In my previous article, I highlighted some strategies that help companies establish a culture devoted to customer value management (CVM). We have seen that over their lifetime, some subscribers generate recurring monthly revenue through simple regular bill payment or using their phones regularly. Others generate cost by using the network unfavourably and utilising call centres in a way that are not efficient for from the organization’s point of view.

CVM-centric organizations should bear in mind that customers give behavioural signals either intentionally or unconsciously all the time. Effective implementation of a CVM program that relies on advanced analytics capabilities will definitely have a strong impact on the overall success of such a program.

Building Analytical Capabilities

In contrast to a traditional mass marketing approach, building analytical

capabilities on campaigns that target specific customers based on their segments and scores showing customers’ probability of churn, provides a wealth of benefits to businesses.

Companies are collecting more and more data on their millions of accounts. Telecommunication operators gather more than 450 different variables for each of their subscribers to analyse and understand what lies beneath that huge amount of data.

By mining detailed usage patterns, operators can micro-segment their customer base into hundreds of relevant segments based upon usage characteristics. Operators can then target specific customer groups with micro-promotions that are more likely to result in increased usage and revenue stimulation. Let’s take a look at some basic steps to build an analytical model in the telecommunications environment.

For an effective analytical model, an organization’s data must be integrated, analyzed and properly prepared to meet the objectives of business. This model can meet many needs, such as customer segmentation or propensity to churn.

The process begins with the creation of an analytical data set which can be referred to as a customer analytic record (CAR). The CAR contains all the variables/attributes for each subscriber. It is important to note that well-defined and agreed-upon attributes will increase the success of the model in the subsequent phases. Examples of some attributes or variables might include the “number of days since the subscriber last logged onto the network”, “ratio of outgoing usage to incoming usage”, “number of days that subscriber uses specific service in a month”, “inactivity flags” and other usage characteristics. These usage behaviours can reveal highly insightful characteristics about subscribers.

Coming back to the model, the CAR structure should be denormalised, consisting of many columns related to particular subscriber for the purpose of analytical processing.

Once the data set is ready, it is time to build predictive models with data mining technologies. The very first stage is called “model training”. In this stage, with the given conditions based on the historical facts, specific target variables are selected and aimed to teach the model. The modelling system is expected to learn from the previous data feeds and historic usage of particular subscriber which then will be used as a deployment model for the future predictions.

Next is the execution phase, which is deployment of the model. Deploying the model enables the prediction of which subscribers are likely to churn based on their usage behaviours, as an example. By the way, it is worth noting that this process is more or less the same regardless of the technology. The data mining tools produce some algorithm or code that, in turn, can be used to produce a score for each subscriber. This algorithm is then passed to data warehousing system where the extracted data lies and queries on that data result in subscriber scores at the end.


What Else is Needed?

Building analytical capabilities that include segmentation of subscribers based on their hidden usage behaviours and churn analysis should be a primary focus for the successful CVM program. Yet, this is only the first step and is ineffective unless followed by a set of campaigns tailored to the individual customer groups.

Deriving knowledge about customers in detail from strong analytics studies is not enough. Organizations have to develop a team capable of designing and deploying marketing campaigns that are relevant to the customers and executing these campaigns successfully. It requires a strong cross-functional CVM team along with superior marketing, statistics and IT skills.

The challenge is to get the relevant teams working together effectively. Simply trying to drive CVM through the marketing department alone will likely to fail. There is a data extraction process which requires some IT expertise. IT needs to manage the data mart extraction process based on the defined rules and mine the data in collaboration with marketers. Even if the data is ready, advanced statistical skills are necessary to build relevant analytical models. After the tested model is approved, marketing can build their strategies and choose which customers to target. It is vital to perform pre-campaign tests on small subset of the target campaigns and compare the results against a control group. After the campaign execution, financial aspects of the campaigns should be measured to determine the financial ROI All these factors allow companies to fully realise the true potential of true CVM process.

Example: European Operator

One major European operator faced a market that had reached 100 percent penetration. Prepaid business accounted for 59 percent of the subscriber base,

but only 18 percent of revenue. Post-paid programs were tied to high acquisition costs, driven by a full subsidization of handsets. The prepaid business faced high churn rates with heavy rotation churn. It proved to be a major challenge when external channels expressed little interest in promoting prepaid. Aggressive subsidy levels offered by a competitor put added pressure on gross additions. The operator’s management had all but given up on prepaid in light of recent marketing campaigns and their lack of success.

In response to these challenges, however, management implemented a CVM program.

  •    The objective: to slow the churn rate and to develop up-sell mechanics.
  •     The approach: to set up a cross-functional CVM team, including members of marketing, IT, customer management and database marketing.
Management tasked this team with jointly implementing a new approach to marketing campaigns. Key activities included building a daily data prepaid data mart, standardized pre- and post-campaign ROI models and a streamlined campaign process. Thanks to the prepaid data mart, highly data-driven analytics could be used to develop consumer insights and micro-segments. This resulted in new campaign ideas. Ultimately, the team targeted a 4 percent increase in revenue. They also increased their campaign capacity from one per month to 10 per month. Post-campaign analysis could now be carried out in a matter of days instead of months. Overall, CVM delivered the levels of rigor and speed needed in a fast-paced and highly competitive market.

Gary Loveman, CEO of analytics competitor Harrah’s frequently asks the question, “Do we think this is true? Or do we know?” I believe organisations should develop that kind of mind-set if they desire to be classified as a “full bore” CVM competitor combining advanced analytical skills.


Link to original post