Anomaly Detection for Call Center Operations

It’s a Friday evening and ABC Corporation call center is running on a lean roster. Friday evenings are generally a good time as the call velocity is less and types of inquiries coming in are basic that even the roster is planned in a such a way that your best resources are not in action. But things are about to change as the Infrastructure team is applying a minor patch in production which has been done many times, and this time has been picked due to low transaction volume at 8pm.

But beyond all this known-known, expected-expected reality today there is a recurring local festival that is happening which is not famous enough to get into a planned roster of holidays, but your data is aware of this as there has been historical spike on this day, this month between 6-10pm for certain merchants.

The ‘expected outage’ on any other day is heading for a major escalation the planned IT migration has gone ahead and impacted certain configurations on production system. The system is already going through an increase in traffic, and then disaster strikes. The load balancer starts to fail, and transactions starts to fail. The merchant is not able to make critical transactions for the festival and the call center is now hit by a storm of calls. Social media starts blasting with red signals, negative customer sentiments start building up.

Now this storyline might look like a starting scene of a Hollywood movie, but it is a fact that every modern business is going through digital transformation and this means that the operational metrics have become like an ECG of your business. You need to keep a continuous tab on metrics ranging from

infrastructure to business points, in order to maintain the health of a modern business. Else, it might end up in issues that might affect business continuity or result in churn.

By investing on an AI enabled anomaly detection tool, a modern enterprise can continuously tap into multiple business critical metrics. AI based algorithms have the capabilities to understand the seasonality in data sets and automatically predict demand curves to detect anomalies, hence allowing businesses to prepare well for unknown-unknowns.

However, detecting anomalies is only one part of the story. The solution should have the capability to correlate metrics from multiple parts of the organization ranging from infrastructure failures to surge in transactions to maintenance events that are happening.

In this particular scenario, the fix for the problem was a change in configuration in the web server, but again, every second that passes by in such a scenario is resulting in lost Dollars, and this is where the automated Root Cause Analysis (RCA) and recommendations come into picture. By correlating historic incidents and mining log patterns, the solution is able to recommend fixes which can be critical.

Let’s hope none of the digital enterprise have to face this issue but a modern anomaly detection tool is like a smart watch attached to your body. It could read critical data from your body and warn you when something is going wrong. You many never know the importance until the next anomaly strikes you.

Author Bio

Nithin Gangadharan is the product manager at CrunchMetrics.  He has more than 10 Years of experience in Fraud Management. He started his career as an Implementation Consultant with Subex Ltd and has been part of many Fraud Management implementations across APAC & Middle East. He has also been a Subject Matter expert & Business Solution Consultant earlier. Nithin is currently working as Product Line Manager for Fraud Management and machine learning developments at Subex.

TechBullion PR: Press releases and sponsored contents, with advertising or promotional materials. | TechBullion is not responsible for or liable for the sponsored contents, accuracy or quality within the press release.