Case study

The challenge

Point-of-sale information is key for our client, a global advertising analytics firm, to understand sales trends and customer preferences. Its brand managers use this valuable data to make decisions about promotions and new sales initiatives.

However, reports on the firm’s sales data were often compiled manually, which required personnel to enter information into forms and spreadsheets. This led to errors, inefficiencies and high costs—including auditing expenses. The company asked us for a better way to extract information from sales receipts using AI to automate the analysis and generate insights.

Our approach

As a first step, Cognizant compared the performance of different optical character recognition tools available in the market to identify the best one for this application. We chose the Microsoft Computer Vision API for its ability to extract consumer purchase information and recognize familiar subjects like brands and moderate content from sales receipts. Collaborating with our client's technical team, we designed and implemented a solution driven by artificial intelligence (AI) that enables this worldwide company to process sales receipts automatically and glean key information more quickly.

The solution uses text analytics and natural language processing to categorize data. The cognitive engine on Microsoft Azure scans and identifies merchant and transaction information—including products, retailers, vendors, promotional offers and brand logos—from retail receipts using natural language processing. It also stores the extracted information in our client's Cosmos sales-tracking database for multiple reports. We also implemented different machine learning algorithms to classify the retailers and identify sales patterns.

Adopting AI to identify vendor and customer purchase patterns

We improved our client's processes without disrupting them and significantly increased their operational efficiency. By using AI to extract customer purchase information from each sales receipt and compilation of useable information, sales executives can now better understand consumers' buying patterns. Using machine learning algorithms provides better, more accurate information for their future-looking marketing strategies and overall business decisions.

95%

accuracy in identifying retailer logo and name

70%

reduction in manual effort due to machine learning

30%

increase in operational processing capacity while maintaining accuracy