Mobile Dev Memo Podcast

Season 7, Episode 11: "Understanding Unity's Vector" (with Felix Tay)

Date: March 24, 2026
Host: Eric Soufert
Guest: Felix Tay, Unity Ads

Episode Overview

In this episode, host Eric Soufert sits down with Felix Tay, who has been with Unity for nine years, to dive deep into Unity’s AI-powered advertising platform, Vector. The conversation unpacks how Vector is transforming UA (User Acquisition) and monetization for mobile game developers by leveraging unique Unity engine data and advances in machine learning. The discussion also explores the broader impact of AI on creative strategy, game design, and developer tooling, as well as shifts in soft launch practices and the future of mobile gaming in the AI era.

Eric and Felix provide candid, insightful commentary about platform evolution, the challenges and opportunities for the mobile game industry, and the technical underpinnings of Unity's growth strategy. The tone is friendly, thoughtful, occasionally technical, but consistently practical for practitioners in mobile advertising and game development.

Key Discussion Points & Insights

1. Introduction to Vector: What It Is and Why It Matters

[02:40 - 05:12]

• What is Vector?
  • Felix describes Vector as “Unity’s AI powered growth and user acquisition platform. It is the fastest growing part of our business” (04:00).
  • Represents a complete rebuild of Unity's ML and data infrastructure for advertising.
  • Integrates gameplay, monetization, and campaign performance data into a “continuous learning model" tailored specifically for games.
• Why does it stand out?
  • Uniquely leverages Unity’s position as both engine and ad network: “We understand how games are built and how players actually behave inside them” (04:13).
  • Focuses on ingesting massive volumes of gameplay data, enabling rapid learning and better predictions.
  • Has driven strong revenue growth (72% YoY in January 2026).

Quote:
"Vector specifically designed to ingest and interpret very large volumes of gameplay related data... by pairing the gameplay inside with our ad network at scale, Vector can learn faster, make better predictions, translates into stronger results for developers and advertisers."
— Felix Tay, [04:11]

2. The Continuous Improvement & R&D of AI/ML Models

[05:12 - 09:08]

• Research at the Core:
  • Ongoing, never a single “big release”—improvements roll out incrementally.
  • Felix explains: “There's elements of the model that you need to update sequentially... We just saw the majority of the benefit happen in this big moment around Q2, Q3 of last year” (04:38).
• Model & Data Iteration:
  • Two primary axes: model improvement (better architectures or outcomes) and data expansion/augmentation.
  • “Based on the availability of data that you have and the way that data is represented...modeling also change, right?” (07:28).
• Regular, Iterative Deployments:
  • Even small changes (e.g., optimizer tweaks) can make measurable difference.
• Not Just a Version Number:
  • “The Vector that we released from 2025 is not the Vector that people experience today...every time there's a better way to do something...we release it" (08:45).

Quote:
"Every time there's a better way to do something, either through modeling or data, with modeling, we release it and we just let the customers experience the benefit."
— Felix Tay, [08:45]

3. Unique Power of Unity's Runtime Data

[09:48 - 16:27]

• Unity’s Advantage:
  • Sits at the intersection of creation and growth; has high coverage of mobile games (~70% of top 1000 use Unity).
  • Access to high-fidelity, time-ordered runtime data direct from the Unity engine, not just SDK event streams.
  • Resolves causality/sequencing issues inherent to SDK-based approaches: “Causality is important. If causality is broken, advanced modeling techniques...become moot” (13:03).
• Reduces Developer Burden and Error:
  • Standardized event collection, less risk of mistakes or missed events.
  • “The lack of standardization across events makes it really hard for AI products to use. So I think standardization and consistency...is what I would say one of the benefit of runtime data” (14:46).
• Unlocks Better Modeling:
  • Rich, ordered, and clean data sets drive more sophisticated sequential and cohort-based modeling.

Quote:
"The power of the Unity runtime...is coverage [and] quality of data signal...we get access to the cleanest quality data straight from the runtime and the boot time when applications got loaded."
— Felix Tay, [11:26]

4. Rationale and Timing for Building Vector

[17:27 - 19:05]

• Why Now?
  • Unity has had an ads product for years, but several factors converged for Vector: new modeling techniques, new data/information, better ML frameworks, and next-gen inference tech.
  • “You're flying a plane when you're building a new one at the same time and you need to find a way and chart a course for a smooth cutover. It's not simple” (17:47).
  • “It became clear that rebuilding was the right course of action...we're pretty pleased with the outcome" (18:56).

5. Big Models and the Future of Game Ads ML

[19:05 - 21:22]

• Preference for Big Unified Models:
  • Felix: “I believe in big models, not small...the future is about big models” (19:22).
  • Big models allow for shared understanding, autocorrelation, and efficient inference.
• Sequential/Personalized Prediction is the Goal:
  • Unlock comes with using full player data sequences to predict “what’s next.”
  • Advances in inference and compute cost management now make this practical.

6. Vector’s Roadmap & Future AI Workflows

[21:12 - 23:23]

• Next Steps:
  • Integrate more runtime data to unlock further gains (“early signals that this is something that we can utilize” [21:22]).
  • Broaden AI utility from predictive ML to intent-driven workflows, letting campaign or monetization goals be expressed as pure intent for AI to execute.
  • Continued push toward bigger, more powerful models.

