How to Harness Big Data and Cognitive Computing
By Micah Singer, CEO, VoIP Logic | June 02, 2016
The growth in computing power and storage capacity is providing the underpinnings of another evolution in information technology – access to big data and the ability to turn this data into operational intelligence. One startling statistic about the scale of data collection purports that in the past two years we have generated 90 percent of all the data produced by humans. Admittedly, much of this data is not produced directly by humans – it comes in the form of log files generated by servers – not exactly on the level of the Magna Carta in terms of elegance or import, but useful in its fashion if applied correctly.
We (along with our machines) are generating such a vast digital record that the overwhelming challenge for technology professionals is to make this mass of information useful to improve how we conduct business. Unfortunately, like with many other industries, telecom is behind the curve where big data is concerned. To make it productive there are three discrete steps – analysis, interpretation, and cognition. In brief, this involves crunching the data, extrapolating meaning, and then reacting to the data with action while learning from the process to improve future interpretation.
Business hosted/cloud communications is positioned to see value from this process. We generate inordinate amounts of data from usage (CDRs), human to system interactions (user logs), and system performance (machine logs), our clients value service performance and customer support above all else (including price) and providers in our sector are continually challenged to both scale larger and to retain the personalized approach to each customer.
Here are a few areas where intuitive use of big data coupled with cognitive computing can improve your bottom line.
Identify Fraud and React
I have written extensively about the scourge of toll fraud and fraud perpetuated through denial of service. Rapid analysis of large quantities of call record data can tell us with quite a bit of precision when fraud is occurring. Couple this processing power with cognitive assessment of a user’s historical use and business-type profile, and accurate fraud assessment should improve dramatically. Further layer on call path control and dial plan management, and you can shut off these users before they do damage. The same is true with denial of service attacks – IP addresses can be blacklisted on the fly to prevent network disruption.
More Efficient Customer Service
Each time you have a service call there is an associated hard cost in time spent. Analysis of the user data ahead of a call allows the CSR to answer most service issues with minimal on-the-fly diagnostic work. Delivering more information to the CSR about the incoming call – user name, account, log history, call history, service uptime, billings and payments – and parsing this information (or even highlighting potential problems) can predict the nature of the coming call. For instance, if a user has tried to make a call or pay a bill or upgrade/downgrade service multiple times and this request is quickly followed by a call to support, the user’s questions are often self-evident. The goal is not to deflect customers to web self-service (that can undermine the appreciated personal touch), but to use the data and cognitive processing of the data to route the call to the right support person who can respond more precisely to the presumptive issue at hand. Make the data your precogs (a la Minority Report) to catch problems before they generate bad will or churn.
Identify Revenue Opportunities
User behavior can be very instructive on when and what to upsell. If data indicates a user makes calls from the same geographic location and experiences intermittent packet loss, it could be time for a managed bandwidth upsell. If, on the other hand, a user makes calls from a range of locations on a regular basis, a mobile phone app might come in handy. Well-interpreted data can lead to a more thoughtful and productive sales effort. Over a large enough sample size – with success tracking metrics (click rate or close rate) – cognitive computing can further refine the correct moment for the upsell.
Complex systems measure uptime in 9s. 5 9s means a technology is functioning 99.999 percent of the time. The 5 9s standard of uptime is impressive and highly respected as a success credential in most fields of technology including business communications. As the table stakes for service quality get higher, however, competition in the margins – for that .001 percent – is getting more and more important. Given the nature of VoIP infrastructure, big data analytics will allow Internet telephony service providers in particular the opportunity to improve uptime by analyzing network traffic in real time, and optimizing packet flow.
Technologies like Splunk and others that process gobs of data and extract operational intelligence are the tip of the iceberg. At VoIP Logic, we have been using Splunk for several years. While it is powerful in the analysis, it has been hard, so far, to couple the lessons that can be found in the data with intelligent cognition available from systems like IBM’s Watson.
However, unlocking this equation promises to usher in a world of more proactive and thoughtful business operations management in many ways. Make sure you ask your technology vendors how they are starting to harness data and cognition in the system they offer you.
Micah Singer is CEO with VoIP Logic (www.voiplogic.com).
View the original article here.