Your AI partner for scalable ROAS optimization—Samba 2.5 is here
AI — 26 Jun 2025
How SplitMetrics United AI & Experts to Empower User Acquisition Teams
SplitMetrics 
The rapid expansion of artificial intelligence (AI) has come at just the right time for app marketing. User acquisition teams must accelerate growth while facing increased competition and market complexity than ever before. They must push their limits to deliver expected results.
To answer this challenge, we created Samba. It’s our expert-trained family of AI models, built to empower app marketers in performance optimization and break any limits in app growth, allowing for better ROAS and profitability of user acquisition.
Samba is a key milestone for the SplitMetrics AI research and development center, and a logical next step in our mission to enable profitable growth for mobile apps. One we’ve been following for more than a decade now.
Samba introduces a new workflow for user acquisition. In it, AI handles high-speed data processing and decision-making, while experts guide, review, and steer strategy. With strong results, such as ROAS or revenue improvements of 55% in some cases, we believe that this workflow of human-AI unity and collaboration is the future of user acquisition.
Read on to learn how we established it and how you can empower your app marketing team with it as well.
Effective collaboration between UA experts and AI will be crucial in addressing the evolution of the mobile app market, a key industry on a global scale. It’s projected to grow at a compound annual growth rate (CAGR) of 14.3% until 2030, according to Grand View Research.
The App Store alone is estimated to have over 43,000 new apps added to it monthly. At the same time, each mobile app category presents a unique competitive environment, characterized by its seasonal trends, average performance metrics, and varying user behavioral patterns that influence app discovery and consequently, app marketing.
The consequence? Information overload remains a persistent concern in online marketing, as highlighted by Bloomberg, Forbes, and Adverity, which report that 67% of CMOs admit to being overwhelmed by the volume of data.
Logically, as the AI Marketing Institute 2025 State of Marketing AI Report informs, 82% of marketers want AI’s help with daily, data-driven tasks, 65% need more actionable insights from data, and 63% wish to accelerate revenue growth.
The increasing number of signals that need to be processed for effective decision-making and then effectively implemented may give even the best app marketing leaders pause unless AI backs them.
Empowering UA teams with technology to address the increasingly diversified, nuanced, and dynamic market has always been our mission. Samba, as a family of AI models for managing bids and budgets in Apple Ads, is a logical leap forward for us.
Currently, it analyzes over 700,000 keywords daily to optimize approximately 250,000 bids, relying on more than 30 million data points (source: SplitMetrics AI). Consequently, it’s capable of delivering a staggering 55% improvement in ROAS in some cases. As an AI, Samba can extract valuable signals from a vast amount of market and ad campaign data. It can detect anomalies before they impact performance and suggest optimizations in real-time.
This efficiency is grounded in human expertise and a long history of development of data-driven app marketing solutions.
Samba is built on a decade of marketing innovation, primarily in Apple Ads automation. SplitMetrics Acquire, our platform for managing and optimizing Apple Ads, was launched the same year as the advertising platform (then known as Apple Search Ads).
This experience was pivotal in enabling us to develop our artificial intelligence models. Robust automation and market intelligence features drew top apps to our platform. This enabled us to aggregate extensive and high-quality data on keyword and ad performance, which is critical for training AI. The high standard for data was strengthened by our joining the Apple Ads Partnership program in 2019.
The founding of SplitMetrics’ Agency in 2021 was also critical, as it enabled us to establish our user acquisition team, providing us with the best possible understanding of how UA teams operate on a day-to-day basis.
SplitMetrics has been continuously evolving to stay one step ahead of the market, which has ultimately given us a unique opportunity to develop our own AI models. We had all the assets:
We were ready to push our research and development further.
SplitMetrics AI, our research & development center for building AI technologies designed to solve marketing performance challenges for mobile-first companies, was established in 2022. It gathered Data Scientists (including multiple PhDs), Machine Learning Engineers, and specialists who have managed multi-million dollar budgets for user acquisition.
The first achievement of SplitMetrics AI was the introduction of condition-based automation strategies for bidding. It was a significant milestone that provided advertisers with a model analyzing past performance to optimize bidding for a particular target.
Condition-based bid strategies represented a significant advancement over rule-based automation. Their capabilities make them a valuable intermediate solution between rule-based automation and AI, as a broad segment of apps could easily adopt them.
Condition-based strategies proved to be effective, but to remain ahead of the market, we knew a leap forward was necessary. Performance analysis of these strategies showed us a clear direction and the next steps required to create more advanced models:
To achieve higher levels of accuracy, our data science team sought to consider data points and market signals coming from beyond the App Store. To remain a step ahead of the dynamically changing market, we recognized that our models would also need to predict changes and act on them before they occur.
The launch of Samba, our artificial intelligence model family, took place in 2023. Driven by the data science team, with significant involvement from our user acquisition experts, we developed a system of prescriptive and predictive AI models.
The collaboration between UA experts and Data Scientists was critical to preparing Samba to monitor auction dynamics at a keyword level, just as a human specialist would do. This meant examining:
Research and training extended even to datasets beyond Apple Ads. Experts’ participation helped Samba generalize across various verticals, including education, finance, and gaming.
Additionally, our UA experts helped us establish a workflow relying on effective collaboration with the AI, while maintaining control and transparency.
From a UA team’s perspective, Samba is a logical evolution of existing automation workflows in SplitMetrics Acquire. This means that they retain control over campaigns through various goal-related settings and can review all of Samba’s activities, due to individual keyword changes being accessible in our Ads Manager, along with all other campaigns.
Samba will empower your team in the same way it now empowers our experts, who have contributed significantly to its accuracy and reliability. Here are the principles:
In essence, by minimizing setup and configuration, Samba and its new workflow take automation to a new level. At the same time, its scope aligns with current UA workflows established using our platform. It optimizes performance while remaining transparent and responsive.
Samba can make 10x more bid adjustments to a single keyword in the same period than manual changes, an advantage over even the most sophisticated automation-rule configurations. Overall, Samba 2.5 can save more than 40 hours every month, adjusting bids 14 times faster, compared to manual optimization over 90 days.
However, from a UA specialist’s practical perspective, all methods that we offer can provide a satisfying degree of automation, although Samba and condition-based strategies are much easier to set up and maintain.
Automation configuration time in Samba 2.5 can be completed in under 2 minutes. All settings can be verified & changed just as quickly.
In this context, we believe that Samba’s most significant value lies in the accuracy of its prescriptive and predictive capabilities to optimize bids. It relies on 30 million data points daily to run Apple Ads campaigns, making those bid adjustments when needed, and to best effect regarding driving goals and ROAS.
Effectively, Samba, as an AI solution, addresses data overload and complexity. It’s capable of making fast and accurate decisions based on goals set by the advertiser. With predictive modeling, it can also forecast how each keyword will perform.
Automation still matters, as liberating time for more strategic activities is and will be increasingly essential for any user acquisition team. However, our AI will, first and foremost, create a competitive advantage for app marketers, allowing faster decisions, more confident scaling, and improved profitability.
Human expertise played a crucial role in the development of our AI. We believe both will be critical to app growth in the future.
For us, Samba is an accelerator, the next logical step in Apple Ads automation, liberating UA specialists to further shift their attention to the strategic and creative aspects of optimization, which has been a goal of using our platform for countless advertisers since its inception.
Samba is also just a beginning, as we’re building toward a vision where every step of the UA workflow has a supporting AI working alongside your team.
Are you ready to see Samba in action? Let’s talk. We’ll show you how expert-trained AI can power your next phase of growth – on Apple Ads and beyond.
