Retention-Driven Marketing for Music Apps Datasets

Posted on Nov 10, 2020

Photo by C D-X on Unsplash

Github Repository | LinkedIn: Rob Davis, James Welch, Sita Thomas


For this project we were tasked with designing a marketing strategy for KKBox, a streaming music service. We were given four datasets describing user demographics, transaction history, listening history, and churn rate. This project explores which users are the most important to target, how churn impacts revenue, two potential marketing plans, the differences between user segments, and future improvements to the data set.

Key Users

The datasets above shows that the vast majority of users are on 30-day plans, and the vast majority of those users have auto-renew. This group accounts for nearly 96% of all users we had data for, therefore, we isolated our research to only 30-day auto-renew users.

We further isolated the users by price of plan, shown on the graph below - 98 percent of users have one of 5 pricing options out of many more. Two of these plans are NT$99 and NT$100, so we categorized them as low-tier plans. One plan is NT$129, so it is our mid-tier plan. The last two plans are NT$149 and NT$180, so they make up the high-tier plans.

Marketing Strategies

The plot below shows the amount of churn among users in each plan. Our low-tier plans have the least amount of churn while our high-tier plans have the most churn from the datasets. However, both the number of customers in a plan and the amount of churn for that plan influence the monetary value of that plan. Because of this, we have suggested two marketing strategies - one that minimizes churn, and one that maximizes revenue.

Churn appears to be directly related to plan price given the information from this dataset, so our recommendation to reduce churn is to market low tier plans, since churn increases as plan price increases.

In order to maximize revenue, taking into consideration the amount of churn per plan, which is different for each of the five plans, we looked at the annual value of a single subscription for each of our five plans, shown on the graph above. For example, on the $99 plan, a single subscription would be worth NT$99 the first month. The next month, however, due to the user churn for that plan price, that single NT$99 subscription would only be worth NT$97. Ultimately, single subscriptions of high tier plans net $346 more annually on average than low tier plans, despite the variations in user quantity and in the amount of churn per plan.

It should be noted that to calculate the annual revenue of each plan, we made two assumptions: that a single subscription would remain active for twelve months continuously and that the subscription would be for the same price plan all twelve months. However, there are many more factors that go into the true value of a subscription plan, which we do not have access to, so we were unable to consider those variables for this project. As a result, this plan value comparison should be considered a starting point only, with further data collection and analysis required to make a confident and accurate conclusion.

User Segmentation Patterns

After grouping pricing plans into tiers and identifying churn patterns, we looked for patterns that would indicate user segmentation. Two strong indicators are registration method (shown above) and payment method (shown below). This dataset uses numbers as stand-in values for categories, so we cannot be specific in this project about which registration method is from the app and which is registration from a referral link, for example, but we can still explore patterns.

Users in low tier plans predominantly use registration method 7 and payment method 41. High tier users have a larger variety of both registration and payment methods. Interestingly, the NT$140 plan and NT$180 plan payment methods are exclusive to each other. We do not know if these patterns are simply artifacts of method availability at a certain point in time, an intentional segmentation by KKBox, or a reflection of user self-segmentation. With this information, we could leverage registration method to expand customer reach and payment method to drive higher user retention.

Plan price is very weakly correlated with the limited engagement metrics this dataset has, such as the number of skipped songs in a listening session or the amount of time listened. However, existing literature on streaming music service users asserts that how people listen and engage with the app is extremely important to understanding segmentation for customer retention and reach, as well as important for ad buyer confidence that their messaging is reaching desired segments. The literature outlined seven user segments, while Spotify created three metrics to categorize their users.

Unfortunately, using these metrics to perform segmentation for KKBox requires far more information than we have, so we cannot confidently provide a data-driven marketing strategy based on user engagement. Therefore, we have compiled a list of dataset improvements to generate the information necessary for making accurate assessments and suggestions specifically about user listening behavior.

Future Work

We identified several areas that would help us better understand our customer segments. We want to know more about listening context - having timestamps in addition to date stamps tells us if and how frequently users are engaging with the app during a morning commute or at a late-night party, for example. Understanding other user behaviors is also key in creating accurate segments, including likes and dislikes, artist/user follows, shares, downloads, and playlist creation and editing behavior, as well as information about song discovery method - when a user accesses a song via personal playlist, a KKBox-curated or other user’s playlist, an album, an artist page, app radio, search, or a direct link.

