How a leading media and online technology giant reduced cash leakage by using machine learning to spot duplicate invoices.
Challenges
Duplicate invoice payments often plague enterprises resulting in cash leakage. The customer, a leading technology services provider with a focus on media and online business, wanted an automated machine learning-based solution to identify duplicate invoices and prevent multiple payments of the same invoice.
- Thousands of vendor invoices and employee transactions are generated and processed each month
- Numerous duplicate invoices produced due to duplicate vendor names, duplicate payment transactions, errors in invoice numbers, disparate dates and invoice amounts
- Existing manual and ad hoc processes to spot and correct duplicated invoices could not effectively review all invoices
Outcomes
- Automated invoice validation to detect duplicates across the spectrum of vendor invoices that eliminated manual work and prevented cash leakage
- Ability to review audit positive records daily thus catching duplicate payments much faster
- Accurate spotting of duplicates between expenses submitted and supplier invoices
- Streamline and automated reporting
- Assured data security through encryption keys
- Scalable, robust and secure machine learning infrastructure for all future machine learning use cases
Technologies Used
- Built an automated machine learning-based solution to rapidly spot duplicate invoices and employee transactions
- Used fuzzy logic and various comparison algorithms and rules to identify potential duplicates
- Built a highly scalable and robust machine learning infrastructure with separate development and production servers for all future machine learning projects within the enterprise domain
- Used leading technologies including Python, Flask, uWSGi, Yapache (Apache), Athenz-CKMS, ScrewDriver, RHEL servers and Oracle Database to develop the solution
