Product descriptions:
Founded in Brazil in 2001 by Guilherme Benchimol and Marcelo Maisonnave, XP Inc. (XP) set out to make financial markets clearer and more accessible to everyday investors. Two decades later, XP has grown into one of Latin America’s leading financial platforms, serving over 4.7 million active clients with investment, banking, credit, and brokerage solutions. To drive growth and stand out in a competitive market, XP overhauled their customer acquisition strategy using first-party data for predictive AI modeling in Databricks, leading to better segmentation and performance media optimization.
The challenge
Amidst growing competition and stricter privacy regulations on digital advertising methods, XP faced increasing pressure to stand out in a crowded market. To grow efficiently, XP needed to rethink their customer acquisition and advertising strategy from the ground up.
Building a secure, scalable foundation for predictive AI
For XP, customer acquisition is more than just account creation. Their true measure of success is when a new user makes their first investment in the platform. XP initially optimized their ad platforms for account sign-ups, which was the easiest event to track in real time. But this created an acquisition challenge: it could take up to 14 days for 80% of new users to invest. This limitation caused them to miss out on acquiring more valuable users who were likely to invest later.
The solution
To close this gap, XP decided to build predictive AI models that could identify individuals more likely to invest. These predictive machine learning models required access to accurate, high-quality customer data while maintaining strict compliance with privacy regulations. XP also had to ensure that they did not expose sensitive customer data or introduce unnecessary complexity and risk.
Driving revenue with predictive AI
With their data foundation in place, XP’s data science team used Databricks MLFlow to train a predictive model that estimated each customer’s likely investment behavior based on early signals, such as browsing patterns, sign-up flow responses, and regional trends. Hightouch syncs these predictions into XP’s marketing and customer engagement tools.
XP chose Databricks as the foundation for data science and governance, enabling teams to develop advanced machine learning models with built-in data security management. They paired Databricks with Hightouch, which enabled XP to sync data and predictions directly into marketing tools, including advertising platforms, without duplicating sensitive information. With Databricks and Hightouch together, they could calibrate paid media bids based on the predicted investment behavior.
The results
Ad bidding driven by predictive AI results in a $38M payoff
To send these predicted investments to ad platforms, XP turned to Hightouch. Previously, XP had relied on client-side web pixels to send conversion data back to each platform when customers opened an account. XP replaced client-side tracking with server-side conversion APIs powered by Hightouch, and sends conversion data based on predicted investments rather than account creation. (Server-side tracking enables more comprehensive, privacy-compliant data sharing between systems—learn more in this blog).
Armed with these early signals, XP’s ad platforms could confidently bid more for high-quality leads and avoid overinvesting in low-value traffic. As a result, XP improved lead qualification by 62.5%, dramatically increased advertising efficiency, and drove an additional $38 million in revenue from performance media.
AI-powered audience segmentation adds $28M in investor revenue
XP could also use AI predictions to segment audiences in Hightouch based on likely future behaviors. For example, customers predicted to invest large sums could be routed to expert advisors or receive tailored onboarding content, while others received more education-focused journeys. This approach unlocked $28 million in incremental revenue by improving match quality between customers and advisors and delivering more relevant experiences at scale.
What’s next: scaling predictive AI across the business
XP is now expanding the reach of their predictive AI models to support more teams and use cases across the business. They also plan to enhance their AI models with second-party data from sources like Experian and the Brazilian Stock Exchange, plus partner data sharing with independent financial advisors via Delta Sharing and Clean Rooms. For internal business users, they are democratizing data access by using AI/BI Genie to make data more accessible.
With a strong data foundation and centralized infrastructure already in place, XP is moving faster, scaling more efficiently, and building solutions that adapt to changing market demands.