AI-Driven Marketing: the rise of Machine Learning

Alex Mari
AI in Marketing Researcher @University of Zurich

Marketing professionals are facing an increasingly complex challenge. On the one hand, we are witnessing a proliferation of digital touchpoints and an objective difficulty in extracting value from extensive data collections in terms of volume, speed and variety; on the other hand, we have consumers with high expectations of interactions, content and personalized offers.

Machine Learning (ML) fits perfectly into this context, as demonstrated by the independent research "The Rise of Machine Learning in Marketing", approved by SwissCognitive, which collected the insights and experience of over 30 international AI Marketing experts including Jim Sterne, A.K. Pradeep, Scott Brinker and Andreina Mandelli. 

ML algorithms learn from data through non-predefined mathematical models and these cognitive systems, whether they’re integrated or not in marketing software, succeed in transforming information into seamless interactions with the consumer, predicting their behaviors, anticipating their needs and hyper-customizing messages. The result is a deeper one-to-one relationship.

 

ML diffusion in 3 key areas


ML is spreading into three key areas of Marketing: 

  • Within Marketing Technologies

Scott Brinker, who every year compiles the long-awaited Martech Landscape, states that AI algorithms will be increasingly integrated into different levels of marketing software (enterprise solutions, individual tools or apps) to turn historical data into actionable insights.

marketingai1 

The benefits can be: 

- direct: when an AI strategy is proactively implemented, defining ML models (built in-house, purchased from third parties or As-a-Service) so that they meet specific business needs;

- indirect: when marketing technologies incorporating AI components are used. 

  • In all Marketing functional areas

TML is affecting every single functional area of Marketing. The ROI is higher in performance marketing (paid media, marketing analytics and SEO), while it is lower in traditionally human-intensive activities (such as content creation, social media and service management). However, these dynamics are changing rapidly: in the future, ML is going to be used more and more to support creative and relational activities.

marketingai2 

  • Along the consumer journey

The different stages of the consumer journey require individual but coordinated ML algorithms, solving specific problems. Human intervention is required to set the algorithms from time to time, according to the goals (we’re not talking of "set and forget").


What are the objectives?

  • Automation

It is not just a cost-cutting automation of human work, this paradigm is also being transferred into customer experience. This applies not only to internal processes that are invisible to the user (targeting/segmentation) but also to proactive services that can be offered to improve user interaction with the brand (e.g. Vodafone's chatbot TOBi has a 100% conversion rate compared to the website).

This also includes automated recommendations that use predictive functions: they account for 35% of Amazon's revenue and 80% of films viewed on Netflix. The level of automation is bound to grow: Gartner says that already 25% of the customer relationship with a business is managed without human intervention. 

  • Optimization

AI allows marketers to engage with consumers across different channels more effectively, while reducing manual work times and boosting the productivity of the entire marketing department. Optimization can be a key driver for a more granular approach to targeting (if I don’t own a cat, there’s no point in showing me cat food adv). 

  • Augmentation

The man-machine synergy is convenient. Algorithms will not replace people, but will enable them to work better, more efficiently and more intelligently. While algorithms are faster and more accurate in data processing, humans have better abstraction skills. This hybridization translates into greater consumer satisfaction and a significant reduction in costs.

  

ML processes in Marketing

marketingai3

  • Data

AI algorithms learn inductively from data analysis. However, there are two myths to dispel: 

- Algorithm models WON’T provide insights from insufficient data. In fact, we must pay attention to the quality of the data we feed the machine. The principle is garbage in, garbage out. 

- a model WON’T keep triggering over time without human control. In fact, the external environment is dynamically changing so ML must be reviewed regularly with new data sets, to avoid dangerous distortions. 

  • Action

In order to get insights from data, which in turn become real-time actions, the entire organization needs to be digitally mature and able to build data infrastructures, where to collect the specific characteristics of each consumer. 

  • Interaction

Users have high expectations and interaction is the focus. The brands that can successfully predict their desires and deliver relevant offers and customized services on time can earn the trust of their users.
 

Expected benefits


- Prediction

Predicting consumer behavior allows brands to offer value at the right stage of the consumer journey. This is obviously not possible with traditional marketing tools, unable to analyze the huge amount of data produced today.

- Anticipation

The ability to anticipate the users’ needs influences the offer of products and services. Think about Netflix, which through AI analysis of the creative elements of successful films is able to produce original television programs, doubling their success rate (from 40 to 80%). 

- Hyper-customization

It’s the ability to deliver relevant messages at the right time, on the right channel: relevance is a fundamental driver. For example, L'Oreal Paris customizes its videos with AI data on audience preferences: this campaign has shown a 109% increase in brand interest and a 30% increase in market share.

Marketing professionals are facing an increasingly complex challenge. On the one hand, we are witnessing a proliferation of digital touchpoints and an objective difficulty in extracting value from extensive data collections in terms of volume, speed and variety; on the other hand, we have consumers with high expectations of interactions, content and personalized offers