Home > Broadband > Wireless Churn, Metrics and Big Data

Wireless Churn, Metrics and Big Data

Churn has a simple definition for a wireless operator – it is the number of net deactivations (i.e. gross adds minus net adds) divided by the average number of the subscriptions during the year. Mobile telecommunication market has changed from a rapidly growing market, into a state of saturation and fierce competition. The focus of telecommunication companies has therefore shifted from building a large customer base into keeping customers ‘in house’. Customers who switch to a competitor are so called churned customers. Churn prevention, through churn prediction, is one way to keep customers ‘in house’. In contrast to post-paid customers, prepaid customers are not bound by a contract. The central problem concerning prepaid customers is that the actual churn date in most cases is difficult to assess. This is a direct consequence of the difficulty in providing an unequivocal definition of churning and a lack of understanding in churn behavior. In the telecom service industry, churn can be of several types -

  • Involuntary churn: This occurs when subscribers fail to pay for service and as a result the provider terminates service. Termination of service due to theft or fraudulent usage is also classified as involuntary churn.
  • Unavoidable churn: This occurs when customers die, move or are otherwise permanently removed from the market place.
  • Voluntary churn: Service termination on the part of the customer when leaving one operator and possibly for another because of better value

In reality, it is very unlikely that MNOs could differentiate unavoidable and voluntary churn and predict them separately.

Nokia Siemens Customer Acquisition & Churn Study

 Churn can be shown as follows: 

Monthly Churn = (C0 + A1 – C1) / C0

Where:

C0 = Number of customers at the start of the month

C1 = Number of customers at the end of the month

A1 = Gross new customers duringh the month

 As an example, suppose a carrier has 100 customers at the start of the month, acquires 20 new customers during the month, and has 110 customers at the end of the month. It must have lost 10 customers during the month, 10 percent of the customers it had at the start of the month.

According to the formula:

Monthly Churn = (100 + 20 – 110) / 100 = 10% 

Causes for Churn

In an intensely competitive environment, customers receive numerous incentives to switch and encounter numerous disincentives to stay. There are many reasons to leave a provider but there might be just one reason to stay. Brand loyalty, customer service and great experience are some of the main reasons for customers to stay on with the current MNO.

Price: Particularly in the wireless and long-distance markets, carriers often offer pricing promotions, such as relatively low monthly fees, high-volume offerings (fixed number of minutes at a reasonable fee per month), and low rates per-minute. These price incentives can provide residential customers, in particular, with powerful incentives to change carriers.

Service quality: Lack of connection capabilities or quality in places where the customer requires service can cause customers to abandon their current carrier in favor of one with broader reach or a more robust network.

Fraud: Customers may attempt to “game the system” by generating high usage volumes and avoiding payment by constantly churning to the next competitor.

Lack of carrier responsiveness: Slow or no response to customer complaints is a sure path to a customer relations disaster. Broken promises, long hold times when the customer reports problems, and multiple complaints related to the same issue are sure to lead to customer churn.

Brand disloyalty (or loyalty to another): Brand issues may arise due to service or other issues experienced over time, mergers or acquisitions involving the incumbent carrier, or entry into the market of another carrier with strong brand recognition and reputation. Marginal brand loyalty can often be overcome by competitors’ incentives.

Privacy concerns: Consumers have an increasing awareness that companies they deal with have a lot of information about them, including their spending habits, personal financial information, health information, and the like. Breaking of privacy promises, publicized privacy problems, telemarketing, and other issues are causing many customers to consider their personal privacy as an asset and they are holding their service providers responsible for keeping privacy promises.

Lack of features: Customers may switch carriers for features not provided by their current carrier. This might include the inability of a particular carrier to be the “one-stop shop” for the entire customer’s Communications needs.

New technology or product introduced by competitors: New technologies such as high-speed data or bundled high-value service offerings—create significant opportunities for carriers to entice competitors’ customers to switch.

New competitors enter the market: The mere existence of viable competitors to the incumbent carrier may cause certain disloyal customers to churn. Further, as competitors enter new markets, they often offer short- or long-term incentives to new subscribers to build market share.

Billing or service disputes: Billing errors, incorrectly applied payments, and disputes about service disruptions can cause customers to switch carriers. Depending on the situations, such churn may be avoidable.

Nokia Siemens Customer Acquisition & Churn Study

MNOs can address churn in four ways, by

  • increasing subscriber satisfaction
  • increasing subscriber lock-in, independent of satisfaction
  • launching measures that specifically aim to improve customer retention
  • Addressing the artificial forms of churn, such as box breaking.

The four strategies can be pursued independently, but MNOs need to be aware that they are interlinked and each will have an impact on the others. Specifically, the measures are part of a linear development. Improving subscriber satisfaction will increase lock-in and lead eventually to lower churn rates.  However, operators can increase the amount of customer lock-in without improving satisfaction – for example, through the use of longer contract periods or the bundling of multiple services. Alternatively, MNOs can seek to address churn directly without addressing the underlying issues of subscriber satisfaction or loyalty, typically through heavy price discounting.

