Interviews
Aman Sareen, CEO of Aarki – Interview Series
Aman Sareen is the CEO of Aarki, an AI company that delivers advertising solutions that drive revenue growth for mobile app developers. Aarki allows brands to effectively engage audiences in a privacy-first world by using billions of contextual bidding signals coupled with proprietary machine learning and behavioral models. Working with hundreds of advertisers globally and managing over 5M mobile ad requests per second from over 10B devices, Aarki is privately held and headquartered in San Francisco, CA with offices across the US, EMEA, and APAC.
Could you share a bit about your journey from co-founding ZypMedia to leading Aarki? What key experiences have shaped your approach to AI and AdTech?
My adtech leadership odyssey began with co-founding ZypMedia in 2013, where we engineered a cutting-edge demand-side platform tailored for local advertising. This wasn't just another DSP; we built it from the ground up to handle high-volume, low-dollar campaigns with unprecedented efficiency. Think of it as the precursor to the hyper-localized, AI-driven targeting we see today.
As CEO, I steered ZypMedia to $20 million in SaaS revenue and processed $200 million in media transactions annually. This experience was a crucible for understanding the sheer scale of data that modern ad platforms must handle — a challenge tailor-made for AI solutions.
My stint at LG Ad Solutions, post-ZypMedia’s acquisition by Sinclair, was a deep dive into the world of device manufacturers and how the control of viewership data can shape the future of Connected TV (CTV) advertising. We used a lot of AI/Machine learning in building the LG Ads business, where the data collected from devices was used to generate targeting segments, inventory blocks, and planning software.
As CEO of Aarki since 2023, I'm at the forefront of the mobile advertising revolution. I can say that my journey has instilled in me a profound appreciation for the transformative power of AI in adtech. The progression from basic programmatic to AI-driven predictive modeling and dynamic creative optimization has been nothing short of remarkable.
I've come to see AI not just as a tool but as the backbone of next-generation adtech. It's the key to solving the industry's most pressing challenges; from privacy-compliant targeting in a post-device ID world to creating genuine and personalized ad experiences at scale. I firmly believe that AI will not only solve the pain points the advertisers face but also revolutionize how operations are run at platforms like Aarki. The lessons from my journey — the importance of scalability, data-driven decision-making, and continuous innovation — are more relevant than ever in this AI-first era.
Can you elaborate on how Aarki's multi-level machine-learning infrastructure works? What specific advantages does it offer over traditional adtech solutions?
My experiences have taught me that the future of adtech lies in harmonizing big data, machine learning, and human creativity. At Aarki, we explore how AI can enhance every aspect of the mobile advertising ecosystem; from bid optimization and fraud detection to creative performance prediction and user acquisition strategies.
At this stage, Aarki's multi-level machine learning infrastructure is designed to address several critical aspects of mobile advertising, from fraud prevention to user value prediction. Here's how it works and why it's advantageous:
- Fraud Detection and Inventory Quality Control: It's designed to protect our clients' performance and budgets. Our multi-layered approach combines proprietary algorithms with third-party data to stay ahead of evolving fraud tactics. We ensure campaign budgets are invested in genuine, high-quality inventory by constantly evaluating user behaviors and maintaining an up-to-date fraud database.
- Deep Neural Network (DNN) Models: Our core infrastructure utilizes multi-stage DNN models to predict the value of each impression or user. This granular approach allows each model to learn features most crucial for specific conversion events, enabling more precise targeting and bidding strategies compared to one-size-fits-all models.
- Multi-objective Bid Optimizer™ (MOBO): Unlike simple bid shading used by most DSPs, our MOBO considers multiple factors beyond price. It uses dynamic variables such as campaign and inventory attributes, predicted user value, and CPM segmentation to optimize bids. This sophisticated method maximizes ROI while balancing multiple objectives, finding optimal bids that win, meet KPI goals, and pace correctly to utilize campaign budgets fully.
