• Facebook
  • Twitter
  • LinkedIn
  • share

A data-driven digital strategy to manage churn for the media and publishing industry

Author: , Date:
Media and publishing industry needs to rethink new ways using a data-driven digital strategy to deal with Churn.





Media and Publishing companies primarily offering subscription-based services are facing market saturation, low entry barriers for competitors, and widespread price transparency, thus impacting growth, causing higher churn, and shrinking profits1. To counter the hypercompetitive market, media and publishing companies are embarked in digitalization i.e. adopting digital technologies to customize content and offer high-quality customer experiences at a lower cost, thereby potentially changing their business model and identifying new revenue and value-producing opportunities. 


However, digitalization has its own set of challenges such as growth of diverse customer base. This growth results in higher acquisition and retention costs. Companies will be successful in this effort only if they embrace a holistic view of a customer and scrutinize customer life cycle. As per CIO.com survey, for top performing digital marketers post-sale portion of the customer lifecycle is on relatively high priority. They invest 52% of the marketing budget on customer retention and expansion, whereas average marketers allocate 44%2. Customer retention strategy account measures churn reduction and cross-sell/ up-sell initiatives.  


This article discusses Churn Analysis and Prediction (CAP) strategy for digital media and publishing companies.




Search analyst Greg Sterling puts the average annual churn rate for search marketing clients at or above 50 percent, with some companies seeing even higher churn rates3. In general, customer churn occurs once a customer stops doing business with a company. Definition of customer churn is industry specific and is refined based on the product/service offerings and financial parameters. 


When does customer churn occur? 


In the context of digital media and publishing industry, customer churn occurs when any of the conditions are satisfied:


  1. Cancel subscription services/products and decide not to continue subsequent year  
  2. Did not re-order fixed term services/products and decide not to order subsequent year
  3. Continuous default of payments excluding bankruptcy 


Quantification of customer churn


To calculate customer churn, companies opt for standard metrics and KPI’s indicated below:


  • Sample Standard Metrics – These metrics are typically used to monitor the trend on a periodic basis. They are usually based on churn count and extrapolated for different periods.  Other groupings could be based on individual sales channels.
  • Sample Standard KPIs: These metrics are key in helping executives gauge severity of customer churn and take decisions based on their trends. They could be annual customer churn rate, Y-o-Y churn rate change, and lost sales per customer churn. 


Metrics and KPIs should be augmented with associated business functions that directly impact customer churn. They could be leading or lagging indicators. Some of them are:


  • Churn count due to complaints – These metrics should help provide a better understanding of the complaints due to a specific product or service such as frequency of complaints, resolution time, source of the complaints, and geographical distribution of the complaints. Many of these metrics could be a KPI after further analysis and could be a leading indicator for product or service acceptance. 
  • Proportion of churn due to complaints to the total churn count – These metrics should help companies determine if more efforts should be allocated towards customer service and post-sales support.
  • Churn due to sales representative inefficiencies – These metrics could help analyze sales rep performance better. Companies could focus on improving the sales training or process. 
  • Customer churn propensity – These metrics, as a best practice, is calculated using predictive modelling techniques and is a value range between 0-1.






As explained earlier, typically, in most companies, simple Business Intelligence reporting is available to determine churn. If the organization processes, technology and skills are advanced, predictive analytics could be incorporated in order to determine the factors that cause the churn. But, what is lacking is a clear strategy map to understand causal relationship between the overall corporate strategic goals and how the corresponding financial and business function objectives lead to those goals.


A strategy map is a visual representation to understand and communicate the objectives and perspectives within a company and facilitate organization to achieve specific goals or objectives. These specific objectives are grouped into four perspectives – Learning & Growth, Internal Business Processes, Customer and Financial. For example, refer flow A-B-C in the strategy map. Wall Street Journal reported that, an increasing number of advertisers attempting to land in places like Google are complaining about online marketing firms’ sales tactics and broken promises4. Sales training curriculum factoring customer switching behavior will develop sales team competence in terms of efficient sales visits planning and product recommendation. This will result in high CSAT and eventually low customer churn. 






Frame the hypotheses. Before one begins the strategy map exercise, it is important to frame a few high-level business hypotheses to understand the success criteria and measurements. For example,


  • Churn rate is uniform across sales channels 
  • Sales rep experience has direct correlation to customer churn count 
  • There is a correlation between count of certain complaint types & the churn rate
  • Monthly loss value due to churn is uniform for individual channels


Financial Perspective. The objectives of financial perspective must ask the following questions,


  • What do our shareholders expect us our financial success to be?
  • What are our most important financial outcomes?
  • Does the company strategy, implementation, and execution achieve the desired outcome?
  • What is your customer retention strategy? How can the retention cost be reduced?
  • What are the retention cost attributes?


