From call detail records and network logs, to device and GPS information, there’s an ever-increasing stream of machine-generated data in the telecommunications industry. More cell phones, mobile devices, clickstream data – and anything with a sensor or RFID tag (News - Alert) – are being instrumented, a phenomenon dubbed the Internet of Things.

This presents both an opportunity and a challenge, as companies now have the ability to analyze more and richer information. While it can help organizations improve network performance and planning (among other things), it also creates an overwhelming amount of data to analyze.

Applying Investigative Analytics

Investigative Analytics is a term coined by Curt Monash as a function for “research, investigation and analysis in support of future decisions.” It can help telcos ask a series of quickly changing, iterative questions to better understand what happened or might happen, why, and what to do to optimize for a particular outcome in the future.

Monash defines investigative analytics as “seeking (previously unknown) patterns in data.” It supports the ability to discover a pattern of past activity that points to some likely outcome in the future across any type of data, regardless of its form and origin. For example: Is this data part of a pattern that indicates recurrence? If so, what is that pattern, and how can we can leverage it for business purposes?

As the Internet of Things becomes more of a reality, the proliferation of non-human devices (including cell phones) contributing data into networks, will force organizations to rethink how their analytic systems are designed and deployed. As more of these devices participate in the mobile ecosystem, the Internet of Things will vastly expand the opportunity for telcos to apply investigative analytics to monitor and analyze network traffic, including:

• Discovering the root cause of dropped calls and determining what to do to prevent it going wrong in the future.
• Discovering why something went right so that you can build processes to support an increased likelihood of repeating that behavior in the future. This could include monitoring and analyzing mobile usage, such as sales of location-based services, to encourage up-sell or cross-sell opportunities.
• Planning mobile capacity to cost-effectively support requirements and service level agreements. This is critical in the telecommunications industry, where forecasting and meeting future demand are essential.

Investigative analytics is a good fit for gaining insight gleaned from complex, machine-generated data sets. Anywhere that machines or devices generate information, there is scope for investigative analytics.

Investigative Analytics in Telecommunications

For both performance and planning purposes, telcos need to monitor and analyze traffic. Key elements are hot spots (areas with particularly high usage) as well as failures. A proper understanding of hot spots, and how these develop over time, will be critical to future investments in new infrastructure. Conversely, any failure within the network is an immediate problem that needs resolution as quickly as possible. Failures may lead to connections being dropped (a well-known potential indicator of customer churn) or reduced service.

By applying investigative analytics to areas such as hot spots and failures, companies can combine usage trend data with location-based and demographic data to improve planning for future infrastructure investments.

For example, instead of merely analyzing a canned query of “what happened?” – e.g. when and where was the failure? – mobile network operators can dig into the “why did it happen?”

An investigative analytics series of questions around mobile networks might follow a flexible sequence of interrogation like this:

• “Where are we seeing the most network failures in California?”
• How about social media: “Are we experiencing a rise in customer service complaints that align with those geographic areas in California?”
• Or, drilling in on an individual basis: “Why is this loyal user all of a sudden complaining about bad service on Twitter (News - Alert)?” By combining different data sources, maybe you can determine that the customer has moved and needs to access a cell tower closer to his new home.
• “In what states are we seeing the biggest increase in cross-sell opportunities?” And: “Are those tied in with social media engagement? Is there a pattern between cross-selling/up-selling and network performance?”

Data Store Requirements

Because telcos need to store and analyze a lot of data, they will need a technology platform that enables them to exploit this data to reduce costs, identify new revenue streams, and improve competitive positioning. However, because investigative analytics must support fast, interactive queries, a simple batch-based analytic environment will not be sufficient.

Recommended database requirements to effectively power machine-generated data analytics are:

1. The solution must be scalable enough to hold all the data you need for long-term analytics. And it must have good enough compression to store this economically.
2. It must be fast enough to ingest the data within a reasonable timeframe, providing for real-time query processing and alerting.
3. To minimize costs and resources, the system shouldn’t require manual tuning or administration such as the creation of indexes. Indexing data as it is loaded significantly slows down loading processes and increases database size.
4. The system needs to be fast and flexible enough to efficiently support ad hoc queries and complex analytics. Databases that use indexes or other constructs, such as projections, to achieve fast query performance will not typically provide adequate performance for ad hoc queries.
5. Since it must support mission-critical applications that drive real-time operations, it needs to be highly available (catering for unplanned downtime without stopping) and, preferably, continuously available (catering for planned downtime as well as unplanned stoppages).

Investigative Analytics in Action

The Internet of Things calls for new analytics solutions. Telcos have the opportunity to evolve analytics from simple network monitoring to all areas of operations, product development and customer service.

Take Polystar, for example, a leading supplier of service assurance, network monitoring and test solutions for the telco market. Polystar (News - Alert) embeds Infobright into the company’s Jupiter visualization application suite, providing network and mobile operators with real-time data access and investigative analytics for insight into subscriber behavior.

While Polystar initially used Infobright’s analytic database technology to deliver data access for network monitoring, the company has since expanded its use to help customers glean customer insight. For example, marketing departments can compare subjective survey input on quality of service with actual network measures to determine effective outreach. Customer care agents, as well, perform in-depth analytics to determine why subscribers from a certain geographic area may be changing carriers. 

Improved customer insight drives improvements in SLA root cause monitoring, customer support resolution, marketing research, roaming analytics and product development, thereby reducing churn and boosting revenues. With Infobright, Polystar mobile network operators are equipped to keep pace with unprecedented levels of detail on network and user behavior as the velocity and volume of subscriber data grows exponentially.

Philip Howard is research director at Bloor Research (www.bloorresearch.com).