We trust Mechanical Turk results. When Workers complete the Human Intelligence Tasks (HITs) and we complete our quality verification step, data is injected directly into the production systems and managed like any other data source. The accuracy, breadth, and completeness of this data are our core business.
Vice President of Operations, Channel Intelligence
Channel Intelligence helps manufacturers, retailers and affiliates understand their data to sell more products, and help shoppers find and buy the products they are looking for, both online and in local stores. With product data as the core foundation of its services, Channel Intelligence continually looks for innovative ways to store, clean, optimize, and manage the hundreds of millions of products in its database.
Channel Intelligence adds 50,000-100,000 new products per day, and each product must be categorized. However, Channel Intelligence's automated process was able to successfully categorize only 80% of the items—the other 20% required manual intervention.
Capacity for manual categorization was not sufficient to keep up with demand, and the backlog exceeded 73,000 uncategorized products. Channel Intelligence estimated that with its current process it would take hundreds of hours to correctly categorize these items, and decided to look for an alternative.
Facing this massive hands-on categorization project as well as a number of other large work-intensive projects, Channel Intelligence decided to leverage Mechanical Turk to test the quality and efficiency of the Workers. If Mechanical Turk delivered against its promise of speed, and low cost, Channel Intelligence could further integrate Mechanical Turk into its operational processes.
The uncategorized products were converted to Human Intelligence Tasks (HITs), which were performed by multiple Workers so they could automatically validate the accuracy of the responses when Workers agreed on the category. In four days, the Workers were able to complete the entire project with 88.3% accuracy, and at 22.1% of the normal cost. Channel Intelligence was convinced that with some additional optimization Mechanical Turk was their answer.
Since the initial test, Channel Intelligence has used Mechanical Turk for nine different projects including categorizing products, extracting product attributes to make them searchable, validating product URLs and identifying retail store locations. These projects have improved both their bottom and top line.
Learn more about Channel Intelligence by visiting their website, www.channelintelligence.com