Reducing Churn R ate and Increasing ROI of Customer Loyalty: Russian e-commerce giant Ozon.ru shares its experience of successfully using Machine Learning technologies in online marketing
The R Churn ate in or outflow rate, is considered one of the most important marketing metric that shows the viability of the business model of retaining loyal customers . Companies strive to keep it to a minimum, increasing the return on stimulating repeat calls, i.e. ROI is a return of investment, since attracting a new buyer costs 5 times more than returning an old one. And the probability of selling goods and services to an old client is 60-70%, while a new one is only 5-20% .
What Is Preventive Online Marketing
To make a potential sale real, you need to be clear about the customer’s needs. For example, if a user sets the search query “olive oil”, online stores, first of all, offer him products of this particular category, and not toys, electronics or household appliances.
This is the traditional approach to marketing – the formation of a sales proposal based on demand. An innovative strategy involves predicting and stimulating customer needs.
For example, when a customer bought olive oil last month, they are most likely to need it again in 30 days and repeat the purchase. Therefore, it is advisable to pre-form and send the buyer an appropriate sales offer through a personal appeal in the personal account, push notification on the website of the online store,
Such a method of maintaining repeated requests actively using online store Ozon.ru, applying machine learning technology ( Machine the Learning ) to generate preventive promotional offers based on the accumulated information about the user’s behavior . This refers to the field of predictive or predictive analytics of big data ( Big Data ) . Let’s look at a schematic sequence of steps for this strategy.
Stages Of Building Predictive Advertising Offers Based On Machine Learning
- Collecting information about the behavior of a site visitor: category and frequency of purchases, region of residence, channels of calls and about 300 more parameters.
- Creation of an intelligent machine learning model that trains on historical data and predicts the likelihood of a future purchase that every site visitor who comes to Ozon in the last 3 months will make a future purchase.
- Sorting users by the calculated value of the likelihood indicator and their distribution into segments of solvency. Ozon distinguishes 20 segments: the most paying customers are in the 20th segment, and the customers who bring the store’s average revenue per visit are in segments 10-12.
- For each of the selected segments, its own advertising model is built, which is implemented through contextual advertising systems: Google, Yandex and other sites.
Results Of Applying Machine Learning In Online Marketing For Predictive Demand
As a result of the implementation of the described approach, there is a negative churn rate and a significant increase in sales through the search advertising channel, as well as the share of buyers purchasing goods of different categories through cross-selling.
Predictive analytics made it possible to personalize the site, i.e. for each visitor, its own unique page is generated, depending on his interests, as well as a notification (email, push message or banner) in the form most promising for this client. Such a detailed classification of buyers allowed to increase ROI from referral to a client by more than 30-50% . This, in turn, improved another important metric for business – LTV. (Lifetime Value) – the total profit of the company, which it receives from one client for the entire period of work with him.
This is just one of many successful examples of the implementation of Big Data and Machine Learning technologies in the retail and e-commerce sector. Learn how to select, customize, and use specific techniques, tools, and analytics to address these and other marketing challenges in our hands-on courses for Data Scientists – Big Data Analysts and Machine Learning Professionals.