These components offer significant advantages over traditional AdTech solutions:
- Superior fraud detection
- More accurate predictions and better ROI through multi-stage DNNs
- Granular creative hyper-targeting with multi-objective bid pricing
- Scalability to handle vast amounts of data
- Privacy-first targeting with contextual cohorts
Our AI-driven approach allows for unprecedented accuracy, efficiency, and adaptability in mobile advertising campaigns. By leveraging deep learning and advanced optimization techniques, Aarki delivers superior performance while maintaining a strong focus on privacy and fraud prevention.
How does the Dynamic Multi-object Bid Optimizer function, and what impact does it have on maximizing ROI for your clients?
The Dynamic Multi-object Bid Optimizer is a sophisticated system that goes beyond traditional bid shading algorithms. Unlike simple bid shading algorithms that focus solely on pricing just under the predicted winning bid, our optimizer considers multiple objectives simultaneously. This includes not just price but also campaign performance metrics, inventory quality, and budget utilization.
The optimizer takes into account a range of dynamic variables, including campaign and inventory attributes, predicted user value, and CPM segmentation. These variables guide the optimization process around client-specific KPIs, primarily ROI. This allows us to tailor our bidding strategy to each client's unique goals.
One of the key strengths of our optimizer is its ability to balance between acquiring high-value users efficiently and exploring new, untapped user segments and inventory. This exploration helps us discover valuable opportunities that more rigid systems might miss.
In practice, this means our clients can expect more efficient use of their ad spend, higher-quality user acquisition, and, ultimately, better ROI on their campaigns. For example, it might make sense to pay 50% more to bid for a user who is 5 times more valuable (ROAS). The optimizer's ability to balance multiple objectives and adapt in real-time allows us to navigate the complex mobile advertising landscape more effectively than traditional, single-objective bidding systems.
Aarki emphasizes a privacy-first approach in its operations. How does your platform ensure user privacy while still delivering effective ad targeting?
I'm proud to say that privacy-first engagement is one of the core pillars of our platform, along with our AI platform. We've embraced the challenges of the no-device-ID world and developed innovative solutions to ensure user privacy while delivering effective ad targeting. Here's how we accomplish this:
- ID-less Targeting: We've fully adapted to the post-IDFA landscape and are SKAN 4 compliant. Our platform operates without relying on individual device IDs, prioritizing user privacy from the ground up.
- Contextual Signals: We leverage a wide array of contextual data points such as device type, OS, app, genre, time of day, and region. These signals provide valuable targeting information without requiring personal data.
- Massive Contextual Data Processing: We process over 5 million ad requests per second from over 10 billion devices globally. Each request has a wealth of contextual signals, providing us with a rich, privacy-compliant dataset.
- Advanced Machine Learning: Our 800 billion row training model database correlates these contextual signals with historical outcome data. This allows us to derive insights and patterns without compromising individual user privacy.
- Dynamic Behavioral Cohorts: Using machine learning, we create highly detailed, dynamic behavioral cohorts based on aggregated contextual data. These cohorts enable efficient optimizations and scaling without relying on personal identifiers.
- ML-driven Creative Targeting™: For each cohort, we use machine learning in collaboration with our creative team to devise optimal creative strategies. This approach ensures relevance and effectiveness without infringing on individual privacy.
- Continuous Learning and Adaptation: Our AI models continuously learn and adapt based on campaign performance and evolving contextual data, ensuring our targeting remains effective as privacy regulations and user expectations evolve.
- Transparency and Control: We provide clear information about our data practices and offer users control over their ad experiences wherever possible, aligning with privacy best practices.
By leveraging these privacy-first strategies, Aarki delivers effective ad targeting while respecting user privacy. We've turned the challenges of the privacy-first era into opportunities for innovation, resulting in a platform that's both privacy-compliant and highly effective for our clients' user acquisition and re-engagement campaigns. As the digital advertising landscape evolves, Aarki remains committed to leading the way in privacy-first, AI-driven mobile advertising solutions.
