AWS for Industries

How WNS Malkom is Redefining Less-Than-Truckload (LTL) Billing Using AWS

Most important document in less-than-truckload (LTL) is the bill of lading (BoL). This is a physical paper document between shipper and carrier acknowledging the receipt of goods for transport. A shipper is a person or company who is the supplier of commodities being shipped. A carrier is a person or company that transports goods for any person and/or company and is responsible for any possible loss of the goods during transport. The BoL describes the nature of cargo, amount of cargo by weight, size and/or number of pieces, the origin and destination, and the party to be billed. Correct and timely processing of this document is essential to an LTL carrier in running an efficient operation.

Key challenges faced by carriers in processing BoL are:

  1. Paper Format and Non-Standardized Bill Formats: The billing process (entering the BoL details into the Freight Management System of Record) is extremely manual and non-standardized. Every shipper can have its own format for the BoL, increasing the complexity for achieving automation.
  2. Accurate Digitization: Accuracy in digitization is extremely critical as any error in it can result in the shipment being incorrectly routed, handled, or billed.
  3. Handling Spike Workload: Inflow of these bills is not consistent as truck drivers collect multiple shipments in one go and deliver all of the bills at the terminal at the end of their trip. Hence the solution needs to scale up to handle these spikes where 70–80% of the bills will arrive during a 3-to-5-hour window.
  4. Large Number of Business Rules: LTL business typically caters for varied types of cargos and shippers. As a result, there is a need to cater to a large number and wide variety of business rules dependent on shipper, consignee, billing parties, payment terms, types of goods, and more during the digitizing process.

WNS Malkom, a proprietary digital billing platform, addresses these challenges using AWS by enabling a digitization desk for automating processing of LTL BoLs. The solution consists of three parts that are combined together for each customer to deliver an end-to-end digitization desk on AWS.

Part 1: BoL Ingestion

  • Primary Purpose: Ingest and process scanned bills as they arrive and extract text/data/information within the required turnaround time (TAT). Typical arrival volumes for large carriers can be in the range of 10,000 bills per hour during the peak time. Typical TAT ranges from 15–60 minutes.
  • Key Non-Functional Requirements:
    • Ability to ingest varied formats, types, and resolutions of bills
    • Ability to ingest bills from multiple channels, e.g., email, file upload, API, etc.
    • Ability to handle spikes in load so that processing can be completed within the required TAT
    • Ability to extract information automatically from the bills with nearly 90%+ accuracy for shipper, consignee, and bill to information
    • Ability to continuously improve extraction accuracy with time based on corrections and feedback from the billers enabling touchless processing as much as possible
    • Ability to apply business rules based on shipper or consignee and bill to parties during extraction process

Part 2: BoL Validation

  • Primary Purpose: Present digitized bills to an agent or auditor for validation or verification.
  • Key Non-Functional Requirements:
    • Ability to on-the-fly extract information from bills by selecting a rectangular area in the image
    • Ability to show the agent visually from where in the bill has the information has been extracted
    • Ability to configure business rules in terms of special instructions, soft and hard validations based on shipper, consignee, bill to, and type of goods being transported

Part 3: BoL Submission

  • Primary Purpose: Integrate with the system of record and submit validated bills for downstream processing in the Freight Management System.
  • Key Non-Functional Requirements:
    • Ability to integrate leveraging multiple methods, e.g., RPA (Robotic Process Automation), file, API etc.
    • Ability to handle exceptions

The following image is a reference architecture of the WNS Malkom BoL solution:

WNS Malkom BoL solution

In order to achieve these requirements, the solution leverages AWS as follows:

Capability Solution Services Leveraged

Ability to ingest varied formats across multiple channels

 

Ability to ingest bills from multiple channels, e.g., email, file upload, API, etc.

 

Multiple channels to receive scanned images

  • API interface
  • Cloud storage
  • Email
Amazon API Gateway, Amazon S3, and Amazon Simple Email Service (Amazon SES)
Ability to extract content automatically from bills of multiple types (images, PDF, Word, Excel, etc.)
  • A serverless extraction pipeline triggered by event originating from either an email coming in, API being called or a file being pushed into cloud storage.
  • Leverages the AWS fully managed machine learning service Amazon Textract to extract content from images and PDF.
  • Leverage WNS’s proprietary Shipping contextualization engine to extract entities like shipper, consignee, bill to, commodities, weights, cargo type, various types of shipping account reference numbers (ARN), and shipping instructions.
Amazon Textract, AWS Lambda, Amazon SNS, Amazon SQS, WNS’ proprietary Shipping contextualization engine

Ability to handle spikes in load with consistent performant throughput

 

 

  • Leverage serverless capabilities of the AWS Cloud. Currently the platform seamlessly can scale to handle 12,000 documents per hour.
  • Leverage AWS Well-Architected Framework best practices for scalability, reliability, and high availability.
Amazon Textract, AWS Lambda, Amazon SNS, Amazon SES, Amazon SQS, Amazon Aurora Serverless, Application Load Balancer, Auto Scaling Groups, Amazon CloudWatch
Ability to improve extraction accuracy over time
  • Automatically detect shipping ARNs to populate for shippers, consignee, and bill to parties based on information extracted from the bill. This leverages the following three techniques:
    • Populate from short bill data and reference data in case available
    • Leverage machine prediction based on past data that has passed through the syste
WNS’ proprietary Shipping contextualization engine

Ability to on-the-fly extract information from bills by selecting a rectangular area in the image

 

JavaScript-based OCR enabling the agent to perform OCR on-the-fly Open-Source JavaScript OCR Library
Ability to show the agent visually from where in the bill the information has been extracted Visually highlight sections in the image based on coordinates of extracted text within the document Amazon Textract
Ability to integrate through multiple methods, e.g., RPA, file, API, etc. Generate digitized representation of the BoL in either JSON or XML format, which can be posted in downstream system through a serverless event-based architecture. Amazon API Gateway, Amazon S3, AWS Lambda

For more information, see Serverless on AWS

Conclusion

Leveraging the techniques mentioned previously, we are able to accurately digitize 70–90% of a BoL without any human intervention and any compromises in accuracy. The serverless architecture enables WNS Malkom to scale from few bills per hour to more than 12,000 bills per hour without any human intervention.

To learn more about AWS for Transport and Logistics, contact your AWS account team.

Puneet Sachdev

Puneet Sachdev

Puneet Sachdev is Corporate Vice President at WNS Global Services, spearheading all technology solutions across Travel, Shipping & Logistics, Utilities, DMRT and Customer Experience. He also leads cloud competency at WNS.

Sachin Sapkale

Sachin Sapkale

Sachin Sapkale is Enterprise Solutions Architect at Amazon Web Services with over 21+ years of hands on architecture and development experience in multiple industries covering Media & Entertainment, Financial Services, and Healthcare. He is focused on helping customers design cloud-based solution (Financial Services and Healthcare) to operate in regulated environments.