“Retaining customers is more difficult than attracting them”
This is a statement that any marketer can relate to. It is said that attracting new customers costs six times more than retaining existing customers. What data are companies analyzing to retain current customers?
‘Retention analysis’ and ‘funnel analysis’ are the most commonly used analytical methods to identify and retain customers at their point of origin. Retention rate is a metric that measures the number of users who return to use a service over a specific period of time. Improving retention has the effect of improving key metrics such as active users, loyal customers, and sales.
A channel is a model for the customer journey starting from user awareness of a service to purchase. The process of tracking where users leave at each stage and reflecting that in strategies that help users enter the next stage is called funnel analytics. Through funnel analysis, you can know at what stage customers churn and improve customer experience.
Retention and funnel analytics have been used very effectively in being able to analyze the points at which customers have stopped using the service. However, this method analyzes customers who have left , and from a marketer’s perspective, it can be said that this is not a highly effective method.
Best of all, customers who leave your service are more likely to switch to a competing service. There is no more painful moment for a marketer than losing a customer who has been through a difficult process. Rehiring them required more money and resources than expected.
If customer churn can be predicted, marketers have the opportunity to re-engage them by sending them a powerful hyper-personalized message. In other words, there is a golden window to market before the last customer churns.
In the battle for timing to market, predicting churn is a powerful weapon.
With the recent announcement of Google’s Privacy Sandbox , the importance of internal marketing based on a company’s first-party data is being emphasized even further. Many companies are adopting platforms like CDP and establishing AI organizations to execute on their data strategies and are conducting active research using data.
CDP DataS provides ‘churn prediction function’ through self-developed artificial intelligence algorithms to help marketers check future customer churn rates in advance. Now, anyone who makes good use of the platform can use predictive indicators based on customer behavioral data.
DataS bounce rate prediction screen
DataS’s churn prediction function predicts future churn using an LSTM (Long Short Term Memory) model, which uses customers’ past traces, such as access, purchase history, events performed and visit time, as the main variables. If only the last 90 days of customer data are accumulated in your CDP, you can immediately apply an algorithm to test the data to predict churn for each customer.
This predicted churn rate is updated automatically every 24 hours in a 360-degree customer profile that shows customer data at a glance, and marketers use this to create integrated marketing communications strategies. go more extreme, such as sending personalized messages to prevent customers from churning. You can do like this:
Predicting customer churn rates focuses on marketing usability
DataS’s bounce rate prediction model breaks user bounce rate into chunks so marketers can easily use the data in practice.
By providing user dropout rate classification into 10 parts through complex training, this provides convenience for marketers to present different hypotheses and actions to users in each section. This reflects the nature of marketers’ work, which involves constantly generating hypotheses and trying to verify them.
Instructions for using the bounce rate prediction function 🔎
When it comes to predicting bounce rate, simply checking the numbers is meaningless. You can only achieve meaningful results by building hyper-personalized marketing scenarios that target specific customers or customer groups in 10 detailed sections and take real actions. To achieve this, there must be the ability to link easily and conveniently with the marketing solutions that marketers use.
Determine’s bounce rate prediction functionality is aligned with the functions provided by Determine, showing high usability.
Create segments, Audience Studio
This is a solution to create user segments according to analytical conditions and create desired audiences by combining each segment. By enabling bounce rate prediction in Audience Studio, you can create and use the following segments.
- High-risk group with predicted churn rate above 80% among app customers in January
- Risk group with a predicted churn rate of 50% or more among customers who make three or more purchases in 2021
- High-risk group with a predicted churn rate of over 70% among customers purchasing in February
- High-risk group with a predicted churn rate above 85% among customers who viewed details of a particular product in January and February but did not purchase it
Personalized messages, growing actions
This is a solution that generates marketing hypotheses and automatically sends personalized messages through various channels and devices, text and email according to user behavior and analyzes effectiveness to verify effectiveness. effectiveness of marketing activities. Different messages can be constructed depending on the predicted likelihood of churn.
While many marketers are pursuing marketing strategies in a reactive manner by looking at analytical metrics, there is a shift in marketing strategies towards a proactive approach through the introduction of Data infrastructure such as CDP, has recently received attention and the application of artificial intelligence algorithms.
Through the bounce rate prediction function, marketers can protect the golden time to market, retain hard-to-attract customers, and turn them into loyal customers. In addition to the bounce rate prediction function, DataS plans to introduce artificial intelligence algorithms to address more marketers’ concerns, such as predicting purchase quantity and predicting purchase probability a specific product.