Case Study: Pixability achieves 90% accuracy on large-scale multi-class NLP for optimized video ad targeting with Snorkel AI

Pixability - Customer Case Study

Pixability, a leading YouTube and Connected TV ad platform, needed to expand its NLP capabilities to help customers maximize reach and optimize video ad spend but was blocked by the time-consuming process of manually labeling high‑cardinality training data. They worked with Snorkel AI, using Snorkel Flow to address that bottleneck.

Snorkel AI used Snorkel Flow to distill knowledge from foundation models and keyword analysis into programmatic supervision, generating 500k labels with zero ground truth and rapidly building deployable classifiers. The vendor produced a 600+ class multi‑label NLP model and achieved over 90% accuracy (including on a model with 26× more classes and a 50‑class model), improving ad performance and brand‑suitable targeting while cutting model development time to days.

Pixability