{"id":1908280,"date":"2023-01-19T11:58:30","date_gmt":"2023-01-19T16:58:30","guid":{"rendered":"https:\/\/wordpress-1016567-4521551.cloudwaysapps.com\/plato-data\/churn-analytics-in-telecommunications-company\/"},"modified":"2023-01-19T11:58:30","modified_gmt":"2023-01-19T16:58:30","slug":"churn-analytics-in-telecommunications-company","status":"publish","type":"station","link":"https:\/\/platodata.io\/plato-data\/churn-analytics-in-telecommunications-company\/","title":{"rendered":"Churn Analytics in Telecommunications Company"},"content":{"rendered":"

Introduction<\/span><\/h2>\n

What is Churn Analytics? And how do telecommunication companies effectively use this analysis in day-to-day activities? Learn from the industry expert Sakshi Gujral who will take you through all the essential details and give you some tips on improving churn analytics results when used practically. <\/span><\/p>\n

[embedded content]<\/p>\n

About Speaker:<\/strong> Sakshi Gujral is currently working as a Data Scientist at Concentrix. She is also pursuing her Ph.D. from IIIT- Delhi. Sakshi is a GATE Scholar and UGC-NET qualifier, and an alumnus of the Defense Research Development Organization. She has 5 years of experience in Industries like TCS, and Genpact, solving Data science problems in the domains of finance, health care, and telecommunication use cases. Sakshi has done research work in the field of Machine Learning, NLP, Internet of Things.<\/span><\/p>\n

Connect with Sakshi on Linkedin.<\/a><\/span><\/p>\n

Table of Contents<\/h2>\n
    \n
  1. Churn Analytics: Telecommunication Industries<\/li>\n
  2. What is Churn Analytics?<\/span><\/li>\n
  3. Using Data Science and Analytics to Analyse the Churn<\/li>\n
  4. Extensive Exploratory Data Analysis<\/li>\n
  5. Dataset Discussion<\/span><\/li>\n
  6. Hands-on Python Notebook<\/span><\/li>\n
  7. Conclusion<\/li>\n<\/ol>\n

    Churn Analytics: Telecommunication Industries<\/span><\/h2>\n

    Customer retention is important for businesses to analyze their growth and effective working strategies. In this DataHour, Sakshi will discuss factors that constitute the degradation of business due to churn, especially in Telcos.<\/span><\/p>\n

    What is Churn Analytics?<\/span><\/h2>\n

    To understand in easy terms, suppose there was a boy named Rahul who used to buy groceries from retail stores. But nowadays, Rahul orders his groceries from online stores. Rahul shifted from retail stores to online stores. Hence, Rahul is the \u201cChurn\u201d for Retail shops. When a person who is using a service from company X suddenly stops using it and shifts to another service providing company Y, then that person is called Churn for company X. Reasons could be more benefits, better options, accessible customer care services, and much more. Churn analytics help to determine all these problems.<\/span><\/p>\n

    When a company faces high churn analytics<\/a> rates, then eventually revenue of the company decreases. It impacts the share market value of the company also. Due to this, a company loses its Brand value, which is the main reason for the layoffs.<\/span><\/p>\n

    In the past few years, you have noticed that millions of people shifted to some particular Telecommunication company because it was providing free data and calling services. Due to easy online application and doorstep delivery of sim, people frequently port their service provider for better experiences.<\/span><\/p>\n

    Hence, looking after your churn rate to improve your business strategy before going bankrupt is necessary.<\/span><\/p>\n

    Few companies where churn analytics is showing its vast impacts are Telecommunication, Gaming Industry, Local shops, restaurants, Banks<\/a>, Shopping Malls, etc.<\/span><\/p>\n

    Using Data Science and Analytics to Analyze the Churn<\/span><\/h2>\n

    Now we will understand this problem from a Data Science and analytics perspective.<\/span><\/p>\n

      \n
    1. Data Set Procurement and understanding:<\/strong> First, we need the complete Data in a digitized form containing all the features, which will help analyze churn. For this, we will use python code with a standard data set, which will help you to get an overview of how your chunk data should look, especially for the telecommunication sector.<\/span><\/li>\n
    2. Data in enrichment and preparation:<\/strong> In real-time scenarios, you always receive the data in a very messy form. So first, you need to enrich it and prepare it so we can understand it easily.<\/span><\/li>\n
    3. Exploratory Data Analysis<\/a>:<\/strong> Analysis means finding the hidden Trends within the data.<\/span><\/li>\n
    4. Handling Imbalance in Data-set:<\/strong> Often, we see an imbalance in the data set; it may occur due to the biased nature of a particular customer class or group.<\/span><\/li>\n
    5. Performing Modeling:<\/strong> We will use machine learning and deep learning modeling for better understanding.<\/span><\/li>\n
    6. Evaluating and Analyzing Results:<\/strong> At the end, we will evaluate our findings from the above processes.<\/span><\/li>\n<\/ol>\n

