Machine Learning-based Ticket Triaging with 82% Accuracy for a Digital Media Solutions Provider
One of the world’s largest personalized campaigns provide, coupons and digital media solutions to retailers and CPG brands
Over 5,000 service tickets were being raised each month. Tickets were coming from a vast network of retail stores. It was important to triage them and assign to the respective teams for rapid resolution.
Dedicated triage teams were working in shifts round the clock to support and address each of the tickets at the earliest.
The firm sought to automate this human intensive task of ticket triaging which involved reading each ticket description. The data on the tickets was unstructured.
Mindtree’s cognitive technologies team evaluated the past tickets data and processes to check if machine learning could be applied to achieve automated triaging.
A machine developed a learning model using Artificial Neural Net Framework to analyze tickets raised in the past 10 months, as a resolution. This model helped predict the team that was best positioned to address the ticket.
The prediction model was then applied to new ticket descriptions to predict the best-fit resolution team.
- Machine learning model was able to predict the resolution teams instantaneously for the new tickets
- The prediction was accurate for 82% of the tickets raised
- Machine learning-based automation of triaging has the ability to reduce the load on the triage team multi-fold