Get fresh updates from Hortonworks by email

Once a month, receive latest insights, trends, analytics information and knowledge of Big Data.


Sign up for the Developers Newsletter

Once a month, receive latest insights, trends, analytics information and knowledge of Big Data.


Get Started


Ready to Get Started?

Download sandbox

How can we help you?

* I understand I can unsubscribe at any time. I also acknowledge the additional information found in Hortonworks Privacy Policy.
closeClose button
January 04, 2017
prev slideNext slide

HDB: Driving Better Customer Experiences in Mobile Network Services

Bob Glithero
Analytics Product Marketing Manager, Pivotal

Over the last five years, mobile network operators (MNOs) realized 15% lower compound revenue growth on average than other types of communication service providers. With few exceptions, MNOs globally have seen a long-term decline in average revenue per user (ARPU).

To reinvigorate growth, innovative MNOs are searching for new ways to exploit data to better serve their customers and drive new revenue opportunities. But pulling together data scattered across legacy transactional and BI systems is difficult and slow. Worse, these systems were not designed to analyze the volumes of unstructured and semi-structured data exhaust from user plane and control plane traffic. But this data offers the potential for deep, integrated insights for operational, customer care and marketing needs. Hortonworks HDB, powered by Apache HAWQ (incubating), is a Hadoop-native SQL database that provides a new approach for analyzing mixed petabyte-scale data with blazing speed for timely, customer-centric insights.

Fighting Market Headwinds

A number of market headwinds are having profound effects on MNOs’ ability to continue to grow revenue. These include:

  •  Competition from free or ad-supported apps. Revenues from bread-and-butter services like voice and text are being displaced by over-the-top (OTT) messaging apps like WhatsApp, Messenger, and Skype.
  • Market saturation. In developed markets, most subscribers already own a mobile phone and a data plan. There are simply fewer net-new subscribers to target. Increasing ARPU from the current base is a key to top-line growth.
  • Prepaid subscribers. Low-margin prepaid subscribers are a large part of most operators’ customer base. Unlike subscribers on contract, there’s typically very little information about prepaid subscribers in CRM systems that can be used for marketing intelligence. Converting these prepaid subscribers to post-paid contracts, or developing new ways to monetize their prepaid segments, is a key challenge for operator marketing teams.
  • Data usage outpacing revenue. Even as new services like LTE and Voice over LTE are increasing capacity and reducing the cost per megabyte of delivery, these services also drive massive increases in signaling traffic. In fact, by some accounts signaling traffic on LTE networks is growing more than twice as fast as user data traffic; little of this unstructured data is monetizable using operators’ legacy BI infrastructure.

To compensate, MNOs typically run pricing promotions or other financial incentives to improve subscriber acquisition. Unfortunately, the impact is usually short-term: financial incentives train subscribers to shop on price alone and position network services as commodity products. This tends to make subscriber churn even worse. MNOs instead need to find ways to position their services as high-value offers worthy of higher pricing.

Advanced analytics put the focus on customers
Innovative MNOs are trying to change the rules of the industry from competing on price to competing on quality of experience, in an attempt to keep subscribers loyal and justify premium margins for next-generation services. In addition to investing in advanced services like LTE and 5G, this includes looking for new ways to exploit MNO’s data assets for insight. Specifically, operators are enriching structured data from legacy BSS/OSS systems (e.g., CRM, billing, provisioning, device inventories, or trouble ticketing) together with unstructured and semi-structured data from clickstreams, data flows, signaling data, sentiment analysis, and voice of the customer solutions (e.g., post-call surveys). Innovative MNOs analyze these new data sets to support a number of emerging use cases, including:

  • Service assurance. Understanding which customers and VIP groups are impacted by service problems is essential for solving problems quickly and maintaining customer loyalty. However, network operations teams increasingly wonder how well their fault and performance dashboards correlate with their subscribers’ actual experience. Most network management tools lack any type of customer context — they give an engineering perspective on device-level issues but not the impact on subscribers, or whose issues to prioritize. Worse, in a multi-vendor environment, network teams are forced to manually piece together the end-to-end service picture from data trapped in fragmented, siloed collections of fault and performance dashboards. None of this is conducive to rapid resolution of customer issues.
  • Customer experience management. Deflecting calls, avoiding device exchanges, and accelerating problem resolutions are top of mind for customer support teams. They need at-a-glance intelligence on key service quality indicators, rapidly aggregated from granular performance indicators such as call setup times, latency, throughput, call drops, jitter, packet loss, and other detailed metrics. They also need device-level insights to quickly identify and fix common issues such as troublesome apps and phones.
  • Customer profiling. Subscribers that take a mix of digital apps and services offer good revenue potential from new, high-margin telemetry services like home automation, connected cars, and IoT devices. Understanding who these subscribers are and how to target them is essential to drive revenue. Innovative MNOs are increasingly analyzing clickstream and app usage data to help marketing teams develop behavioral insights to better segment subscribers. This type of analysis can even reveal inferred demographics and useful segmentation approaches for otherwise anonymous prepaid subscribers. These insights pinpoint subscribers likely to respond to treatments from advertising or promotional campaigns, or from new revenue strategies based on data brokerage to partners, like sponsored data.

