India-based KYC Platform
How the India-based KYC platform reduced the response time in invoice inspection by 70% and cut the overall costs by 50% through reducing the efforts of manual data entry
0
%
Reduction in response time for invoice inspection
0
%
Decrease in overall costs through reducing the efforts of manual data entry
Company Description
- Based in Malaysia with a team of hundred people and a customer base of over a million
- One of the fastest-growing travel companies in Asia
- Offers its customers the ability to find and purchase affordable travel options
- Acts as a broker and travel-agency-as-service company
Challenges Faced
- Needs to process a wide spectrum of invoices throughout the entire operational flow, especially its airplane tickets brokerage services
- Recruit a few contractors to process all documents manually
- Manually processed all documents in PDF formats, which is time-consuming and highly error prone
Solution
- All invoices are processed through XXX, which serve as the core operational platform in the finance department’s Accounts Payable team
- Use pre-trained model to capture key information on the invoice
- Pair vendor and cost center records
- The company’s backend team is responsible to integrate the pipeline with original Oracle system, while XXX provides all-rounded and quick technical assistance to the team
Impact
- Largely reduce manual work which originally require full-time effort of nearly 10 contractors every month
- Now, the team is spending only about 1 hour per person daily
- Lower error rates by 90% and reduce the number of exceptions by 70% for manual review
- Without cutting headcount, the company can now assign more advanced tasks to the contractors, such as past due reports, financial analysis and credit notes to reduce costs
Company Description
- India’s digital identity ecosystem, providing automated age and identity verification solutions
- Process over 1 million identity documents in India annually
- The company helps its partners on verification process when onboarding new customer to his/her employers
- Most commonly submitted identity documents are driving licenses and passports
Challenges Faced
- The company is now using traditional OCR engine along with a team of 5 contractors to manually review the results, which is an extremely time consuming and expensive process
- Also, the identity documents often have different templates or formats by country or even by regions. Therefore, traditional rule-based solutions failed. Not even mention sometimes the scanning is of poor resolution, tilted, not adhering to the standard photograph protocols.
- The company also does not have enough resources to maintain an in-house OCR solution and with little improvement in accuracy even after repeated training
- The company has tried solution from international technology giants, but the solution is not tailor-made and the overall accuracy level is not very high
Solution
- MachIDP has developed a custom trained-model using customer data to train on top of the pre-trained model
- The last few layers of the neural network learn from the real customer data to give accurate prediction in production stage
- The company only needs to provide a few hundred samples to train the custom model, and they have the freedom to define the fields to be extracted
- The solution can identify the structure of the identity documents with key data fields captured
- MachIDP’s data outputs can be extracted in any formats that can be easily integrated with the company’s existing application
- To ensure data privacy, MachIDP’s models can be configured to run on-premises using docker containers so the user data never leaves the company’s infrastructure
Impact
- Reduced the response time by 70% in invoice inspection and cut the overall costs by 50% through reducing the efforts of manual data entry
- Also, the accuracy has improved when compared to human review and the number of manual reviewers has therefore reduced along with the number of passes required for each identity documents