Case Study: Cisco achieves 15x faster propensity-to-buy model production with H2O.ai

Cisco cuts model production from a month to 2 days with H2O.ai

Cisco's advanced analytics team faced the challenge of rebuilding and deploying 60,000 propensity-to-buy models every quarter. Their process was slow, taking over a month, and relied on small data samples and simple decision trees, which limited prediction accuracy for scoring their database of 160 million companies. To keep pace with shifting demand, they needed a faster, more powerful solution and turned to vendor H2O.ai for its H2O-3 open-source machine learning platform.

H2O.ai's solution enabled Cisco to implement a new model factory using H2O's in-memory distributed engine. This allowed the team to train on their entire dataset using more advanced algorithms like GLM, random forests, GBM, and deep learning. The result was a 15x faster model production time, cut from a month to just two days. Cisco now deploys 60,000 highly accurate models per quarter, work that has been associated with influencing over $3 billion in revenue since the factory's inception.

Cisco