How new AI tools can transform customer engagement and retention

As the cookieless future continues to gain momentum, the global digital advertising sector is experiencing a tectonic shift. Companies are being forced to reimagine the way they reach out to customers.

Online marketing has been dominated by third-party cookies — tracking codes posted on websites to extract users’ information — and data brokers who sell the information in bulk.

However, this multibillion-dollar business, perpetuated for decades, is now in checkmate by a perfect trifecta: new privacy laws, big tech restrictions, and global consumer privacy trends.

While the end of cookies is inevitable, businesses still struggle to find new advertising techniques. Statista’s January report reveals that 83% of marketers still depend on third-party cookies, spending $22 billion on this outdated technique in 2021.

In this report, we’ll dive into the complexities of digital advertising transformation and reveal how new technologies, machine learning (ML), and AI present new opportunities for the industry.

The challenges, risks and new trends of digital marketing

Using third-party data has become a high-stakes risk strategy. Companies that do not observe data privacy laws can face millions in fines for data breaches or misuse. For example, defying the General Data Protection Regulation (GDPR) can cost up to €20 million (about $21.7 million) or 4% of a company’s annual global turnover in 2023.

And the legal data landscape goes well beyond the GDPR; it is diverse, constantly evolving, and growing. From state laws like the California Consumer Privacy Act (CCPA) to federal laws like the Health Insurance Portability and Accountability Act (HIPAA), businesses must identify which laws apply to their operation and know the risks.

The dangers of running third-party data campaigns do not end with courts. Brands that do not align with consumer expectations risk losing clients and business opportunities. A 2022 MediaMath survey revealed that 84% of consumers are more likely to trust brands that prioritize using personal information with a privacy-safe approach.

The issue is not new — privacy concerns have been growing for years. In 2019, Pew Research reported that 79% of Americans were “concerned about how companies use their data.” In 2023, privacy has become a top priority, and customers expect companies to protect their data. Failing to do so brings devaluation of brand perception and potential loss of customers and business partners.

The most significant barrier to third-party data is coming from online giants themselves. Companies like Apple, Google and Microsoft are leading the way towards ending cookies. Increasing restrictions make it harder for marketers to obtain consumers’ data daily.

First-party data — obtained under consent in a direct relationship with the user, for example, when making a payment transaction or agreeing to the terms when signing up — is trending and expected to replace third-party data. First-party data is also better-quality, as it goes beyond limited information based on age, location and gender. Furthermore, companies can use first-party data to create modern data marts.

ML and AI: From raw data to value to action

First-party data such as that collected through endpoints like point of sale (PoS) terminals can generate data and significant potential to target lifetime value (LFT) customers. LFT campaigns are trending as companies like Uber, DoorDash and Spotify find new ways to reach their customer base, Reuters reports.

The challenge both startups and big companies share is building, maintaining and managing the first-party data they collect from their customers in what is known as “data marts.”

Imagine the vast amount of raw data that a company can generate. Even when this is first-party data — sourced directly from their customers — not all of it can be used, is accurate, or is valuable. And that is what LFT campaign managers have to deal with. They must scan a sea of raw data to find very specific information.

This is where AI and ML come into play. AI/ML applications can find that needle in the haystack and do much more when managing data marts.

Understanding data marts

Data marts are a subset of information found within data warehouses. They are built for decision-makers and business intelligence (BI) analysts who need to access client-facing data rapidly. Data marts can support production, sales and marketing strategies when they are compiled efficiently. But building them is easier said than done.

The challenge with first-party data marts is the amount of raw data analysis needed to build them. This is why the automation, augmentation and computing processing power of machine learning (ML) and AI have become the tip of the sword in the new era of data-driven marketing predictive analytics.

Feature engineering: Building consumer buying signals

Feature engineering is a crucial component for AI and ML applications to effectively identify features — valuable data. Selecting the right features that the AI algorithm can use to generate accurate predictions can be time-consuming. This is often done manually by teams of data scientists. Manually they test different features and optimize the algorithm, a process that can take months. ML-powered feature discovery and engineering can accelerate this process to just minutes or days.

Automated feature engineering can simultaneously evaluate billions of data points across multiple categories to discover the critical customer data needed. Companies can use ML feature engineering technologies to extract essential information from their data marts, such as customer habits, history, behaviors, and more. Companies like Amazon and Netflix have mastered feature engineering and use it daily to recommend products to their clients and increase engagement.

They use customer data to create what is known as consumer buying signals. Consumer buying signals use relevant features to build groups, subsets or categories using cluster analysis. Usually, signals are grouped according to customers’ desires, for example, “women and men who practice sports and have an interest in wellness.”

But developing and deploying the AI apps or ML models to run signals-based targeting marketing campaigns is not a once-and-done task. AI/ML systems need to be maintained to ensure they are not drifting — generating inaccurate predictions as time progresses. And data marts need to be updated continuously for data changes, new data additions and new product trends. Automation in this step is also essential.

Additionally, visualization is key. All stakeholders must be able to access the data the system generates. This is achieved by integrating the ML model into the business intelligence dashboards. Using BI dashboards, even those within the company who do not have advanced data science or computing skills can use the data. BI dashboards can be used by sales teams, product development, executives and more.

Final thoughts

While AI and ML have been around for decades, it is only in the past few years (and months for generative AI) that they have truly taken quantum jumps. Despite this accelerated pace of innovation, companies and developers must strive to stay ahead of the game. The way forward is simple. Businesses must look into ways the tech can be used to solve real-world problems.

In the case of data privacy, the end of cookies and the end of third-party data, AI can be used to revisit this original problem and innovate its way to a new, never-thought-of-before solution unique to every company. But planting the seed of AI ideas is but the start of the journey. Craft and hard work are needed to follow through. The potential of ML and AI is, in this perspective, endless and highly customizable, and capable of serving each organization to achieve its unique goals and targets.

Ryohei Fujimaki is founder and CEO of dotData.