Case Study: Cisco achieves 15x faster production of 60,000 predictive models with H2O.ai

Cisco - Customer Case Study

Cisco’s 20-person advanced analytics team runs a Predictive Model Factory that builds 60,000 propensity-to-buy models each quarter for Cisco’s 160 million-company database, but faced severe speed and scalability limits: recreating all models from scratch took more than a month, forcing small training samples and constraining algorithm choice. To modernize the pipeline, Cisco adopted H2O.ai’s open-source in-memory distributed machine learning platform (H2O), integrating it into an R-controlled process to streamline data prep, training and scoring.

Using H2O.ai’s H2O, Cisco moved to ensemble, GBM and deep-learning workflows and scaled from training on ~100k samples to using tens of millions of records, enabling full scoring of 160 million companies. The result: production time dropped from over a month to two days (≈15x faster), 60,000 models are produced each quarter with higher accuracy, new buying patterns are incorporated immediately, and the P2B factory has helped drive more than $3B associated revenue—delivered with fewer technical resources thanks to H2O.ai.

Cisco

Lou Carvalheira

Advanced Analytics Manager