Enhancing Customer Analytics with Big Data Integration  

Oleh:

Sehgal, Rajesh ; Srivastava, Shalabh

Jenis: Article from Proceeding
Dalam koleksi: 12th ANQ Congress in Singapore, 5-8 Agustus 2014, page 1-7.
Topik: Big Data; Telecom; Predictive Model; Churn and Tool
Fulltext: PM2-2.1-P0354.pdf (840.77KB)

Isi artikel

As a part of its global analytics operations Wipro is engaged with a global telecom service provider to take care of their multiple service lines. This includes the dedicated complaints queue based out of multiple geographies where customer service centers handle the incoming requests of troubleshooting complaints. A key factor in analyzing the complaints data was to dynamically identify the propensity of a customer to withdraw from the network with least latency from the time of his/her calling. For analyzing this propensity to stay or not stay in network (in Telecom parlance frequently referred to as “customer churn”) details pertaining to service package type, account type, type of queue, tenure of the customer with the provider etc. were looked into. Using this customer geo-demographic data a dynamic, robust and in-situ churn measurement framework was developed and to reduce systemic latencies it was integrated with the providers in house cloud based CRM. Two specific end goals were looked into: - Specifying the CRM capabilities vis. a vis. the churn tool integration - Analyzing the current retention policy to identify the best way of mapping churn prediction value (for a customer) with the best possible retention offer To this data, which emanated from proprietary relational databases, a plan of overlaying unstructured data was proposed. This included: - Customer comments in textual format (emails, flat files etc.) - Social media buzz pertaining to VAS, customer plans etc. on customized provider portals and public sites A combination of public data, streaming social media data and relational data was analyzed using a combination of open and proprietary big data platforms, which included: - HDFS for data storage and use of Apache SQOOP to funnel structured data in it - Using Mongo DB/ NoSQL databases to store the unstructured text data - Using open source platforms/languages like R, Hive to query the HDFS (or any other storage platform) The outcome was measured both in terms of information gain (that in addition of big data bought forth) and the dependency and cost savings achieved by building in-house expertise on platforms as opposed to subscribing to third party platforms

Opini Anda

Klik untuk menuliskan opini Anda tentang koleksi ini!

Kembali