysis and machine learning. As we explore the transition to marketing mix models, it's essential to also familiarize ourselves with the associated Python libraries.

Two of the most popular Python libraries for developing and implementing MMMs are:

Google's LightweightMMM: Developed by Google, this Python package is designed specifically for marketing mix modeling. LightweightMMM uses Bayesian structural time-series models to estimate the incremental impact of marketing tactics on sales or other metrics.
Facebook's Robyn: Developed by Facebook, Robyn is an iterative, semi-automated marketing mix modeling package. While originally designed for the R programming language, they have recently released an API that allows you to use Robyn with Python.

Case Study: Success with Mix Models

By identifying over and underinvestment in marketing budgets across channels, our client could reallocate resources more effectively to maximize ROI.

A leading travel agency needed to evaluate the effectiveness of marketing efforts across different channels and improve their allocation of marketing budgets.

Methodology

We used marketing mix modeling to analyze historical data and quantify the impact of different marketing activities on sales outcomes.

We considered factors like how the adverts were communicated, the devices they were viewed on, and whether they were targeted at specific products.

In addition, using MMMs also allowed us to analyze and visualize how these sales outcomes changed over time and whether there were any seasonal trends that we needed to incorporate in our models.

Using backtesting techniques, we were able to validate our models, ensuring a high degree of reliability in our results.

Results

We enabled the client to gain valuable insights into how different marketing channels reacted to marketing activities, allowing for informed decision-making.

By identifying over and underinvestment in marketing budgets across channels, our client could reallocate resources more effectively to maximize ROI.

In TV advertising, for example, we were able to show that there was still room for additional investment since ROI had not yet reached its maximum potential.

On the other hand, we demonstrated diminished ROI in radio advertising due to over-investment in this channel.

Understanding the most effective ways to communicate with customers also helped our client tailor their messaging and targeting strategies for improved engagement and conversion.

Lastly, the analysis identified interactions between different marketing channels, uncovering synergies and opportunities for integrated marketing campaigns.

Preparing Your Organization for the Transition

1. Educate and train teams

Conduct workshops to educate marketing teams about the upcoming phasing out of third-party cookies and its implications.

Offer training sessions focused on the new tools and techniques involved in marketing mix modeling.

Consider bringing in external experts for specialized training or subscribing to online courses that cover advanced marketing analytics.

2. Gradually integrate new models

Start with pilot projects that use new attribution models in parallel with existing methods. This dual-running phase allows teams to compare outcomes and gain confidence in the new models.

Before moving on to more complex Bayesian or agent-based modeling, consider first starting with basic regression-based models in early pilot projects.

You could also implement the new models in phases, starting with less critical campaigns to manage risk and gradually expanding to more significant campaigns as the organization gains more expertise and confidence.

3. Enhance data infrastructure

Ensure that data management practices are robust enough to handle the types of data required for marketing mix modeling while ensuring high levels of data quality.

Assess and upgrade technological resources if necessary, such as data storage solutions and analytical tools that support the new attribution models.

Embracing the Future of Attribution Modeling with Agilytic

The core question at the heart of attribution modeling is how to best allocate marketing spend to maximize ROI.

The transition toward alternative attribution modeling techniques can be a smooth process with the right tools and strategies.

At Agilytic, our experience in implementing marketing mix models—as demonstrated in our case study with a leading travel agency—has equipped us with the expertise and insights necessary to guide other businesses through similar transitions.

Our success in enhancing marketing strategies through advanced attribution models proves the value of moving away from outdated, cookie-dependent methods.

We understand that the transition to new models can seem daunting. That's why we offer tailored solutions that fit seamlessly into your existing workflows, backed by comprehensive training and support.

Get in touch with our experts

Attribution Modeling in a Cookieless World