DigitalFacebookFeaturedInsightsInstagramMarketingOpinionSocial Media

Marketing mix modelling: A view from Meta’s Rasheeqa Jacquesson

Rasheeqa Jacquesson_Marketing Science Partner for MEA, Meta

By Rasheeqa Jacquesson, marketing science partner for MEA at Meta.

Today’s digital landscape enables marketers to seemingly access a limitless amount of data. But are they taking advantage of that data to the fullest extent possible? This is not the case for many organisations. For the past decade, direct response marketers have used multiple data sources, such as cookies and mobile device IDs, to measure advertising results. This data is becoming increasingly difficult to access, and with it, measurement systems, such as channel-level reporting and cross-channel attribution, are becoming less effective.

Digitally native brands accustomed to making fast, frequent optimisations based on granular data feeds need to reconsider their measurement strategy. In this category, are marketers of disruptor brands – startups that were born online and deliver value in new and innovative ways to customers. The time has come for marketers, including digital natives, to evolve their existing marketing analytics. With an eye to 2023, Marketing Mix Modelling (MMM), a statistical analysis of marketing and business, will thrive as a relatively future-proof measurement solution for brands seeking actionable insights into cross-channel marketing effects.

With privacy laws changing the way marketers gather and leverage data, here is how brands can use marketing mix modelling to gain insights and measure performance.

Marketing mix modelling in a new era

 Thanks to the innovation of machine learning algorithms, MMM is allowing advertising actionable insights at the desired level of granularity and speed. Not relying on individualised data, it is emerging as the only truly holistic and resilient measurement system, in a changing data ecosystem. It’s great to see marketers moving towards continuous MMM – developing capabilities and hiring talent to manage this in-house. A good indication of this shift is the growth of the Robyn MMM open-source code community both on Github and Facebook. More than 1000 marketers, analysts and data enthusiasts are already connecting, discussing and learning as they progress on their MMM journey.

In a recent study by Accenture, they ran multiple experiments using custom-developed MMM to test its suitability for marketers of disruptor brands, who need a budget-friendly, reliable system to inform granular media optimisation.

Their findings show that MMM, when used with advanced machine learning techniques and innovations, provides benefits including:

1)   Robust MMM is attainable for marketers of all sizes and categories: MMM’s success is largely measured by the model’s ability to predict the dependent outcome. Accenture’s experiment, which used 1,200 data features across 5 different sources commonly collected by marketers of traditional brands and by those of disruptors, proved a high level of prediction accuracy, meeting industry-standard measures of 90% in R-squared values and 5% or below in mean absolute percentage error (MAPE).

2)   MMM can produce granular and actionable results: With the advances in machine learning, MMM can now utilise techniques such as the gradient descent algorithm to break down data and produce actionable insights based on these variables. In Accenture’s research, the model broke down two years’ worth of data into the day-of-the-week level, a necessary dimension for marketers making daily-level budget shifts (as shown in Figure 1). This is one example of how MMM can produce actionable and granular insights that marketers can use to optimise marketing effectiveness.

3)   MMM demonstrates cross-channel synergies without user tracking: Disruptor brands are often keen to understand their customers’ paths to conversion so that they can optimise their cross-channel marketing efforts. Marketing mix models, when integrated with advanced machine learning techniques, can provide similar insights on cross-channel synergies. Accenture’s experiment produced clear insights into cross-channel impacts. Results are summarised below in Figure 2’s ‘web’ of contributing factors.

A method that will sustainably meet the chief measurement needs

MMM combines and evaluates all online and offline marketing activities, builds a picture of relationships between them, and expands to monitor factors such as promotions, seasonality, or competition. Today, MMM requires less resourcing and budget to implement and uses commonly-collected data to deliver quick and detailed cross-channel insights, making it suitable for marketers of brands of all sizes – including those focused on direct response advertising. Even marketers of disruptor brands, who navigate frequent business changes and tend to use various non-paid marketing tactics, benefit from the comprehensive view delivered by MMM.

Here are four best practices to achieve actionable, accurate and sustainable results:

1)   Align on key objectives prior to modelling: Given the variety of questions, MMM can answer, it is important to build an MMM learning agenda and focus on addressing one question at a time. Aligning the key objectives is the fundamental first step in the MMM process. All the subsequent steps in the MMM building process will benefit from a clear understanding of the key objectives of MMM.

2)   Ensure data is relevant and comprehensive: Create separate variable needs for each strategy that the marketers want to quantify the ROI. In addition to preparing the media-related variables, it is equally important to include a comprehensive list of non-media variables that can also influence the brand’s business outcome. Such variables will vary from brand to brand, but some common ones include economic factors, seasonality, competition, etc.

3)   Choose the MMM option that addresses your questions: The most common questions are typically already addressed by the different MMM options on the market, whether it is an open-source solution such as Robyn, a partner solution, or a self-service MMM SaaS solution. When evaluating MMM options, make sure their capabilities can effectively answer the questions identified in step 1, in order for brands to generate actionable insights from MMM.

4)   Regularly refresh and calibrate MMM to reflect business changes: Investing in a data infrastructure that allows you to automatically ingest new data into the MMM model will help marketers and model practitioners gain long-term efficiency in refreshing their MMM models. It is important to build a framework to calibrate and choose the most robust MMM model. Incrementality studies are the industry’s gold standard to measure ground truth. Marketers should run incrementality studies on their marketing channels in parallel with running MMM to improve model accuracy and strengthen the trust in MMM adoption.

Resilient measurement solutions need a solid foundation

MMM and its evolutions are here to stay. It is time for disruptor brands to stop waiting and get started on their MMM measurement journey. The best way for marketers to understand advertising efficacy is to adopt rapid modelling alongside causal testing. Setting up marketing mix modelling now, before the doors to individual-level data close, will help prepare brands for the changes to online privacy ahead. Time and resources invested in building this now; will pay off for those who do.