Can you explain the concept of ML-driven Creative Targeting™ and how it integrates with your creative strategy?
ML-driven Creative Targeting™ is our methodology for optimizing ad creatives based on the behavioral cohorts we identify through our machine learning models. This process involves several steps:
- Cohort Analysis: Our ML models analyze vast amounts of contextual data to create detailed behavioral cohorts.
- Creative Insights: For each cohort, we use machine learning to identify the creative elements that are likely to resonate most effectively. This could include color schemes, ad formats, messaging styles, or visual themes.
- Collaboration: Our data science team collaborates with our creative team, sharing these ML-derived insights.
- Creative Development: Based on these insights, our creative team develops tailored ad creatives for each cohort. This might involve adjusting imagery, copy, calls-to-action, or overall ad structure.
- Dynamic Assembly: We use dynamic creative optimization to assemble ad creatives in real-time, matching the most effective elements to each cohort.
- Continuous Optimization: As we gather performance data, our ML models continually refine their understanding of what works for each cohort, creating a feedback loop for ongoing creative improvement.
- Scale and Efficiency: This approach allows us to create highly targeted creatives at scale without the need for manual segmentation or guesswork.
The result is a synergy between data science and creativity. Also one of our core pillars, Unified Creative Framework, ensures that our ML models provide data-driven insights into what works for different audience segments. At the same time, our creative team brings these insights to life in compelling ad designs. This approach enables us to deliver more relevant, engaging ads to each cohort, simultaneously improving campaign performance and user experience.
What role does your creative team play in developing ad campaigns, and how do they collaborate with the AI models to optimize ad performance?
Our creative team plays an integrated role in developing effective ad campaigns at Aarki. They work in close collaboration with our AI models to optimize ad performance. The creative team interprets insights from our ML models about what resonates with different behavioral cohorts. They then craft tailored ad creatives, adjusting elements like visuals, messaging, and formats to match these insights.
As campaigns run, the team analyzes performance data alongside the AI, continuously refining their approach. This iterative process allows for rapid optimization of creative elements.
The synergy between human creativity and AI-driven insights enables us to produce highly targeted, engaging ads at scale, driving superior performance for our clients' campaigns.
How does Aarki's AI infrastructure detect and prevent ad fraud? Can you provide some examples of the types of fraud your system identifies?
As I mentioned earlier, Aarki employs a multi-layered approach to combat ad fraud. We are approaching fraud deterrence as a pre-bid filter with post-bid analytics of the data that comes through our systems. While I've already outlined our general strategy, I can provide some specific examples of the types of fraud our system identifies:
- Click flooding: Detecting abnormally high click rates from specific sources.
- Install farms: Identifying patterns of multiple installs from the same IP address or device.
- Abnormal click-to-install time (CTIT): Spotting abnormal click-to-install times as a signal for bot activity.
- Low Retention Rates: Identifying users from publishers that repeatedly exhibit low retention rates after install.
Our AI continuously evolves to recognize new fraud tactics, protecting our clients' budgets.
How does Aarki’s approach to user acquisition and re-engagement differ from other platforms in the industry?
Aarki's approach to user acquisition and re-engagement sets us apart in several key ways:
- Privacy-First Strategy: We've fully embraced ID-less targeting, making us SKAN 4 compliant and future-ready in a privacy-focused landscape.
- Advanced AI and Machine Learning: Our multi-level machine learning infrastructure processes vast amounts of contextual data, creating sophisticated behavioral cohorts without relying on personal identifiers.
- ML-driven Creative Targeting™: We uniquely combine AI insights with human creativity to develop highly targeted ad creatives for each cohort.
- Dynamic Multi-object Bid Optimizer: Our bidding system considers multiple objectives simultaneously, balancing efficiency with exploration to maximize ROI.
