In the last few years, artificial intelligence has transformed the way companies handle data and interact with customers, opening up new opportunities for marketing. Indeed, AI makes it possible to analyse large volumes of signals, enabling a deeper understanding of preferences and behaviour. This level of knowledge enables the creation of highly personalised campaigns, improving customer experience and engagement.
In this context, the adoption of AI to identify and activate user interests represents an innovative and potentially disruptive tool in identifying the intent of the customer base.
In the past, user interests were identified by examining the number of contents visited in certain sections of the site. This approach, widely used by DMPs for contextual targeting, was essentially statistical deterministic and exploited the simple information architecture of a site.
Subsequently, large advertising platforms such as Google released very refined algorithms to understand the content of the page read by the user and extract its meaning. Today, thanks to the spread of AI on a global scale, we have the possibility of exploiting very similar approaches to large DMPs, with two major advantages.
The first is that we no longer have to rely solely on information architecture: AI is able to transversally analyse the site, understand and interpret content, extract its essence and attribute it correctly to the user, regardless of its location.
The second advantage lies in customisation: AI allows us to precisely select the interests we want to extract, aligning them to the business and customer needs.
As recounted in detail in one of our last articles, at Bytek, we have implemented an embeddings-based approach and developed a proprietary algorithm to assign interests to users accurately. The onboarding process is managed in a collaborative manner: we organise workshops with customers to identify the most relevant topics for their business (custom interests), which can be tracked within specific pages and content, and then associated with individual users. This approach also allows us to identify particular life events, which can influence preferences and behaviour.
In addition to tracking life events and custom interests, we also identify interests related to specific products or services, using the same methodology. We analyse users’ behaviour on pages, previously labelled using AI, and compare it with both personal navigation and that of other users with similar preferences.
This data, based on first-party information, enriches our understanding of the user and can be activated in various platforms: CRM, marketing automation systems, campaigns on Google Ads and Meta, programmematic and even on the website for targeted personalisation.
In the following paragraphs, we will examine in detail the ways and strategies to activate interest analysis, highlighting practical and customised approaches that can optimise the engagement and results of marketing campaigns.
CRM Activation and Cross-Sell Experiences
An area of great potential in activating the results of user interest analysis is cross-selling. Through the integration of interests directly within the CRM, this data becomes immediately available and actionable for the sales area. This allows the sales team to act proactively. For example, if the analysis reveals an interest in life insurance, the salesperson can act quickly, activating a dedicated strategy: he or she can contact the customer directly, or use pre-configured email templates, ready to be sent when a specific interest is identified. Moreover, these operations can be fully automated by grouping all customers in the CRM with similar interests and sending them personalised communications.
These strategies are an essential practice in modern marketing. The ability to understand in real time when a user manifests a specific purchase intent is crucial. By working precisely on interests, we can identify these opportunities and act accordingly.
Furthermore, communications based on specific interests and observed behaviour are not perceived as intrusive by the customer. On the contrary, data show that engagement and conversion rates improve significantly, as the content resonates with the exact moment and needs of the user, creating a more valued and relevant experience.
Personalisation with LLMs
The use of Large Language Models for personalisation represents another opportunity to exploit the analysis of user interests. Through prompt engineering, it is possible to include variables such as specific customer characteristics, membership segments, detected interests, customer value (e.g. customer lifetime value) and tone of voice, which can also be identified through interest analysis.
In fact, it is possible to create a taxonomy of tones of voice and understand which communication style is actually used on one’s website. Often one is convinced that one is using a certain tone of voice, but content analysis may reveal that another is being used. This insight can then be integrated into the prompts to achieve highly effective customisation results.
Reporting
Reporting based on interest analysis provides an in-depth understanding of one’s customer base, enabling informed decision-making. For example, if after calculating the interests of individual users we apply advanced clustering systems such as k-means models, we can identify the most relevant customer clusters and discover the interests they have in common. These insights provide valuable indications for refining marketing and sales strategies.
One of the key strategies is to adapt the editorial plan according to the prevailing interests of the best customers.
Also, if users show interest in certain topics but do not purchase related products, there may be a disconnect between the offer and their preferences. In these cases, it is useful to review the product strategy and adapt the communication to the real needs of customers.
Another important variable is the change of interests over time, especially as customers age. Customer cohorts, like generations, evolve: a company that ten years ago had a predominantly 30-year-old customer base will now be faced with a group of 40-year-old customers with different needs and interests. The ability to track these changes through dashboards that provide real-time data is essential to keep marketing strategies aligned with the evolving needs of the customer base.
Activating Interests with Event-driven Marketing
Event-driven marketing is also a possible strategy for activating data from the analysis of user interests. Thanks to artificial intelligence technologies, it is possible to transform identified interests into events that can be activated in real time. This approach makes it possible to develop reactive actions based on specific triggers, which are activated when certain user behaviours or manifestations of interest occur.
For instance, it is possible to configure triggers that, when a user shows a particular interest, trigger the sending of customised emails, generated by AI through prompt engineering. In parallel, retargeting strategies can be activated on platforms such as YouTube, creating specific campaigns towards clusters of users who have recently expressed a certain interest.
A further level of customisation can be achieved through the use of tools such as Mutiny, which allow the user experience to be modified in real time. In this context, a company can dynamically change the appearance of landing pages or specific content on its website according to the interests expressed by the user. When a user visits the page, the system recognises their interest and immediately adapts the content, presenting tailored offers or promotions to maximise the effectiveness of communication and engagement.
Understanding the different decision windows is crucial, as it allows media pressure and personalisation activities to be optimally calibrated, exploiting the most opportune moments to influence consumer choice.
Conclusion
As we have seen, companies that fully exploit first-party data, such as customer preferences and interests, can personalise communications across multiple channels (email, video, web) and significantly improve the user experience.
According to a McKinsey study, companies that excel at personalisation can see a 10-15% increase in revenue, with some companies achieving increases of up to 25%, depending on the industry and the ability to effectively execute personalisation. This is because offering tailored content and promotions at the right time improves customer engagement and loyalty, creating a ‘flywheel effect’ where recurring and relevant interactions lead to increased loyalty and long-term value growth.