MNOs’ legacy BI systems remain essential for user productivity with core analytics, but they are not optimized or capable of coping with the variability, velocity, and volumes of data generated by clickstreams, flows, and other unstructured sources.

This data also needs to be analyzed in time to be relevant and useful. For example, operators would typically expect their customer service agents’ dashboards to be updated with current performance statistics in a sub-minute time frame. A high degree of concurrency of requests for aggregation or filtering from different users running in parallel is also necessary.

As a result, business users struggle to apply general purpose BI tools to these new use cases. Querying tools in the Hadoop ecosystem are capable of working with petabyte-scale data, but they typically offer a reduced SQL dialect, no in-database machine learning functions, and limited ability for rapid, iterative querying.

HDB: The Hadoop-native SQL database to bridge BI and advanced analytics
To bridge the worlds of traditional BI and unstructured Big Data – and support the types of innovative use cases highlighted above – we offer HDB, a high-performance analytics database that runs within a Hadoop cluster. HDB combines the familiarity of a full ANSI SQL interface, the performance of a massively parallel processing (MPP) engine, and the power of in-database functions for advanced analytics, data science, and machine learning at scale.

To deliver faster time to insight, HDB executes SQL inside the Hadoop cluster. It operates directly on Hadoop Distributed File System (HDFS) data without translating queries to MapReduce. HDB can query all data in HDP, regardless of format. HDB can quickly and efficiently federate queries across Hive, HBase, and HDFS via the HAWQ eXtension Framework (PXF). It also works with file formats like JSON, Parquet, ORC, and Avro.

To deliver advanced analytics use cases, like predictive analytics, data scientists can access Apache MADlib (incubating), a library of highly parallel machine learning algorithms, from within HDB via familiar SQL syntax. By running analytics within the database, analysts can leverage the entire data set where it resides in the HDP cluster, rather than relying on sampling. This helps data scientists create and deploy better models more quickly.

For ease of management, HDB integrates with Hadoop ecosystem components such as Ambari system management and YARN for resource management, to provide a completely open, integrated, enterprise-class data platform.

Taking the low-risk road to analytics-driven innovation
HDB is proven with mobile operators. For example, one mobile operator wanted to achieve more real-time reporting on call faults, analyze network traffic patterns for optimization, and broker subscriber data on web and app usage to advertising partners. To meet their requirements for fast querying and analysis, the customer built an application using the Spring IO development framework, Pivotal GemFire for real-time event notifications, and HDB for analysis of aggregated data in Hadoop. With this advanced analytics application, the customer was able to improve reporting on call faults, and to broker analytics on customer footfall traffic for their enterprise customers.

In another case, an operator was suffering from poor network quality, dropped calls, and loss of customers. Its legacy data warehouse could not keep up with the volume of network data being generated — about 2 billion call detail records (CDRs) per day. A lot of valuable network data was simply dropped due to lack of storage and a lengthy ETL process. With HDB and HDP, the operator was able to mine all their CDRs, analyze dropped calls, and improve network service. Reports that had taken two months to generate are now delivered in a day. Better yet, the operator was able to do all this at a much lower cost than with their legacy data warehouse.

Innovative operators are starting to mine the vast troves of unstructured data now available to them to help develop compelling customer experiences and uncover new revenue opportunities. On February 7, join us for a webinar to learn how HDB’s in-database analytics enable advanced use cases in network operations, customer care, and marketing for better customer experience. Register now, and get started on your advanced analytics journey!


ohhani baek says:

It is a Nice Article.Thanks for Providing information about the Hadoop advanced analytics mobile network services is a very good idea.

whatsapp++ ipa says:

Hadoop is the trending technology in recent years and it has a lot of features to overcome all the problems.Really it is an informative post to know about Hadoop advanced analytics in mobile network services. WhatsApp plus is the best messaging app for better communication and also for sharing.

Leave a Reply

Your email address will not be published. Required fields are marked *