Case Study: Tractable improves annotation quality and scales QA with Encord

Building Computer Vision Models for Insurance with Tractable

Tractable, a company that develops AI for visual damage assessment in the auto and property insurance industries, faced a challenge with scaling its complex data annotation projects. Their previous platforms could not handle sophisticated, multimodal ontologies and lacked the robust quality assurance functionality needed as their remote annotator team grew. To build accurate models, they required a user-friendly training data platform that supported granular annotation, annotator monitoring, and efficient QA workflows, leading them to partner with Encord.

Encord provided a solution with its training data platform, which featured the advanced quality assurance and annotator review tools Tractable needed. The platform’s user-friendly interface allowed for building complex ontologies quickly, and its API integrated with Tractable's S3 servers for secure data governance. As a result, Tractable efficiently trained and scaled its team to over 40 annotators, achieving a critical balance between annotation speed and strong quality control. Encord's direct support and rapid feature development based on feedback were also cited as key benefits.

Tractable

Camilla Gilchrist

Head of Operations