7. AI's Transformational Impact on Creative Strategy

[23:54 - 29:54]

• AI-Empowered Creative Production:
  • Huge increase in speed and variety of creative asset generation (“It’s very liberating…AI reshapes the creative strategy for gaming advertisers...” [24:13]).
• BUT — Risk of Shifted Costs:
  • GenAI can just move bottlenecks from asset production to testing; testing budget can become the limiting factor.
  • Felix: “You want to actually get productivity without the added cost of marketing. So...the right kind of AI driven creative is...built with performance advertising in mind” (25:23).
• True Innovation is in Personalization:
  • Future is not just volume, but narrowing creative to small, highly-targeted cohorts for higher conversion.
  • "The holy grail of creative is...truly personalized because then people will convert" (27:32).
• Unity’s Approach:
  • Emphasis on outcome (better ROAS), not just more assets or spend: “We want them to actually be able to spend and achieve success. So for us...the KPI for a good creative gen AI is better OS that simple?” (29:05).

Memorable Exchange:
Eric: “We’ve just got this machine gun now...we’re not being very strategic with how we deploy it.”
Felix: "Meaning like just spread and then hopefully one of them will land...We don’t want to transfer their cost of creative to the cost of marketing."
[28:46 – 29:36]

8. Emerging Trends in Game Design & Soft Launch

[29:54 - 34:32]

• Shift in Soft Launch Optimization:
  • Traditional process: build full game, soft launch, tweak if KPIs are off—risky and costly.
  • New process: test vertical slices (playables) earlier to measure engagement, fail and pivot faster, reduce sunk R&D cost.
  • “The thing that will make a lot of impact...is minimizing your pivot and your change at the 11th hour...get that signal early as they're still iterating and developing the game” (30:18).
  • If engagement score is weak, “the cost of pivoting is very low” (31:37).
• Industry is Moving Toward Failing Fast:
  • Collect real player data earlier, avoid investing deeply until a concept is validated.

9. AI in Unity’s Developer Tooling

[35:05 - 41:48]

• Unity’s Philosophy:
  • Open ecosystem; Unity as “assembly point” for any tool, first or third-party (35:25).
  • First party Unity AI Assistant tailored for game development tasks (e.g. building NPCs); keeps inference cost low and domain-specific (“It is called LLMs, I guess for a reason. Sometimes...using a large language model...is a bit of an overkill” [36:53]).
  • AI assistance focused on keeping “human in the loop," incremental edits vs. regenerating from scratch.
  • “You want to change a small little bit but don’t touch anything else outside...That requires a deep understanding of how human and AI interactions [work] and that’s our environment” (39:29).
• Cautions About GenAI Practices:
  • Repeatedly “one-shotting” code/games is infeasible for larger projects due to cost and lack of scalability.
  • “The reality of video games, it’s an enterprise grade entertainment...all these things require thought...you can’t one shot a game [at] professional grade” (39:29).

10. Optimism for the Future of Mobile Gaming

[41:54 - 43:28]

• Mobile as the Natural Home for Fun:
  • “At its core, gaming has never been about having the best graphics…It’s about fun. Mobile...continues to be probably the best place for fun to show up” (41:54).
• Strengths of Mobile Platform:
  • Penetration, flexibility in monetization (ads/IAP/paid), robust discovery, technical capability, and scale.
  • Unity is positioned at the intersection enabling this at scale.
• AI to Further Proliferate Content:
  • “With content generation, with AI becoming more and more accessible, more and more devices will get entertainment content."

Notable Quotes & Memorable Moments

• On the uniqueness of Unity’s data:
  “Our runtime data technically doesn’t have to worry about quality issues...”
  — Felix Tay, [11:47]

• On Class Imbalance in Monetization:
  “If your classifier just says everyone's not going to make an IAP, it's correct 99% of the time…But it's wrong for the group of users we care about.”
  — Eric Soufert, [16:57]

• On the big model approach:
  “The future is about big models, how to optimize your cost on the big model and not be scared...”
  — Felix Tay, [20:47]

• On true innovation in creative:
  “I think the holy grail of creative is like a combination...where one day you could narrow it down to a very small cohort and it's truly personalized because then people will convert.”
  — Felix Tay, [27:32]

• On shifting creative testing burdens:
  "You may have convinced yourself that this is very productive because just look at the number of assets we’re creating...but actually just shift in the cost elsewhere and you’re probably not better off."
  — Eric Soufert, [29:36]

• On soft launch evolution:
  “Minimizing your pivot and your change at the 11th hour...some developers have started to...not actually follow the traditional soft launch...they try to minimize the cost of development by porting some existing title [and] try to launch that as a playable unit...”
  — Felix Tay, [30:18]

Timestamps for Key Segments

• Introduction to Felix Tay / Unity background: [01:12 – 02:22]
• What is Vector and why now? [02:40 – 09:08]
• Unity’s unique data and modeling advantages: [09:48 – 16:27]
• Big models vs. small models—future of ML in ads: [19:05 – 21:22]
• Vector’s roadmap, using runtime data: [21:12 – 23:23]
• AI and creative production – opportunities & pitfalls: [23:54 – 29:54]
• Shifts in game design & soft launch best practices: [29:54 – 34:32]
• Unity’s support for AI in development tooling: [35:05 – 41:48]
• Optimism for mobile gaming & closing thoughts: [41:54 – 43:28]

Conclusion

This episode of Mobile Dev Memo offers a comprehensive, behind-the-scenes look at Unity’s major AI-powered push in mobile advertising via Vector, with practical insights spanning AI/ML, data strategy, creative, game design, and the developer experience. Felix Tay’s technical and product knowledge, combined with Eric Soufert’s probing questions and industry context, makes this episode essential listening—or reading—for anyone interested in the business and technology of mobile gaming.