How Standard AI Powers Retail Intelligence At Scale With Encord

Introducing Customer: Standard AI

Retailers today face some of the toughest challenges in physical stores — from tracking shopper behavior to optimizing promotions and in-store media. 



Standard AI solves these challenges by equipping retailers and CPG brands with AI-powered computer vision technology that unlocks real-time insights on promotions, media impact, new product launches, and more. The VISION platform turns data into actionable insights, helping retailers drive sales, improve store performance, and gain deeper visibility into the shopper journey.  



Standard AI is equipping brick-and-mortar stores and their suppliers with real-time in-store analytics, all while maintaining a privacy-first approach. Helmed by an executive team that spans firms like Mars, Sara Lee, Oracle, NASA, and Adobe, Standard AI is at the forefront of applying artificial intelligence and computer vision to physical retail spaces.

The Data Challenge: lower quality, longer time to value

At its core, Standard AI’s business is based on its ability to extract meaningful insights from millions of hours of video footage from store’s security cameras. 

For the ML team, access to high-quality and enriched video data was an extremely cumbersome, time-consuming process that relied on multiple internal and agency hand-off points, lengthy quality control review cycles, and fragmented point solutions. In several instances, custom scripts were needed to glue these workflows together. The ML team spent more time managing fragile data pipelines than focusing on model performance.

The product and ML teams also wanted to find a way to streamline the data curation and annotation processes and spend more time evaluating and fine-tuning their models to continue improving performance. This wasn’t possible due to limitations of Standard AI’s existing tooling, which offered limited debugging, quality control, and statistical insights. 

From a business perspective, Standard AI understood that the way of managing their AI data pipeline wouldn’t scale or support the rapid innovation needed to meet customer demands. Standard AI’s Product and ML team defined strict requirements for their updated approach:

• Native support for large-scale video processing (including fish eye lens) with support for other modalities
• A unified platform for curation, labeling, and evaluation
• Robust API and SDK support
• Ability to automate labeling tasks with HITL
• A simplified user experience for non-technical users
  

The Solution: easy access to pixel-perfect video data to build production-grade AI

• Native video processing — With prior approaches, the Standard AI team spent long cycles manually uploading videos to other platforms and faced significant conversion issues. With Encord’s native video support and API, the Standard AI team is now able to upload and process millions of video files 5x faster.  
• Unified data management platform for AI — Standard AI shifted from relying on outsourced labelers, open-source tooling, and niche labeling platforms to Encord, a unified platform that streamlines data curation and annotation at enterprise scale.
• Robust API and SDK support — The API integration for data upload is intuitive, enabling rapid implementation. and programmatic data annotation at speed with the Encord SDK.
• Granular ontologies — Standard AI was able to rapidly refine complex hierarchical ontologies in response to evolving project insights, continuously enhancing data labeling precision.
• AI-assisted labeling with HITL — Standard AI benefited from automated labeling functions paired with human-in-the-loop reviews to ensure label accuracy whilst speeding up labeling pipelines.
• Seamlessly scalable across the organization — The ML team has democratized access, enabling all team members to add, annotate, and utilize enriched data for their specific tasks. This distributed approach accelerates company-wide progress by eliminating bottlenecks previously created by specialized tooling gatekeepers.

Benefits: Standard AI built production-ready models faster and at lower cost

Since moving to Encord, Standard AI has seen a drastic improvement in productivity across its product and ML teams, significant cost savings, improved data quality, and faster time-to-model improvement. Key benefits:

• Saved $600k a year
• Process millions of video files 5x faster with Encord’s native video platform
• 99.4% faster project initiation