Industry Metrics around Churn

Mobile industry measures some metrics that revolve around churn and a customer behavior. The mobile industry in the US operates in a different model when compared to Asia or Europe. Several regional and behavioral patterns help in making the decision and the play. Also sometimes churn is a relative term, MNOs intentionally would like to lose some customers who are cannibalizing certain aspects of their business and are not profitable eg. High data users on an unlimited data plan, etc.

The cost of acquiring customers has increased incrementally in line with more advanced (and costly) Smartphone equipment. Often reported as CPGA (Cost per Gross Addition), acquisition costs include retail, administration, marketing and handset subsidy costs. While efficiencies in retail, such as driving customers through online channels, have helped to cut acquisition costs, the average cost of subsidizing equipment has grown. It is defined as –

Cost Per Gross Add - A ratio used to quantify the costs of acquiring one new customer to a business. Often, the CPGA ratio is used by companies that offer subscription services to clients, such as wireless companies and satellite radio companies.

Not all subscribers have equal ‘value’ and operators must look carefully at strategies to retain their high-value and loyal customers. ‘Customer Lifetime Value’ (CLV) models provide a useful way to apportion a dollar value to a subscriber (or subscriber group) based on cumulative cash flow from a subscriber relationship and the benefits of loyalty and advocacy that increase over time. They provide marketing teams with a means to gauge the effectiveness of their acquisition costs and retention strategies. CLV modeling may also be used with the goal of changing the behavior of different customer segments to consequently change their lifetime value.

Customer Lifetime Value (CLV) – a financial concept that represents how much each customer is worth in dollar terms and therefore exactly how much a company should spend to acquire and keep each customer.  CLV is calculated using a model and inputting various estimates and simplifying assumptions.  In reality, there are several variations of CLV available due to the complexity and uncertainty of customer behavior.

In wireless, CLV can also be:

CLV = ((ARPU – Variable CCPU) x Tenure) – (SAC + Capacity Charge)

CLV in wireless:

CLV (customer lifetime value) calculation process consists of four steps:

  • forecasting of remaining customer lifetime in years
  • forecasting of future revenues year-by-year, based on estimation about future products purchased and price paid
  • estimation of costs for delivering those products
  • calculation of the net present value of these future amounts
  • Forecasting accuracy and difficulty in tracking customers over time may affect CLV calculation process

Cash Cost Per User (CCPU) – Measure of the monthly cost to serve a customer, derived by dividing total operating costs by average number of users.  It is a measure of the monthly costs to operate the business on a per subscriber basis consisting of costs of service and operations, and general and administrative expenses of consolidated statement of operations, plus handset subsidies on equipment sold to existing subscribers, less stock-based compensation expense.

Cannibalization – A reduction in sales volume, sales revenue, or market share of one product as a result of the introduction of a new product by the same producer. If a company is practicing market cannibalization, it is eating its own market. For example, say Pepsi puts out a new product called Pepsi chill, and customers buy Pepsi chill instead of regular Pepsi. Although sales may be up for the new product, these sales may be eating into Pepsi’s original market, in which case the overall company sales would not be increasing. Because of the possibility of market cannibalization, investors should always dig deeper, analyzing the source and impact of the success of a company’s new but similar product.

Identification of cannibalization is by no means clear-cut and needs to take into account of the dynamics of the market. This needs examination by three methods.

  • Gains loss analysis
  • Duplication of purchase
  • Deviations from expected share movements

Read more on Wireless Metrics from my eBook on Amazon – http://amzn.to/xAzV6W

Churn Analysis – Game Theory, Big Data Analysis models (using Hadoop)

Game theory has been recognized as a cornerstone of micro-economics that can be applied to analyze problems with conflicting objectives among interacting decision makers. In any competitive market, customers may operate as a pool wherein they agree to pay for services that in a way also determine the value of the commodity. Thus, game theory offers a powerful mathematical tool to study different problems in science and engineering from an econometric point of view. It has been extensively applied to various competitive environments including the energy market, airlines industry, and Internet services. Recently, it has also been successfully used to deal with problems in the field of networking and communications .This is because the service quality each user receives in a competitive environment is often affected by the action of other users trying to gain access to the same network resources.

Let us understand game theory as widely used in the economics domain to model interactions among parties with conflicting interests, where each party is called a player. In a game, each player’s strategy has impact not only on his/her own payoff, but also on other players’ payoffs. Depending on whether cooperation is allowed among players, games can be divided into cooperative and non-cooperative categories. The most basic form of non-cooperative games is a two player game in which each player has a set of strategies with associated payoff values; each player makes an independent decision on a strategy so as to get the most out of the game on the basis that the other player is not cooperating. Thus, the outcome of the game is to find a pair of strategies, one for each player, who optimizes the payoffs of both players. Games can be played in two forms

  • Normal form where each player makes a strategy decision without knowing the decision of the other player
  • Extensive form where at least one player has partial information about the other player’s decision

Mathematically, a two-player non-cooperative game consisting of players P1 and P2 is defined by payoff matrices A and B, respectively. Assume P1 has m strategies denoted s1, s2, …, sm and P2 has n strategies denoted as t1, t2, …, tn. Thus, the rows in the payoff matrices represent P1’s strategies, while the columns represent P2’s strategies. More precisely, the element aij of matrix A defines P1’s payoff when P1 chooses strategy si and P2 chooses strategy tj. The element bij of matrix B is P2’s payoff when P1 chooses si and P2 chooses tj. Since this type of game is defined by two payoff matrices, it is also called a bimatrix game.

