In modern telecommunications with the competition that arises between service providers, customer reimbursement and retention is a significant challenge. For the new entrants, acquiring the new customers is the highest priority, while for the incumbents it is important to retain the revenue earned by the customers.
Telecom companies can increase profitability by creating predictable modeling to identify potential churn candidates and non-revenue-generating customers; and can increase revenue and profitability through targeted promotional and promotional offers that will not only retain those customers, but also convert non-revenue-generating customers into profitable revenue-earning customers.
This article highlights the necessity of churn and campaign management and the use of SAS – Telecommunication Intelligence Software (TIS) for the purpose. It also includes various implementation challenges for SAS – TIS in the real – time scenario.
Acquiring and retaining customers is a significant challenge in all industries. In the telecommunications industry, if a customer fails before a company can make a return on the investments it has made in acquiring the customer, it will affect the profitability of the company. Therefore, identifying the profitable customers and retaining them is very critical.
As the telecommunications market becomes more competitive, it becomes increasingly difficult to determine the reasons why the customer leaves the company’s service. In this situation, it is even more difficult to predict the likelihood of the customer leaving in the near future. It is becoming more and more challenging to devise a cost-effective incentive to target the right customer to convince him to stay with the company.
Predictable modeling of core analytics and management aims to generate scores that show the likelihood that customers will perform in the future. This takes into account various aspects of customer sensitivity to churn, including the history of people who have previously struggled and builds a data model that generates an easy-to-understand reference numbers (scores) assigned to each customer. These customers are then targeted for incentives to determine their cancellation. In other words, Churn analysis determines the probable causes of a future cancellation, depending on past items, which will help companies adjust their offerings. For example: whose analysis reveals that many customers have withdrawn from a particular area last month, and a further study has identified that there are frequent call interruptions (service disruptions) in this exchange (or BTS area). It can be concluded that due to the technical inadequacy of the special exchange, frequent call interruptions are experienced which have contributed to customer dissatisfaction and their relocation. So further technical solution to this exchange may prevent future potential churns.
Business definition of Churn Management
Definition of churn is the first and most important activity in the design of Churn Management. Different companies define churn according to their business experiences.
Churn definition differs from a prepaid to postpaid scenario.
In a prepaid scenario, a customer may be considered violated in the following cases:
a) If the customer leaves the network (disabled)
b) If the customer is an active non-user (ANU)
A customer can be considered ANU when:
i. Customer has no outgoing or incoming use for the last (X) rolling days
ii. the customer has incoming use only but no outgoing use for the last (X) rolling days iii. If the customer’s use is below a predetermined (business settlement) amount for the last (X) rolling days.
In a postpaid scenario, a customer pays a rental monthly. So in case of non-use or minor use, the company earns a fixed income from each postpaid customer. Therefore, the customer is considered cone when leaving the network (Disabled).
Churn parameters for business analysis
After defining churn, the next activity is identifying the correct churn contribution parameters. Churn probability or churn score for individual customers can be generated based on the following categorical details:
1. Customer demographics Customer demographics related data is used to segment the entire customer base depending on:
d) Customer account information
e) Subscription life cycle
2. Billing and use:
Billing and usage related information obtained from the switch (Call Data Records) is mainly used for churn probability detection. The following details are used:
a. Price plan
b. Monthly Usage Summary (Call Count, Charged Data Volume, Free Call and Data Quantity)
c. Monthly surplus contribution
d. Declining payment
e. Managing channel information
f. Reloaded channel information
g. Network Product Information (Voice, Messaging, Data)
3. Technical quality:
Service quality is a potential churn driver as calls decline or poorer service quality increases customer dissatisfaction and therefore churn probability. In the case of CDMA, as the customer is closely associated with the handset equipment, the aging of the handset affects the likelihood of the customer getting churn.
The following details are used:
a. Counted call counts
b. Service quality
c. Age of equipment (age of handset in case of CDMA)
4. Contract Details: At the end of the contract period or the delivery period, the likelihood of the customer leaving the connection is high, therefore it has a major influence on the determination of churn. The following details are used:
a. Commitment period
b. Count of contract renewal
c. Current contract and end date
5. Event related:
Loyalty scheme or loyalty benefits are the main drivers of retention. Loyalty form data is used for churn scores.
