Behavioral signals and the art of ethical targeting - James Dickens (Gamelight)

James Dickins, Team Lead of Technical Support at Gamelight, joins Taylor Lobdell to discuss how rewarded UA and behavioral targeting shape user acquisition in 2025’s privacy-first landscape. From the company’s laser-eyed cat mascot to the mechanics of lookalike models, James explains how his team balances precision with ethics and how algorithms are trained on aggregated behavior data. He also discusses creative testing and when human intuition should override machine logic. He breaks down the real limits of automation, why model decay demands constant retraining, and how to build campaigns that adapt as fast as user behavior changes.

Questions that James answered in this episode:

• What is Gamelight, and how does its rewarded game recommendation model work?
• How does the team identify and replicate high-value users?
• What causes lookalike model decay, and how do you avoid overfitting?
• How can UA teams respect privacy while maintaining performance?
• What signals matter most when predicting player retention?
• How can creative teams and data teams collaborate more effectively?
• What are the most common mistakes UA managers make when scaling campaigns?
• How do rewarded ads reshape consent and engagement models?
• What role does human judgment play in interpreting algorithmic outputs?
• How will privacy and regulation continue to shape UA in the next five years?

Timestamps

• (0:00) – Intro: James Dickins and Gamelight overview
• (0:54) – The story behind Gamelight’s laser cat mascot
• (1:35) – What makes a high-value player, and how to identify them
• (2:04) – Using behavior profiles to guide acquisition strategy
• (2:55) – Lookalike models and the value of aggregated data
• (3:51) – Avoiding overfitting and model decay
• (4:50) – The surprise of discovering unexpected audience segments
• (5:20) – Privacy-first UA and ethical targeting
• (6:24) – Rewarded UA and consent-based engagement
• (6:30) – Managing ad fatigue and creative burnout
• (7:12) – When human intuition beats the algorithm
• (8:25) – Balancing optimization with experimentation
• (9:45) – Measuring engagement and long-term retention
• (10:17) – Designing for compliance before regulation hits
• (10:57) – Treating creative testing as data science
• (11:42) – Building the feedback loop between creative and data teams
• (12:52) – What keeps UA leaders up at night
• (13:07) – Predicting the future of user acquisition
• (17:35) – Wrap-up: how to connect with James

Selected quotes

• (3:40) – “A strong lookalike model is always learning and changing, predicting not just who might install but who will stick around.”
• (4:24) – “Lookalike models are amazing, but they’re not magic. Model decay happens, and what worked last quarter might fail today.”
• (5:59) – “Precision targeting and privacy can feel like opposites, but they can work together when you focus on aggregated signals.”

Mentioned in this episode

• Gamelight
• Archers.io
• James on Linkedin