- Contextual Intelligence: We leverage trillions of contextual signals to inform our targeting, going beyond basic demographic or geographic segmentation.
- Continuous Optimization: Our AI models continuously learn and adapt, ensuring our strategies evolve with changing user behaviors and market conditions.
- Unified Approach: We offer seamless integration of user acquisition and re-engagement strategies, providing a holistic view of the user journey.
- Scalability: Our infrastructure can handle immense data volumes (5M+ ad requests per second from 10B+ devices), enabling highly granular targeting at scale.
- Advanced Fraud Deterrence Mechanisms: Our in-house pre-bid fraud filters, post-bid analytics of massive data volumes, combined with 3rd-party data, put us at the forefront of saving our clients’ money from fraudulent traffic.
This combination of privacy-centric methods, advanced AI, creative optimization, fraud deterrence, and scalable infrastructure allows us to deliver more effective, efficient, and adaptable campaigns.
My experiences have taught me that the future of ad tech lies in harmonizing big data, machine learning, and human creativity. I take pride in the fact that, in addition to our technology, we also have an outstanding team of analysts, data scientists, and creative professionals who add human creativity to our tech.
Could you share some success stories where Aarki’s platform significantly improved client ROI and campaign effectiveness?
The AppsFlyer Performance Index recognizes Aarki as a leader in retargeting, ranking us #1 for gaming in North America and #3 globally. We are also rated as a top performer across all Singular advertising ROI indexes. This case study is also a testament to our global leadership. Not just for gaming, but we have recent case studies showcasing our ability to drive results across various app categories.
I'm proud to highlight our partnership with DHgate, a leading e-commerce platform. Our retargeting campaigns for both Android and iOS delivered exceptional results, showcasing Aarki's ability to drive performance at scale.
Leveraging our deep neural network technology, we precisely assembled user segments to maximize retargeting effectiveness. This resulted in a 33% growth in higher-intent user clicks and a 33% increase in conversions.
Most impressively, while DHgate's spend with Aarki increased by 52%, we consistently exceeded their 450% D30 ROAS goals by 1.7x, achieving an outstanding 784% ROAS. This case exemplifies our commitment to delivering superior results for our clients. Read more about it here.
For a food and delivery app, we implemented a retargeting campaign to reactivate users and acquire new customers efficiently.
This resulted in a 75% decrease in Cost Per Acquisition (CPA) and 12.3 million user reactivations. The key to success was utilizing our Deep Neural Network models to target the right audiences with tailored messaging, keeping the campaign fresh and engaging. Read it here.
These case studies demonstrate our ability to drive significant improvements in key metrics across different app categories and campaign types. Our privacy-first approach, advanced AI capabilities, and strategic use of contextual data allow us to deliver outstanding results for our clients, whether in user acquisition or re-engagement efforts.
What future advancements in AI and machine learning do you foresee as pivotal for the mobile advertising industry?
Looking ahead, I anticipate several pivotal advancements in AI and machine learning for mobile advertising:
- Enhanced privacy-preserving techniques: The massive scale of data we process will lead to unprecedented learning capabilities. Deep neural networks (DNNs) will leverage this to create superior privacy-first engagement strategies. In fact, the concept of “targeting” will evolve so dramatically that we'll need new terminology to describe these AI-driven, predictive approaches.
- Generative AI for real-time creative optimization: We'll see AI that can not only optimize but also create and dynamically modify ad creatives in real-time. This will revolutionize how we approach ad design and personalization.
- Holistic Predictive Models: By combining our deep neural networks for product insights with our Multi-Objective Bid OptimizerTM (MOBO) for pricing, we'll develop highly effective and efficient models for both user acquisition and retargeting. These will provide incredibly accurate predictions of long-term user value, allowing for smarter, more strategic campaign management.
These advancements will likely lead to more effective, efficient, and user-friendly mobile advertising experiences.
Thank you for the great interview, readers who wish to learn more should visit Aarki.