Identification of the source systems:
After deciding the Churn parameters, the next step is to identify the source systems from which the respective data is extracted.
Cusomer details from CRM system
Use and billing related details from billing system
Technical quality from Exchange & CellSite
Activation details from the Provisioning system
Data management is the basis for a business analysis. Real data must be present in the right place.
Data management has three parts:
Extraction: involves extracting data from source system and loading into data exchange layer
Transformation: Includes validation of the extracted data (eg, validation for unique keys), creation of joining conditions between the tables, cleaning of invalid data, etc.
Load: Includes data loading in the Business Intelligence Data Warehouse
Data Modeling and Churn Score Generation
When the approved data is available in the data store, the data modeling is performed. It is an iterative process. The quality of the model is available and the model that provides the best business value is considered. This model provides results in the form of churn scores for individual clients, which can be used to determine campaign goals.
Using churn scores for storage campaigns
The data model generates the individual customer’s churn score, ranging from 0 to 1.
0 – Indicates the least likelihood that the customer may black out
1 – Indicates the highest probability of customer cheating.
These scores are weighted components of various parameters such as
Decrement (promotional and key) information
Quality of service
Customer service / complaints
Price Plan Sensitivity
Business decision must be made to determine an upper threshold for churn scores. Customers above this threshold need further analysis (eg: customers with a score of 0.7 and above). The top two parameters contribute to the churn score generated at the individual customer level (for customers who have a churn score greater than the threshold). Depending on these parameters, retention campaign can be conducted. The parameters can be as follows:
Usage Statistics: Usage behaviors can be derived from the combination of decrement (promo and core), balance and recharge information. The customer who has a higher score in “less use” can be targeted with offer pricing quotes to improve his / her usage and convert that customer from non-revenue to revenue.
Higher off-net use: The higher “off-net use” score means that the particular customer has called other networks very often. A targeted campaign can be carried out with the pricing plan that is advantageous for calling other networks. Further analysis of the called off-line numbers may result in identifying often-called off-line numbers that can be targeted by promotions as a business candidate.
Handset Features: The handset used by the customer may be outdated and lack the modern features. In this case, the likelihood of the customer switching to a newer handset is significant and there is considerable sensitivity for that customer to move to another service provider that has assembled a handset offer. A retention campaign can be targeted (to this group of customers who have a high handset churn score) with new service offerings along with handsets.
Customer service / complaints: The higher score in customer service / complaints means that the customer has often called customer care and the probability that the customer is dissatisfied with the service is higher. Further investigation of customer call interaction details may reveal the reason for frequent customer service calls. After performing campaigns on the basis of churn scores and churn drivers, the campaign response must be captured and entered into the campaign to analyze campaign success.
Implementing Churn Management Solution Implementation Step
The following phases are involved in implementing the Churn Management solution:
1. Requirements analysis: At this stage, business requirements are gathered and analyzed and business definitions for core are decided
2. Solution Assessment: In this phase, the business intelligence solutions are evaluated with the high level requirements of the implementation company. The feasibility test is performed depending on the high business requirement and data availability.
3. Detailed Analysis / Detailed Design: In this step, the business requirements for the Churn Management project are analyzed in depth for the design, development and improvement of the project. An exercise is conducted to understand the availability / unavailability of information required to meet the business requirements and data mapping from the source system.
4. Data Analysis – ETL: In this step, the data is extracted from the source system, transformed (cleaned / modified for missing fields and the data quality analyzed) and then loaded into the Data Warehouse in the business intelligence tool.
5. Data Modeling: In this step, the analytical data models are created using statistical methods (eg: logistic regression method) on historical data for predicting churn scores and analytical base tables are populated by data.
6. Reporting: Churn scores (0-1: 0 – means less likelihood of churn, 1- Maximum likelihood of churn) are generated at each customer / account / subscription level and the corresponding report is generated.
