Nov 20, 2014 · 22m · mad
Michael Rubenstein and Catherine Williams, App Nexus // Data Driven #31 // Nov 2014
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this presentation at Data Driven NYC, AppNexus executives Michael Rubenstein and Catherine Williams explore the scale, structural evolution, and real-world algorithmic applications of data science within the programmatic advertising ecosystem. They detail how AppNexus transitioned from makeshift analytics into an enterprise platform processing tens of billions of daily transactions through advanced machine learning and streaming technologies.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 2% of the talking time here. How this is scored →
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Catherine challenges audience member Carter's assumption regarding truth-telling in repeated second-price auctions, noting she does not consider it an ideal end-state.
Hardest push from Matt ▶ 21:58 Rephrasing Audience QuestionMatt Turck steps in to rephrase and clarify Ariel's audience question about click-through rate improvements after Catherine asks for clarification.
Biggest teaching moment ▶ 12:48 Domain Splitting via Google CorpusCatherine details how her team used mathematical word-splitting probabilities based on the Google trillion-word corpus to eliminate false positive porn flags.
Matt holds his own ▶ 17:55 Host Moderating Audience Q&AMatt Turck steps back in smoothly to transition from the presentation phase directly into audience Q&A while keeping time.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| Catherine Williams' Background, Dark Ages, and Renaissance of Data | 0 | 1 | 0 | 0 | Catherine Williams delivers a solo presentation on her background in pure mathematics and her transition to AppNexus quants during a data renaissance. The host does not participate in this monologue segment. | |
| Case Study: Pacing Algorithm for Even Budget Delivery | 0 | 1 | 0 | 0 | Catherine presents a case study explaining her first project at AppNexus, creating a pacing algorithm lever across 1500 disconnected bidder instances. The host is inactive during this presentation segment. | |
| Case Study: Programmatic Adult Content Detection and Word Splitting | 0 | 1 | 0 | 0 | Catherine illustrates how adult content detection was improved through probabilistic domain word splitting using the Google trillion-word corpus. The host does not speak in this monologue segment. | |
| The Modern Era of Data Science at AppNexus | 0 | 1 | 0 | 0 | Catherine outlines the modern infrastructure of data science at AppNexus, including Hadoop clusters and real-time streaming tools. The host is silent throughout this segment. | |
| Looking Forward: Visual Metaphor for Predictive Systems | 1 | 3 | 2 | 1 | Matt Turck opens the floor for audience Q&A, where Catherine and Michael answer technical questions on auction theory and click-through rates. Catherine gently reframes audience questions about truth-telling mechanisms and long-term incentives. |