Information about user devices is also useful, like the type and number of devices used, device’s frequency of use and the datetimes it is used. Lastly, understanding user music taste is a major part of segmentation, which includes song categorization metrics such as genre, mood, instruments, and activity (studying or exercising, for example) (Ha Lee, 2015). All these variables would provide tremendously valuable insights for marketing campaigns, reducing customer churn, improving customer experiences, and increasing customer referrals, which all lead to greater revenues and enhanced company focus and efficiency.


In this project, we isolated key users, explored the effect of churn on revenue to produce marketing approaches for low and high tier plan users, considered user segmentation, and developed an expanded data collection plan for more confidently and accurately generating strategic insights.


Ha Lee, J., Price, R. (2015) Understanding Users of Commercial Music Services Through Personas: Design Implications. Proceedings of the 16th ISMIR Conference, 476 - 482

Understanding people through music. (2017, July 13). Retrieved October 30, 2020, from

The skills I demoed here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.


About Authors

Sita Thomas

Knowledge is power. I leverage fierce curiosity and creativity to deliver immediate and impactful results for mission-driven companies by weaving together 20 years of statistics, engineering, and business development. I've worn many hats across a variety of industries,...
View all posts by Sita Thomas >

James Welch

I was trained as a synthetic biologist and I am working to become a data scientists too. I have expertise in the genetic engineering of a variety of single-celled organisms, DNA and protein design, and industrial process scaling...
View all posts by James Welch >

Related Articles

Leave a Comment

No comments found.

View Posts by Categories

Our Recent Popular Posts

View Posts by Tags

#python #trainwithnycdsa 2019 2020 Revenue 3-points agriculture air quality airbnb airline alcohol Alex Baransky algorithm alumni Alumni Interview Alumni Reviews Alumni Spotlight alumni story Alumnus ames dataset ames housing dataset apartment rent API Application artist aws bank loans beautiful soup Best Bootcamp Best Data Science 2019 Best Data Science Bootcamp Best Data Science Bootcamp 2020 Best Ranked Big Data Book Launch Book-Signing bootcamp Bootcamp Alumni Bootcamp Prep boston safety Bundles cake recipe California Cancer Research capstone car price Career Career Day citibike classic cars classpass clustering Coding Course Demo Course Report covid 19 credit credit card crime frequency crops D3.js data data analysis Data Analyst data analytics data for tripadvisor reviews data science Data Science Academy Data Science Bootcamp Data science jobs Data Science Reviews Data Scientist Data Scientist Jobs data visualization database Deep Learning Demo Day Discount disney dplyr drug data e-commerce economy employee employee burnout employer networking environment feature engineering Finance Financial Data Science fitness studio Flask flight delay gbm Get Hired ggplot2 googleVis H20 Hadoop hallmark holiday movie happiness healthcare frauds higgs boson Hiring hiring partner events Hiring Partners hotels housing housing data housing predictions housing price hy-vee Income Industry Experts Injuries Instructor Blog Instructor Interview insurance italki Job Job Placement Jobs Jon Krohn JP Morgan Chase Kaggle Kickstarter las vegas airport lasso regression Lead Data Scienctist Lead Data Scientist leaflet league linear regression Logistic Regression machine learning Maps market matplotlib Medical Research Meet the team meetup methal health miami beach movie music Napoli NBA netflix Networking neural network Neural networks New Courses NHL nlp NYC NYC Data Science nyc data science academy NYC Open Data nyc property NYCDSA NYCDSA Alumni Online Online Bootcamp Online Training Open Data painter pandas Part-time performance phoenix pollutants Portfolio Development precision measurement prediction Prework Programming public safety PwC python Python Data Analysis python machine learning python scrapy python web scraping python webscraping Python Workshop R R Data Analysis R language R Programming R Shiny r studio R Visualization R Workshop R-bloggers random forest Ranking recommendation recommendation system regression Remote remote data science bootcamp Scrapy scrapy visualization seaborn seafood type Selenium sentiment analysis sentiment classification Shiny Shiny Dashboard Spark Special Special Summer Sports statistics streaming Student Interview Student Showcase SVM Switchup Tableau teachers team team performance TensorFlow Testimonial tf-idf Top Data Science Bootcamp Top manufacturing companies Transfers tweets twitter videos visualization wallstreet wallstreetbets web scraping Weekend Course What to expect whiskey whiskeyadvocate wildfire word cloud word2vec XGBoost yelp youtube trending ZORI