Use Case: Email Analysis

Challenge: Missing Discounts

  • Vendors sometimes offer discounts to retail buyers (i.e., a 10% discount for 50,000 televisions).
  • Auditors manually comb through procurement emails to find offered discounts that the retailer never actually received.
  • Manual auditors still miss about half of the total. One retailer estimates unrecovered savings at $280 million per year.

Solution: NLP Analysis

Use natural language processing to automatically analyze emails and find more savings.

Impact: Millions in Savings

  • Auditors find relevant emails faster.
  • Auditors find new recovery opportunities they may have missed.
  • Retailers measure ROI by tracking the emails and recovery amounts attributable to machine learning.

Infinia ML can help you automate email analysis.
Email us to learn more.

Technical Details

Key Inputs
  • Emails from the procurement department. A retailer may have millions per quarter.
  • A sample of emails relevant to a procurement auditor.
Additional Integrations
  • Vendor list(s)
  • Invoices
  • Purchase orders/receipts
  • API to accounts payable system
Configuration

For one retailer, Infinia ML processed 20 million emails from a one year period and:

  • Used unsupervised learning to surface emails that auditors wouldn’t see otherwise.
  • Produced a temporal topic model to explore conversation topics changes with retail seasons. The model improved on open source topic modeling libraries.
  • Built a visualization of email topics (see image above).
50 topics in a set of 3.8 M emails over 12 months. Each number represents a cluster of words that are highly correlated; for example, 6 is “cost retail costs unit margin costing correct lower sheet current”.