Case Study


Churn prediction

Business challenge: The whole service sector is facing the challenge of customer lifetime value (CLV). Users are constantly exposed to offers and actions of competitors who want to persuade them to change service providers (telco, cable TV and Internet providers are the best examples). Again, there are many products that can meet the same need, therefore there is a refusal of using a particular product, e.g. credit card (banks and insurance are good examples of this). The challenge for providers is to accurately identify those clients who are inclined to shorten their „stay“, in any unwanted sense.
Solution: It is possible to apply advanced algorithms to the historical data of users, owned by the service or product provider, in order to build a model that can be used to express the tendency of some of the current users to leave in the next period.
Benefits: When we make a prediction of those who are prone to “churn”, the company can implement measures to prevent the leaving of its customers. These measures would be strictly used for those customers who show such tendencies, so that resources and time would not be wasted, but would be precisely spent only on those who want to leave.


Next Best Offer

Business challenge: You invest money, effort, and time to create an offer for your customer, but the customer eventually tells you that he is not interested after all. You call the customer again after a while, with a new promotion and he rejects you again with the sentence “where were you a month ago when I bought a competitive brand?”. How do we know WHAT, WHEN, and HOW to offer our customers, whether it is a basic or an additional offer?
Solution: A data-based approach that allows us to create a unique approach to each of the customers, instead of general campaigns, so that the customer feels valuable. Customers who feel valuable become loyal and they can even become brand ambassadors.
Benefits: Increased efficiency of the marketing budget, customers who feel valuable, reduction of the number of missed opportunities.


User profiling

Business challenge: Today’s customer is more and more demanding! Customers expect brands to always be present! The messages they receive are expected to be in line with their desires, needs, lifestyle or priorities. Customers are no longer a homogeneous segment formed based on of demographics or consumer habits, but a more complex system of relationships that needs to be understood, in order to become adapted to that system. How to get to know our customer better?
Solution: What if we told you that now, with the help of predictive analytics, we can create personalized campaigns for each customer to suit their desires and needs. Companies already have all of the information in their data, they just need to link them in the right way so they can use them.
Benefits: Saving time and money, reducing the rate of missed opportunities, reaching the audience with the best individual message, increasing customer satisfaction



Business challenge: No matter how many users or clients we have, each of them deserves attention, in a unique way and to a deserved extent. However, we need to optimize the approach and treat all clients through a limited and balanced number of profiles (segments) that will represent their overall diversity and whose total number will provide space for efficient communication with limited costs.
Solution: By applying advanced algorithms on both demographic and behavioral (user) data, we clearly see the profiles that stand out and at the same time fully cover the entire client portfolio.

Benefits: In this way, we gain insight not only into which clients need a different approach, but also into which clients deserve more attention in general, because they have greater strategic importance. This approach involves not only the development of a segmentation model based on integrated data that gives a complete picture, but also the whole mechanism of simple scoring of users (both new and old), each month, in terms of which segment they will belong to. Each client transition from one segment to another stands out separately. It is important to point out that each client has a segment score, no matter how little information about him we have at the time.


Fraud detection

Business challenge: Not every potential customer is equally desirable in our system, there are those who intentionally or unintentionally do harm to their providers. How to recognize who is the one we don’t want?

Solution: Based on demographics and behavioral data, we can create a model that precisely predicts the clients who are at risk of intentionally or unintentionally becoming part of what we call fraud.

Benefits: Reduction of costs incurred because of losses due to fraud, but without jeopardizing the level of acquisition.


Anomaly detection

Business challenge: When an unwanted event occurs, we are trying to save as much as possible. This process is not simple at all, because it is necessary to detect the anomaly and eliminate it efficiently in the sea of information that we have.

Solution: The anomaly detection system implies a pre-prepared solution for detection and prevention so that an unwanted event does not occur. To detect anomalies, we use prediction methods, scenario analysis, multivariate analysis, pattern recognition in data series.

Benefits: Reduction of the number of unwanted events, savings by early detection of harmful events, savings by system optimization.


Collection process optimization

Business challenge: Unfortunately, many clients do not settle their contractual obligations and because of that, companies are often forced to enter into long and expensive debt collection procedures, which have many steps. A common practice is that the billing process works according to the same scenario for all clients, which is certainly not the most optimal because it generates unnecessary costs for those who do not aim to avoid payment, but are late for some reason, and also prolongs the time for those who really want to avoid payment.

Solution: The solution is based on creating groups of defaulters, according to past data, with clearly defined scenarios for each group of defaulters.

Benefits: Shorter time and lower billing costs, while avoiding creating additional dissatisfaction.


Prediction of employee turnover

Business challenge: Employee turnover is a problem that we often encounter in modern business. This phenomenon usually causes a large cost to the company because it is necessary to cover the costs of the employee’s departure, the overtime hours of someone who took over his job, the costs of recruiting a new employee, his training, and many other things. On average, a company must invest between four weeks and three months to train and introduce a new employee to the business. It often happens that the same new employee leaves the company in the first year, which represents an additional cost. The situation is even more delicate when it comes to consulting companies and smaller companies, where there is a deterioration in client satisfaction due to frequent changes of consultants or key account managers, with whom the client is used to working.

SolutionBased on internal data and specialized questionnaires, we create a model that predicts the probability for each employee to change the company.

Benefits: This solution helps the company to predict employee turnover and take proactive measures to prevent employee departures, therefore, saving company funds.