A customer churn has a direct impact on lost sales, thereby, a financial objective could be to minimize lost sales due to churn.


Customer Perspective. The objectives of customer perspective must focus on the customer value proposition and areas for improvement in customer satisfaction and Life-time Value (LTV). As summarized by LTV equation5, reduction in churn rate directly influences the overall LTV of a customer:

Lifetime value of a Customer = [Average Annual Sales per Customer – Average Servicing Cost per Customer] / Annual Churn Rate


  • Why should a customer buy a product or service?
  • What are the most important product and service attributes? Relationship related attributes? Brand?


These objectives must clearly assist in determining the LTV of the average vs good customer.


Internal Business Functions Perspective. The objectives of internal business functions perspective must focus on areas for improvement in operations cost reduction, simplification of order management process, and product & service offerings.


  • How can the customer service be improved?
  • How can the product and pricing structure be simplified?
  • How can the customer facing website be improved?


For all the perspectives mentioned above, the objectives should have a corresponding measure, targets to achieve, and corresponding initiatives to achieve them.




Identify analytical data needs. As the objectives are framed, measures are defined, and targets are identified, consideration for the underlying business data attributes and their corresponding granularity is important. For example, historical order data, churn data, complaints data, employee & customer master. An analytical data set is then based on this data. 




Below diagram illustrates CAP approach:




Churn analysis & prediction approach begins with design activity whichrequires: 


  • Analysis of inter-functional dynamics to identify key business activities which can impact customer behavior and can cause churn
  • Analysis of time horizon – Identify e.g. analyzing data of past 6 or 12 months and providing insights and business action for future
  • Sampling techniques to identify training & testing data set for churn prediction model
  • Analysis-to-action lead time
  • Result output format e.g. flat file, excel sheet or file format compatible with existing applications


Next step is to leverage statistical techniques to manage outliers & missing values followed by descriptive analysis to visually check churn trend/ seasonality, correlation between KPI’s/metrics e.g. correlation between churn loss value & price sensitivity. Descriptive analysis helps in framing final set of churn related hypotheses. 


Churn Prediction Modelling: Once the analytical data set is defined, churn prediction modelling begins with exploratory data analysis (EDA) to statistically validate the hypothesis & identify significant variables based on information value. Based on output from the EDA, customized predictive models can be developed at region or Sales Channel (local, tele sales, high value) level. Depending on data size suitable for open source tools such as R/Python can be used to develop predictive model/s. 


For continuous provision of accurate insights and observations, it is absolutely necessary to monitor CAP outputs with actual churn data. In the following cases, analysis & churn model need to be recalibrated by conceptualizing/ including new KPI’s/ metrics.


  • Significant variation in the predicted churn count and actual churn count. In general, variation can be 5% or as defined by business users
  • Change in the calculation logic of existing KPI’s & metrics
  • New KPI’s & metrics included due to business transformation such as Digitization, change in product/service portfolio 


Operationalization of CAP output: Post descriptive analysis & churn prediction modelling, insights & observations can be translated into intuitive reports and dashboards with drill-down feature. As an example, below indicated dashboards highlights Y-o-Y churn trend, churn analysis across sales channels, sales loss due to churn, impact of complaints on the churn etc. Subsequent dashboard indicates drill-down at sales channel level to analyze trend of sales loss due to churn.   



Further insights & observations of CAP framework can be leveraged by critical business functions as follows:


  • Marketing
    • Design retention campaigns by factoring customer types or channel contributing significant churn count, redefine pricing strategy considering price sensitivity of customers who are about to churn. 
    • Calculate Customer Lifetime Value (CLV) for the customers who are likely to be active in the future. Churn prediction output can be used to filter customers with low or zero churn propensity to estimate CLV. 
  • Sales – Plan customer visits for high value customers who are most likely to churn, sales representatives (reps) can recommend relevant products/ service bundles, or re-structuring sales rep team
  • Customer service – Redefine complaint resolution SLA, optimize customer service team composition in view of complaint types resulting in high churn rate   




Venkatesh Thyagarajan

Senior Director, Global Business Analytics Head



Romi Malik

Manager - Data Scientist