Leading Supply Chain Company improves its invoice data extraction efficiency by 95%
Client Profile
The end client is a leading logistics company based in the USA, providing various logistics and transportation services across the globe.
Business Problem
The company receives thousands of invoices each month from its suppliers, with each invoice having a different format. The client was looking for a solution that can help it automatically extract certain key fields from the supplier invoices. With its present solutions, it was unable to get the desired extraction accuracy. Finally, it sought the assistance of Optimum Data Analytics (ODA) to overcome the issue.
Key Challenges
- Myriad of invoice formats made it difficult to extract required data from invoices in a given time frame.
- Unstructured data and complex formats required conceptualizing a deep learning model that could work equally efficiently with each invoice format.
- Timelines to conceptualize, train and deploy the model to produce accurate results were stringent.
Technologies Used
Inference APIs from Document Data Extraction Platform, Document On-boarding Services, Python, OpenCV, TensorFlow, Deep Learning, Tesseract, CTPN
Solution
- Identified 10 distinct and complex invoice formats for pilot implementation.
- Built a bespoke application using Inference APIs from our Document Data Extraction Platform for data validation and audit.
- A deep learning model was trained on each supplier invoices format. Then, the model was deployed as an API, on cloud. The client integrated these APIs into their application.
- The APIs enabled the client to feed supplied invoice as an input and receive the extracted output as a JSON file.
Business Impact
- The data extraction accuracy improved remarkably to 95%
- Streamlined supplier invoice data extraction enabled the client to divert more attention towards core strategic activities.
- Positive results fostered a strong relationship as the client decided to award ODA a long-term contract.