I provide consulting services to firms seeking to manage their data and gain insights about their customer's behavior, including satisfaction, interactions, subscriptions and microtransactions. Most of this work is with video game developers managing game populations and seeking to improve the player experience, but the models are widely applicable and adaptable.
I work with collaborators in computer science and social networking to develop the best possible tools as technology advances. At small scales we use standard, proven techniques drawn from the social sciences and our own research. These include regression models, surveys and experimental models (sometimes called AB testing).
At larger scales (2TB+), we use more heavy-duty data analysis using machine learning, social network analysis, heat map visualizations, and regression tools. We are currently developing cloud-scale computing techniques for very large data loads (100+ TB).
Subscription prediction modeling: Who is likely to leave? Why did that person leave? How can I keep them?
Social network modeling: Who are my valuable people--not just because they are valuable on their own, but because they are what keeps others there. Conversely, who is driving others away?
What are the probabilities that person X will leave? Why?
How can I fix these things: What specific elements of the game/service are causing these outcomes (both good and bad)?
Who is generating the most income? What predicts and causes this, and can it be adapted for a larger group?
Email is the simplest: dmitri DOT williams AtSign usc DOT edu