The Only AI Stack You Will Ever Need

The Big Picture

Shoppers take approximately 50 milliseconds to decide whether they like a site or not. For a retailer, this means that the site and images on it need to be of the highest quality in order to grab shopper attention. In fact, 88% of shoppers think product images are crucial for making a purchase decision. As a result, brands and marketplaces have invested a significant amount of resources into ensuring clarity of image and alignment with their guidelines to deliver an optimum shopping experience.

With over 25 million eCommerce sites around the world, and 2 billion shoppers online. The volume of products entering the market, and the level of competition means the speed at which retailers go to market is critical.

The Impact

The team worked closely with Mercado Libre to understand which mission critical challenges they were looking to solve, and thus aligned on key goals to be achieved. Read on to understand about the AI intervention for the customer and how our AI was able to add value to their business.


Reduction in rollback rates


Reduction in number of tickets raised by sellers regarding image QA issues


Reduction in rollback percentage for text moderation


Reduction in false positive detection for text moderation


With 12M sellers in their database, every product image lined up for the website needed to follow specific guidelines. Manual image quality moderation resulted in three critical issues:

  • Slow onboarding time
  • Errors and inaccuracies in results
  • Substantial time and effort required for quality checks

The resource-intensive manual process also slowed down the time taken to go-to market and monetize a product.

Key Goals

With the challenge to assess image quality at scale and streamline image onboarding, the key goals identified for the intervention were:

Streamline and simplify the image qualification process

Automate feedback loops that can train the system on QA based corrections / re-classifications

Decrease the effort taken to manually assess product images

The Intervention

This involved integrating our AI-powered image recognition capabilities into the customer’s vendor onboarding workflow.

This would enable:

Use Cases

Real-time automation that helps vendors receive immediate feedback on images they have uploaded.

Use Cases

Systems to be easily trained to cater to custom guidelines that are unique to the customer’s marketplace

Use Cases

Reduction in effort by the customer’s team for manual intervention

Use Cases

Automated feedback loops to enable re-classifications to be fed back into the system ensuring they are not repeated again

Use Cases

Streamlined workflows that accelerate the image qualification process and enable sellers to go-to-market faster

AI-Powered Workflow Automation

The AI-powered workflow ensured higher efficiency in the product onboarding process and significantly reduced the time and manual effort required to qualify seller images uploaded on the site, resulting in shorter go-to-market periods.

Key Outcomes

By automating an otherwise manual process, our AI was able to significantly reduce the time and cost associated with assessing the images that are uploaded on the customer’s website.

86.5% reduction in rollback percentage for text moderation

70% overall reduction in rollback rates across 3 months

51% reduction in number of tickets raised by sellers to customer call centers regarding image QA issues

56% reduction in false positive detection for text moderation


Stay in the know of the latest and greatest in AI & Business. Subscribe to the Out Of The Blox newsletter today.