Target Behavior in Real Time for Effective Outcomes: Part 1
How real-time, adaptive architectures can drive management decisions for specific use cases
Big data offers the capability to observe a population and collect detailed behavioral data. Many business use cases thrive on the ability to find and treat a small percentage of customers in a special way. After identifying a small segment of customers, an organization can focus on them by collecting additional information, making decisions, and then targeting those decisions in near-real time to achieve desired business outcomes. This first installment of a two-part series takes a look at targeting behavior in the context of some specific use cases, and the
discusses a real-time, adaptive architecture for effective behavior targeting.
Many of these business opportunities are familiar to organizations. However, an unprecedented opportunity now exists to observe an entire population, focus attention where it matters, and individually target chosen customers at the right time with an understanding of their historical behavior.
In recent years, there has been a proliferation of big data sources and advanced analytics techniques for behavioral targeting. In particular, usage and location data from the Internet of Things are getting special attention. Network probes have opened up access to detailed usage data from wireless networks. And wireless service providers have access to these high-velocity data sources. They are interested in utilizing this data for themselves and for third-party organizations. Collating all the network data and acting on useful patterns at the right time can present major challenges for many tier 1 telecommunications wireless service providers.
Behavioral targeting can make a difference
Consider a scenario in which a telecommunications customer service organization would like to analyze usage and service-quality data to decide if a specific device is deteriorating in performance and requires preventive maintenance. Given that a very small percentage of devices require preventive maintenance, the organization needs to determine how to find those devices and drive proactive maintenance for them.
Major telecommunications organizations incur significant costs each year to replace devices that are not faulty because the subscribers perceived they had problems that required obtaining replacement devices. At the same time, a hefty percentage of subscriber churn can be experienced when the subscribers struggle with their malfunctioning devices. Instead of seeking a replacement device, they replace their supplier. How can a customer service organization adequately and proactively handle these two groups to help reduce its service cost and customer churn?
In another example, a fraud detection and management team would like to isolate methods fraudsters use to steal subscriptions and conduct illegal transactions. A very small percentage of transactions are fraudulent. However, according to the Communications Fraud Control Association (CFCA), they add up to USD5.22 billion in annual losses.1 Fraudsters keep finding new loopholes in telecommunications processes and systems to invent new ways of defrauding people. How can a fraud management organization make agile changes to its fraud detection and prevention software to effectively find and deactivate fraudulent subscriptions before they cause excessive revenue leaks?
Looking at another use case, broadcasting marketing campaigns to a wide-ranging market segment can be expensive for many marketing organizations, and they often result in poor campaign yields. Many consumer organizations are moving toward smart campaigns for targeting microsegments. They target a specific set of customers, personalize their campaigns, and learn from customer responses how to improve campaign effectiveness.
The marketers need to identify a specific microsegment that is highly likely to respond to a targeted campaign. In addition, the marketer needs to tailor its campaign based on individual preferences. A study conducted by Responsys found that the number-one reason customers are motivated to opt into mobile marketing campaigns is to receive special promotions. At the same time, more than one-third of subscribers to opt-in messages said they were actually unlikely to take action from a location-based offer.2 To maximize campaign effectiveness, the campaign management system should find the microsegment and personalize the campaign based on individual preferences.
Data flow and shared intelligence present challenges
As intelligent devices using Internet of Things data emerge, behavioral targeting opportunities and concomitant challenges are expected to grow. Automation for such opportunities needs to take into account two specific challenges. First, data flow should allow for the high-volume tsunami of data received at extremely high velocities. The architecture should use advanced techniques for filtering and focusing on data that meets specific conditions. The second challenge is that the intelligence gathered through historical analysis must be shared across the elements to learn from environmental changes and feedback from past actions. Machines need to learn to share intelligence to mimic human learning.
Part 2 of this series
presents a real-time adaptive architecture for effective behavioral targeting and leveraging a framework called D4 to represent its discover, detect, decide, and drive components.
Please share any thoughts or questions in the comments.
1 “2013 Global Fraud Loss Survey,” Communications Fraud Control Association. Note: registration is required to obtain a copy of the survey report.
2 “Mobile Marketing Engagement Study,” Responsys, February 2014. Based on an Ipsos Observer survey of 1,200 US adults from October 25 to November 6, 2013.
|[followbutton username='IBMdatamag' count='false' lang='en' theme='light']|