      Data-Set Description<\/b>: The data set we will use in this project is the \u201cIBM Telco Churn Dataset.\u201d It has 33(Independent Variables) that indicate the characteristics of clients of a Fictional Telecommunication company. The churn column (response variable) indicates whether the customer left in the last months. Class \u2018NO\u2019 indicates the clients who haven\u2019t left the company in the last months. Class \u2018YES\u2019 indicated the clients who left the company in the last few months.<\/span><\/p>\n

      Below is the python notebook that Sakshi has prepared for today\u2019s project. Here is Telco Churn excel; you can see 33 columns.<\/span><\/p>\n

      <\/p>\n

       <\/span>Below is the pic of all the columns by name. We will understand them separately. All these different parameters will help us to understand the churn.<\/span><\/p>\n

      \"project\"<\/p>\n

      Now the target column for us is the \u201cChurn Label,\u201d as shown in the screenshot below. It\u2019s either YES or NO, as described before. <\/span><\/p>\n

      \"Churn<\/p>\n

       <\/span>In the below pie chart, we can clearly see that the data set is biased toward \u201cYES.\u201d So, it\u2019s important to handle the class imbalance here. We are taking help from SMOTE here. SMOTE is a \u2018Synthetic Minority Oversampling Technique\u2019 that helps handle imbalanced datasets. So, from the original datasets, we have generated a few more samples.<\/span><\/p>\n

      \"Churn<\/p>\n

       Extensive Exploratory Data Analysis<\/h2>\n

      In the python notebook, the hidden Trends which are already there in the data will help take out the decision of what kind of machine learning model we need to apply in a later stage.<\/span><\/p>\n

      Below are the points that we are gonna look at in this project.<\/span><\/p>\n

      \"Churn<\/p>\n

      Now, we will start enriching the data to apply machine learning models. We have applied a combination of categorical and float columns; we need to make them in such a format that our machine learning algorithms can quickly process them. After pre-processing, we took out some statistics data like mean and Standard Deviation. In the last 2 columns, you can see \u2018CLTV\u2019 (customers retaining for longer times) and \u2018churn reason\u2019, which are very important. So on this text data, I have applied MP to find the trends making customers move to another company. We can also analyze the data of other companies\u2019 data and design our product so that customers won\u2019t leave in the future.<\/span><\/p>\n

      \"Churn<\/p>\n

       <\/span>Below is the correlation; you can clearly see that highly correlated variables are close to 1. For example, the \u2018Total Charges\u2019 is 0.93, indicating that if the service charges are pocket-friendly, then customers will remain for a more extended period. <\/span><\/p>\n

      \"Churn<\/p>\n

       <\/span>With the below code, we will study the distribution, which is very important in any machine-learning problem. It helps us to decide which algorithm we\u2019re going to use ahead.<\/span><\/p>\n

      \"Churn<\/p>\n

      Now, we need to segregate the medical and categorical values. Below you can clearly see the churning part in terms of gender distribution. We can conclude that churning in terms of gender is having no much difference.<\/span><\/p>\n

      \"Churn<\/p>\n

      In the chart below, we can see that customers who have taken plans for more time periods show more retention. For them, the churn rate is much lower.<\/span><\/p>\n

      \"project<\/p>\n

      Below is an exciting chart showing that as tenure increases, the chance of churn decreases.<\/span><\/p>\n

      \"Churn<\/p>\n

      Through all these graphs, we are trying to figure out which factors are responsible for more churn rates and which are not. So that we can work on the loopholes to decrease our churning; for example, we can tell telecom companies to provide more extended plans with discounts so that the churning rate will decrease. <\/span><\/p>\n

      For the last column, \u2018Churn Reason,\u2019 we will put all customer texts in the form of word clouds for which we have applied some part of NLP. You can see all the reasons that are contributing to churning.<\/span><\/p>\n

      \"Churn<\/p>\n

      The graph shows the relationship between churn score with zip code, latitude, longitude, tenure, etc.<\/span><\/p>\n

      <\/p>\n

      Results<\/b>: The sketchbook plot below shows the distribution of 0 and 1 labels on test data.<\/span><\/p>\n

      <\/p>\n

      Conclusion<\/span><\/h2>\n

      By applying analytics and understanding the insights from it, telco companies can work on improving their plans and reducing churners. Other Sectors like Hotels, Retail shops, and shopping malls can also use this method to stop their customers from being churners.<\/span><\/p>\n

      Below are our takeaways<\/strong> from the above analysis.<\/span><\/p>\n

        \n
      1. Electronic Check medium is the highest churner.<\/span><\/li>\n
      2. Monthly customers are more likely to churn because they can move without contract terms.<\/span><\/li>\n
      3. Customers who don\u2019t feel online security and get no Tech. Support the highest churn.<\/span><\/li>\n
      4. Non-senior citizens are the highest churners.<\/span><\/li>\n<\/ol>\n

        Here is another interesting article: Bank Customer Churn Prediction Using Machine Learning<\/a>.<\/p>\n

        The media shown in this article is not owned by Analytics Vidhya and is used from the presenter\u2019s presentation.<\/b><\/p>\n

        Related<\/em><\/h3>\n<\/p>\n