Consulting giant eliminates a year of labeling time with Snorkel

A 170+ year old global consulting firm with 300,000+ experts across 150 countries needed to flag audit-relevant news signals (fraud, liquidity issues) from a huge volume of accounting/auditing articles, but manual labeling was extremely costly - three experts took a week to label just 500 data points, implying roughly 52 person-weeks for 10,000+ articles. Snorkel's programmatic labeling technology let the firm's data science team build an NLP relevance classifier using labeling functions instead of manual annotation. The firm labeled 10,000+ articles in hours instead of nearly a year, improved accuracy 15% over hand-labeling, raised F1 from 70 to 85 within days, and cut application development time 3x while reducing required SME time by 50%.

The customer is anonymised in Snorkel AI’s own case study — not withheld by us.

Snorkel AI is a data-development and evaluation company whose programmatic labeling tools build training and benchmark datasets used to stress-test whether a model approach is viable before a full build.

This sits in the Design & Stress-Test stage of the AI use-case lifecycle: Named examples from the tools pharma teams use to scope, prototype and benchmark an AI use case before committing to build it — data-catalog, sandbox and evaluation platforms.