7. User Acceptance Testing and Deployment: After completing a successful UAT, the software is rolled out for business users.
There are several challenges when implementing a business intelligence solution on a huge scale from millions of customers.
The best time for implementation is consumed by data management. Data management spends 75% of the total deployment time. Data management includes:
Identification of source systems from which data is to be extracted:
Due to the involvement of multiple source systems (CRM, Provisioning system, Billing, Mediation systems, etc.), identifying the right source system for different data fields is becoming increasingly difficult. Identifying the correct data source and mapping to DIL fields spends most of the implementation time. If the data source mapping is incorrect, the subsequent implementation steps (modeling, analysis) will also be incorrect. Therefore, special care must be taken during the data collection exercise.
Data quality: Data taken from the source systems must be high quality and error free. The biggest challenge when implementing a business analytics solution is to get high quality data. Cleaning up data and filling in the missing fields takes a long time on implementation.
Change Management: With the implementation of a BI solution, users need to change the way they used to perform churn prediction and campaign management. Therefore, the user’s adaptability and user awareness must be built up through appropriate training sessions
To make the Business Intelligence system operational: After implementation, a specific organizational structure for managing BI operations must be planned and resources must be trained within the required areas.
SAS in business analysis
SAS is a leading business analytics software and service provider within the business intelligence domain. It has provided proven solutions to access relevant, reliable, consistent information across organizations to help them make the right decisions and achieve sustainable performance improvement and mitigate risks.
SAS has an expanded capacity to handle large-scale data (using SAS-SPDS – scalable performance data server). This, combined with a strong programming language and enriched graphical interface, has differentiated it from the other analytical tools available in the market. This makes SAS perfectly suited for enterprise use where it requires the management of huge data stores.
SAS – Telecommunication Intelligence Solution (TIS)
SAS has several industry-specific solutions. SAS has packed their business analysis knowledge in the form of models, processes, business logic, queries, reports and analysis.
TIS is the telecom industry specific business analytics solution built specifically for the needs of the telecom industry. This solution helps telecommunications service providers with specific modules, for example:
SAS Telecommunication Campaign Management
SAS customer segmentation for telecommunications
SAS customer storage for telecommunications
SAS Strategic Performance Management for Telecommunication
SAS Cross sells and Up sells for telecommunications
SAS payment risk for telecommunications
SAS’s churn management and campaign management solution includes segmentation of the entire customer base
Discover the root causes
Scoring the individual customer based on their likelihood of churn
This churn score is further used as an input for campaign management.
SAS Data flow (Architecture)
The data must be collected from different source systems.
CRM system: customer / account / subscription related data
Delivery system: Activation date, equipment (handsets) Age Billing system: Billing data
Mediation system: Call record information
The data is collected in the Data Interchange Layer (DIL). Then extract, transform and load detail data (DDS).
The data is used for:
1. Dimensional data modeling: This is used for query, reporting and OLAP (online analytics)
2. ABT (Analytical Base Table): This is the development-specific model that has been developed and can be used for a specific analysis. For example: ABT to core model.
3. Campaign Data Mart: This data is used to target specific customer segments for targeted campaign.
Therefore, it is imperative that core management is a significant challenge in the modern Indian telecommunications industry. Detecting the real cause of the kernel and predicting the kernel in advance can save the company from a significant loss of revenue.
Business Intelligence tools help telecom service providers perform data analysis and predict the churn probability of a particular customer. Apart from predictable core analysis, the tools can be used for various other analytics to help make business decisions.
SAS has the potential to handle huge amounts of data. As a business intelligence tool, SAS allows the company to efficiently manage huge amounts of data and perform analysis of the information available to millions of customers. In addition, with its telecommunications-specific solution (TIS – Telecom Intelligence Solution), SAS helps build the data warehouse to contain the required parameters for further analysis.
Therefore, SAS-TIS can be an effective tool for business intelligence activities in the telecom industry.
Link: SAS company information: http://www.sas.com/
Link: Arindam’s profile: http://in.linkedin.com/in/arinmukh