Download PDF
Cortical.io > Case Studies > How an International Transportation Company Optimized High Volume Email Processing in Customer Centers
Cortical.io Logo

How an International Transportation Company Optimized High Volume Email Processing in Customer Centers

Technology Category
  • Analytics & Modeling - Natural Language Processing (NLP)
  • Analytics & Modeling - Real Time Analytics
Applicable Industries
  • Transportation
Applicable Functions
  • Logistics & Transportation
Services
  • Data Science Services
The Challenge
The company was receiving between 100,000 and 250,000 customer emails per day in 35 countries. A major challenge with such a high volume of incoming emails was to identify the emails that actually required a response (only 50%). So far, the customer centers spent a significant amount of time sorting out emails that do not need to be processed (e.g. out of office messages, FYI mails, etc.), a task that unnecessarily burdened the teams and increased the costs in the customer centers. The task was further complicated through the use of multiple languages in the emails.
About The Customer
The customer is an International Transportation Company that operates in 35 countries. The company receives a high volume of emails daily, ranging from 100,000 to 250,000. These emails are in multiple languages and only half of them require a response. The company's customer centers were burdened with the task of sorting out irrelevant emails, which increased operational costs.
The Solution
Cortical.io, together with its partner PwC Germany, developed a Web service to detect emails that are not business case relevant. The solution flags those emails as “no case”, and in addition categorizes the email topic (e.g. “invoice”) for proper routing. The solution was trained with a small amount of annotated emails and reference material related to the logistics domain (books, pdfs, websites, emails). In addition to classification, the solution uses language detection algorithms to route mixed language emails to the appropriate native language speaker. The solution can easily be adapted to new situations at short notice and has minimal hardware requirements.
Operational Impact
  • The system is so precise that it detected errors produced by human annotators in the training set.
  • The solution allows an “audit-track”–every single decision of the system can be clearly traced and each semantic processing step can be inspected, allowing the company to understand why each email has been classified as “case” or “no case”–an important aspect with respect to international rules like GDPR, which oblige companies to be able to justify any decision based on the use of an automated system.
Quantitative Benefit
  • The company was able to automatically extract relevant terms from hundreds of thousands of emails daily
  • The system was able to label each new incoming email as “case” or “no case”, with indication of the level of prediction confidence
  • The system was able to classify each business-relevant email into categories like invoice, complaint, order, etc.

Related Case Studies.

Contact us

Let's talk!

* Required
* Required
* Required
* Invalid email address
By submitting this form, you agree that IoT ONE may contact you with insights and marketing messaging.
No thanks, I don't want to receive any marketing emails from IoT ONE.
Submit

Thank you for your message!
We will contact you soon.