Games can be divided into zero-sum and non-zero-sum. For a two-player game, if aij + bij = 0 for 1 <= i <= m and 1 <= j <= n, it is a zero-sum game; otherwise, it is a non-zero-sum game. In a zero-sum game, a player’s gain in payoff results in loss in the other player’s payoffs; in non-zero-sum games, a certain strategy change could result in gain for both players.

In a bimatrix game defined by the payoff matrices A = [aij] m×n and B = [bij] m×n, a pair of strategies {si *, tj *} is said to constitute a non-cooperative (Nash) equilibrium solution to the game if the following pair of inequalities is satisfied:  

Intuitively, the Nash equilibrium is the points where no player in the game can improve his/her payoff by changing his/her own strategy, if all other players’ strategies remain unchanged. In other words, the Nash equilibrium is the point where there is no incentive for players to change their strategies if there is no cooperation among them.

Churn analysis requires analysis of customer billing data, network performance data, call detail records and several other factors including but not limited to sales and promotional data for any operator. Again Call detail records and billing data involves a huge data set and requires large databases and analysis engines. This can be done with specialized databases and analysis engines like Hadoop, Netezza, etc.

Apache Hadoop is a software framework that supports data-intensive distributed applications under a free license. It enables applications to work with thousands of nodes and petabytes of data. Hadoop was inspired by Google’s MapReduce and Google File System (GFS) papers. Hadoop is a top-level Apache project being built and used by a global community of contributors, written in the Java programming language. Yahoo! has been the largest contributor to the project, and uses Hadoop extensively across its businesses. The Yahoo! Search Webmap is a Hadoop application that runs on more than 10,000 core Linux cluster and produces data that is now used in every Yahoo! Web search query.

Hadoop was created by Doug Cutting, who named it after his son’s toy elephant. It was originally developed to support distribution for the Nutch search engine project. LinkedIn has been making heavy use of Apache Hadoop and Pig with its People You May Know and skills features (among others), and has pulled together a lot of User Defined Functions (UDFs) for Pig in the process. Amazon uses a cloud computing service that uses Hadoop, to crunch tons of data. This service, called Amazon Elastic MapReduce is designed for businesses, researchers and analysts trying to conduct data intensive number crunching (statement).

Specific advantages of the Hadoop implementation of MapReduce include:

  • Ability to write MapReduce programs in Java, a language which even many non-computer scientists can learn with sufficient capability to meet powerful data-processing needs
  • Ability to rapidly process large amounts of data in parallel
  • Can be deployed on large clusters of cheap commodity hardware as opposed to expensive, specialized parallel-processing hardware
  • Can be offered as an on-demand service, for example as part of Amazon’s EC2 cluster computing serviceOne example of using Hadoop for eScience is implemented by the Survey Science Group at the University of Washington. Click here for more information on Hadoop and Astronomical Image Processing.

These were the top three reasons mentioned for using Hadoop

  1. Mining data for improved Business Intelligence
  2. Reduces the cost of data analysis
  3. Log Analysis

T-Mobile USA utilizes big data as a federated and multi-dimensional dataset. The company has overcome challenges from a disparate IT infrastructure to enable regional marketing campaigns, more advanced churn management, and an integrated single-screen “Quick View” for customer care. Using its data integration architecture, T-Mobile USA can begin to manage “data zones” that are virtualized from the physical storage and network infrastructure.

 

Brett Sheppard from O’Reilly media discuses these details here.

Did you like this? Share it:
  1. Patricks
    July 30th, 2012 at 02:46 | #1

    Great exposition. Good job. How can Telcos benefit from big data without infringing on privacy of customers. How best can telcos survive using data at their disposals as another source of revenue or as a means to reducing cost.

    • July 31st, 2012 at 19:55 | #2

      Hi Patrick,

      A very good question, Telcos have acces to cutomer data including call detail records at all times but they follow an Industry compliance for handling such private data.
      The information that is specially helpful for MNOs are things like dropped calls, lack of coverage and subsequent lack of call origination or data degradation. Another aspect that helps operators is the survey collected from churning customers who leave a certain MNO for mobile handset upgrades. eg. iPhone.

  2. Ron Smith
    December 21st, 2012 at 07:21 | #3

    What have you found to be a "acceptable" cost per acquistion for the wireless space?

  3. Mehmet Erdas
    May 28th, 2013 at 07:40 | #4

    E2E CEM SQM NPM aqnd Cost and Billing CDR and Billing data are the main data sources to prevent Churn. But how about modelling churn process as a stochastic process with transition probabilities, map to garaph as vertexes and edges or use game theory approach to analyze the valuntary churn.
    Met

  1. No trackbacks yet.

Switch to our mobile site

Stop Copying Plugin made by VLC Media Player