Preparing Your Business for a Future Without Cookies
As Quebec’s Bill 25 recently came into effect, many digital marketers are trying to envision the future of their profession in a world where user data is no longer a given. In an Internet once erected as a perfect surveillance system, data was as abundant as grains of sand on the seashore. But spying is going out of fashion, and the sand is mostly slipping through the hourglass as we move towards its complete prohibition.
While it’s evident that a paradigm shift is on the horizon, it’s not easy to predict what will follow the current zeitgeist. What avenues are available today to digital marketers navigating the quicksands of online marketing?
Sacha Benadiba, a data analyst at Click & Mortar, foresees a resurgence of mixed marketing models (MMM). Dating back several decades—invented circa 1950—MMM had lost popularity in favour of cookie-based marketing. But it’s known that the latter is currently undergoing an apocalypse as international laws put brakes on the systematic collection of personal data.
The significant advantage of MMM is that it doesn’t require user data to operate. Instead, it measures the impact of a campaign using consumer-independent variables. No cookies available? This family of marketing tools doesn’t consume them anyway.
Sacha recently submitted his Master’s thesis, which he dedicated to developing two MMM models. By doing so, he became our expert in the field. Sacha took the time to answer some essential questions about this marketing tool, which, according to him, could make a spectacular post-apocalyptic comeback.
Let’s start from the beginning. What exactly are we talking about when we refer to a “mixed” marketing model (MMM)?
In fact, it’s a family of statistical models designed to analyze the outcomes of marketing efforts — that’s the core of the model. For example to determine how a target variable, such as income, is influenced by the marketing activities carried out by our company.
By extension, they allow us to answer questions like “How much effort should I put into my channels, based on how much they contribute to my revenues?” It’s a model that has been around for a long time, and a lot of documentation is available on it. It can be used by any organization engaged in marketing. The only thing to keep in mind is that these are models that require a certain level of digital maturity because there are several parameters to consider. For example, if we run a car ad on television, we don’t expect people to come and buy a car directly. We would try to determine what the time lag is [between advertising and potential purchase]. So, it’s essential to take the time to think about each of these little parameters. However, once we have all that, it’s a very effective tool for improving our marketing and understanding how it impacts our revenue.
Several traditional practices in digital marketing, which rely on liberal and unlimited data collection, will become impossible when Law 25 comes into effect. How can mixed marketing models fill this void?
In fact, this model existed before the digital era. But yes, there’s a significant reason explaining the resurgence of these models: the strengthening of laws around personal data security. In Europe, there has been the GDPR since 2016, and in Quebec, Law 25, which came into effect on September 22. Users are now given control over their personal data, meaning they can manage their cookies. So, for a digital marketing company, it becomes challenging to quickly analyze the impact of advertising campaigns.
This is precisely where MMM comes into play, thanks to the type of statistics on which the model operates. With data that doesn’t concern users and solely belongs to the company, we can perform analyses on its revenue. I can attempt to observe if my revenue increases when we invest more in a TV advertisement, for instance.
You have developed two mixed marketing models yourself. What sets these models apart, what can they measure, and who do they serve?
These are models based on regression; but today, with advancements in machine learning, we can achieve more sophisticated models. I wanted to compare the two, starting with linear regression, which is more basic, to then see what a more advanced machine learning model can bring to the base model.
My second model is called the “Forêt aléatoire,” which is a typical machine learning model. With it, we can go much further and capture the synergy effects between different marketing channels. It allows us to determine, for example, if I run a TV ad and a Facebook ad, what is their synergistic effect—how do they contribute to each other?
Now that the model is developed at Click & Mortar and ready to be used, could it be adapted to any client?
Absolutely. That being said, as I mentioned earlier, it’s important that the model is tailored to the specific context of the company for which it’s being developed, meaning it should be adapted to its reality. If it’s for a service-based company, for example, we would need to consider the weather variable. It also depends on the client’s digital maturity: do they have clean and sufficient data? These are two very important factors to consider.
What type of business would have the ideal context to engage Click & Mortar to explore these MMM avenues?
Firstly, as I mentioned earlier, the quality of the client’s data is important. Without that, it’s challenging to obtain results that accurately reflect reality. Additionally, it can take a lot of time to develop, so having the necessary budget is crucial.
However, it’s suitable for a wide range of clients: anyone looking to understand how their marketing performs and what they could do to improve it. In fact, there are several questions that can be answered with MMM that cannot be addressed using cookies.
Ah, you’ve got my interest. What kind of questions are we talking about?
Saturation, for instance: if I invest more in my Google campaigns or my Meta campaigns, will the returns follow, or would it be better for me to allocate my money elsewhere?
Synergy, too, is more easily identifiable, as well as the time lag between advertising and revenue. It’s also a model that has a great capacity to bridge the gap between digital and traditional. So, it’s an excellent tool for businesses with both an online and a physical presence, to understand how a digital campaign impacts in-store sales and how traditional campaigns influence online sales. Without MMM, such insights are very challenging to obtain.
You dedicated your master’s project to MMM. Did you have any major revelations by studying it so closely?
It’s a highly versatile model: the variety of questions it can answer is quite incredible. The model can be stretched in various directions to address saturation effects, time lag, and their impact on my revenue.
At the same time, I realized how sensitive the model is. It’s essential to study the parameters one by one. The slightest variation in the input parameters can affect the results. So, there are significant calibration challenges at the outset. During the four months I spent building the model, I spent half of that time on the database, making absolutely sure there were no errors in it.
Do you see mixed marketing models evolving significantly in the future? Becoming more precise? Or do you think we’ve reached a point of maximum expertise?
Certainly, artificial intelligence is likely to accelerate the processes. However, the most significant advancement I foresee is an improvement in data quality with the use of “data clean rooms.”
I mentioned earlier the challenge of data quality and quantity required for an MMM model: data clean rooms address this challenge very effectively! The current enthusiasm for them is precisely due to their focus on data privacy and anonymization.
To find out how your data can show us the way to successful campaigns, contact Click & Mortar. It will be a pleasure to discuss it with you.