2 min.
Back to the future: Performance measurement for privacy-centric marketing
1L’art de la gestion de projet2Un projet à succès commence par une bonne gouvernance3Cascade, agilité, demandes de changement?

Back to the future: Performance measurement for privacy-centric marketing

Analytics & Tracking

The loss of non-durable identifiers due to technological changes (iOS and Google) as well as the advent of legislation to protect user privacy has had a significant impact on the usual ways digital performance is measured.

To date, the traditional way that was most commonly used was multi-touch attribution (MTA). This type of conversion modelling was popular given the granularity of the data produced at each stage of the consumer journey. The problem? It necessitated heavy dependence on cookies… which are about to disappear.

The loss of third-party cookies, coupled with the necessity to obtain consent from users in order to capture and use their data (the infamous Law 25), requires marketers to change their ways of doing things to be able to continue making connections between marketing investments and business results. As a result, we are witnessing a return to statistical tools that have already proven their worth in the past: statistical inference models such as marketing mix modelling (MMM), regression-based attribution (RBA) or even autoregressive integrated moving average (ARIMA).

These methods of measuring performance don’t require being able to access very granular data and therefore can stand up to the legal and technological changes currently taking place.

Use of these methods for planning marketing activities

When planning your marketing campaigns, what’s most important is to be able to understand the past to better anticipate the future, as well as put in place the technological elements that will allow you to measure your performance when you are in the activation phase (campaign launch or strategy operationalization).

In concrete terms, a better understanding of the past means having a good understanding of your marketing and the trends occurring within it, your competitors, and the evolution of your customers’ needs. Understanding the customer primarily happens through the analysis of primary data owned by the company. These data are a strategic element in developing high-performing campaigns. To quote the former CEO of General Electric, Jack Welch, on the importance of knowing your customer: “There are only two sources of competitive advantage: the ability to learn more about our customers faster than the competition and the ability to turn that learning into action faster than the competition.”

Better understanding the past also means understanding the performance of past marketing activities as well as the relationships between investments made and revenue generated. Two of the methods of measuring impact cited above, MMM and RBA, have a key role to play in the modern marketing tool box.

A leading, proven model available to marketers, MMM enables measurement of the incremental value of marketing actions and identification of the saturation threshold for ad formats as well as their ROI. By aggregating investment data online and offline, we thereby get a complete view of the impact of marketing actions. A useful tool when it comes to obtaining an overall viewpoint that can guide you in planning marketing activities, this model needs a lot of historical data (generally two to three years’ worth). This makes it less flexible and fairly lengthy to implement.

As an alternative or complement to MMM, RBA has some good arguments for being a solid attribution solution. This model is based on the same statistical methods as MMM but is focused essentially on digital marketing data. This makes it much more actionable, with a much higher frequency execution (weekly or monthly). Using this model, we are looking to understand the relationship between inbound variables (i.e., clicks, impressions, investments) and outbound variables (results) through historical data.

Even if simulation is possible with the RBA model (particularly with budget optimization scenarios), it isn’t the highest performing model when it comes to making predictions about time series. For this use case, ARIMA should be preferred. We use time series data (i.e., the change in business results over time) to predict future trends in these series by taking seasonality into account.

From a technological point of view, given the continuing disappearance of cookies, it has become essential for companies to implement a marketing data warehouse. This lets you centralize all primary customer data as well as marketing data in order to create a 360 view of your customers and marketing efforts. This intelligence then allows you to prioritize the segmentation and targeting of potential high-value customers!

What does performance measurement mean when you're in the midst of an activation?

When activating your marketing campaigns, performance measurement must be handled practically in real time. For this reason, attribution in platforms remains the most efficient way to measure campaign performance. To accomplish this, it’s important to be able to rely on conversions based on primary data (mainly email).

Your digital properties must integrate solutions that use persistent identifiers (email, etc.) to measure conversions (for example, Google enhanced conversions). This method of identifying conversions (using primary persistent identifiers) is available for Meta and Google DV360.

Google Analytics 4 (GA4) integrates certain features to mitigate the loss of data. The name of this feature is behavioural modelling for consent mode.

As you have seen, technological and legal changes are upsetting our traditional approaches to measuring performance online. Marketing specialists must turn to new statistical methods such as marketing mix modelling (MMM) and regression-based attribution (RBA) to continue quantifying the value generated by their activities. At the same time, they need to integrate attribution features that exploit primary data within their advertising platforms. This adaptation is crucial to guarantee a reliable correlation between investments and results in a context where available data are dwindling.