Classified 80,000+ inconsistent, multilingual job titles to make segmentation reliable at scale.
Organic, multilingual job-title input had made calculation-based segmentation unreliable across a 600,000+ record database with 80,000+ unique titles. I normalized titles to alphanumeric form, deduplicated to ~50,000, and defined level and function taxonomies.
Classification ran through Copilot and ChatGPT using an iterative one-percent sample-correct-rebuild loop until samples needed no further correction. Marketo Smart Campaigns apply classifications to new input automatically, and Power Automate flows learn emerging title trends and flag when rules should change — keeping a human checkpoint before any rule goes live.
