Check out our Amazon Connect solution! Learn More Introduces AI-Powered Analytics

By callstats on November 15, 2018

This week at Kranky Geek in San Francisco, we are excited to introduce new AI-powered analytics that automatically analyze your WebRTC metrics and identifies the root cause of trouble, eliminating costly manual data analysis and accelerating time-to-resolution. Available immediately for enterprise and pro plus subscribers, these new capabilities are based on a state-of-the-art artificial intelligence architecture and form a cornerstone in the company’s mission to deliver analytics with answers. 

AI-powered analytics addresses three key pain points we have seen our customers face since we started in 2014: Building a monitoring system consumes specialized and costly manpower, problems are often unanticipated and undetected, and identifying the true root of the issue is complicated and time consuming. Our proprietary artificial intelligence system looks to address all of these issues to save you time, manpower, and money.

Efficient Monitoring and Analyzing

Monitoring a WebRTC service is a large scale task that requires specialized and costly manpower and tools. To efficiently monitor and analyze a WebRTC service, your team must be able to:

  • Decide which of the more than 500 metrics must be monitored;
  • Set the proper threshold to indicate trouble for each metric;
  • Identify external factors (ISP, network type, geographic location) and data sources for monitoring.
With our artificial intelligence system, we take all of that off your plate. We efficiently monitor and analyze large volumes of WebRTC service data, saving man-hours of manual analysis. Our proprietary Objective Quality metric aggregates multiple data points to produce an accurate indicator of the quality of user experience. In addition, round-trip time and call count are automatically monitored by this first release of AI-powered analytics (read our white paper to learn about additional metrics on the roadmap). Four years of experience has shown us these KPIs help operations teams and product managers keep an eye on the overall health of their service.

Using highly scalable statistical inference methods, we analyze data patterns and automatically set thresholds. This eliminates complex and error-prone manual processes that can produce false positives or trouble indicators that are altogether missed.

All of these pieces come together to form an efficient, dynamic monitoring system backed by historical data analysis. The algorithms use historical data collected over multiple time windows to determine when a change is occurring.

Proactive Problem Identification

In a WebRTC service, problems are frequently unanticipated and may go undetected while users either abandon your service or suffer through a poor experience. Most monitoring systems focus on service-level KPIs that don't indicate when localized problems occur. Avoiding trouble requires predictive capabilities that aren't available in contemporary monitoring systems.

We’ve applied machine learning algorithms to alert you to trouble before you start to lose customers. We automatically detect trends and fluctuations in KPIs – including normally fluctuating metrics, such as usage. Our algorithms include time series predictors that help forecast trouble so you can avoid it. With this combination of predictors and detectors, you can feel confident detecting and resolving issues before they affect user experiences.

Figure 1: Trends and fluctuations on

Trends and fluctuations are reported in a new window on the dashboard. The system indicates the metric for which a deviation or trend has been detected, the magnitude of the change and the time period over which it was detected. The system monitors metrics over multiple time windows to detect changes in network operating profile that can occur in minutes or slowly unfold over months.

Identifying the Root Cause of Communications Issues

Identifying the root cause of problems or a change in operating profile requires specialized skills, tools, and time. To efficiently identify issues in a WebRTC service, your team must be able to:

  • Develop a hypothesis for the root cause, which requires expensive and hard-to-find communications expertise.
  • Comb through mountains of data to find correlations and evidence to support a hypothesis, which currently takes way too much time.
With our artificial intelligence roadmap, we look to rewrite the troubleshooting handbook by delivering a root cause diagnosis and eliminating hours of troubleshooting time. Artificial intelligence will develop a holistic view of network operations, identifying correlations across WebRTC application metrics and external factors. Using artificial intelligence, we will provide a reason for detected network behaviors that gives your team a comprehensive explanation for a problem and how to fix it, eliminating the need for expert analysis.

How do we do it?

By combining statistical inference, machine learning and AI algorithms, we help save the time of your most skilled staff members. Their hours of monitoring and interpreting just became a single, easily understood notification. In our free white paper, Instantaneous Answers: How Uses AI to Improve Communications, we describe our architecture and roadmap in detail.

This is an exciting time to be a customer. Our new AI-powered analytics is a cornerstone in our mission to deliver customers analytics with answers. We help accelerate the entire trouble shooting process, from detection to root cause identification, which delivers better user experiences. For more information, visit our page here

Interested in checking out our new AI-powered analytics? Try our free demo.

Try the Demo

Tags: Artificial Intelligence,