AI for customer analytics
Achieve better conversion rates,
lower marketing costs and increase profits
Check our new offer for Data Science Development Path!
Understanding customer lifecycle
with machine learning
Only 1 out of 26 unhappy customers actually complains. The rest just go away without giving any sign they’re displeased. Even companies that have embraced the power of big data may benefit from further automating the data processing done by data analysts.
Applying AI models to predict customer lifetime value, enhancing cross-selling and fraud detection may increase a company’s performance and help it gain competitive advantage.
Our extensive training track gives your team the know-how it will need to build advanced, state-of-the-practice machine learning solutions for customer analytics.
Who it is for
The track is designed for teams of data analysts who work regularly with data sets and seek a hands-on approach for common areas of business.
The examples and projects used during the track are run in Python or R, so knowing at least the basics of the syntax is recommended. We can also provide additional training in the Python or R programming environment the team prefers to use for machine learning. No previous experience in machine learning is required.
Program
Our training program contains six day-long units divided into two levels. Depending on your focus, you either choose specific areas that interest you or go through all the units.
Know your customer 2.0
Customer View
Customer
Segmentation
Customer
Lifetime Value
Proactive customer engagement
Cross-selling
& up-selling
Fraud
Detection
Churn
Prediction
Interested?
Let us know which training units you need to help your team make the most of customer analytics.
Program units details
Customer view 360°
Overview
Customer 360° offers access to all of the information that can be gathered about clients. The aim is to integrate knowledge on customer behavior across any channel, system, device or product.
Having access to the full range of the customer’s journey makes it possible to increase desirable patterns of behavior, reduce customer churn and facilitate personalized marketing offers.
Your team will learn
- How to drive successful project of holistic customer view
- What are common challenges and pitfalls in customer data preparation
- How can customer view 360° help you meet your business challenges
Related workshops
Customer segmentation, Churn prediction
Detailed agenda (1 day)
Part 1: Business requirements of Customer view 360°
- Business vs. customer perspective.
- Key challenges in driving a customer data model.
- Keep up with the client.
Part 2: Data preparation
- Transactional data, website logs, call-center activity, location – combining datasets to improve customer characteristics.
- Data inconsistency – missing data handling.
- Customer profiles – description statistics and data visualization.
Part 3: Hands-on coding: Customer view 360° development
- Iterative process of adding new data sources.
- Real-time perspective.
- Data volume management.
Customer segmentation
Overview
Customer segmentation helps you understand different types of customer behavior and quantify how many of each you’re dealing with. Marketing automation based on customer treatment according to their behavior or product usage is the best way to put forward appropriate offers.
Machine learning delivers algorithms to generate customer clusters based on transactions or other activity. The clusters enable a straightforward definition of different customer types, which can then be addressed with well-targeted marketing actions.
Your team will learn
- How to prepare transactional data to drive successful customer segmentation
- How to perform segmentation using machine learning techniques
- How to transfer results to business teams
Related workshops
Customer view 360°, Cross-selling & up-selling
Detailed agenda (1 day)
Part 1: Business needs and expectations in customer segmentation
- Determining future segmentation usage, main goals and needs.
- Types of segmentations: (lifetime, lifecycle, behavioral, RFV recency-frequency-monetary value).
- Utilizing the results of cluster characteristics.
Part 2: Data preparation
- Feature engineering techniques.
- Working with a dataset: feature scaling and transformation.
- Data exploration, hypothesis testing.
Part 3: Hands-on coding: Modeling customer segments
- Reducing dimensionality (PCA, t-SNE).
- Clustering: K-means, finding the optimal number of clusters.
- Evaluation clustering – model assignment stability in time.
Customer Lifetime Value
Overview
Customer Lifetime Value (CLV) is employed to determine how to invest time, money and resources for your most valuable customers or whether customers are profitable. It measures all of the profits a customer will generate for a company.
CLV helps determine the amount that should reasonably be invested in each stage of the customer life cycle as well as which customer segment should be focused upon to maximize profit. It also helps you define customer churn.
Your team will learn
- How to collect data to drive a successful CLV project in your company
- How to train the best Customer Lifetime Value model
- How to apply results to boost the revenue your customers generate
Related workshops
Customer segmentation, Cross-selling & up-selling, Churn prediction
Detailed agenda (1 day)
Part 1: Business context of Customer Lifetime Value methodology
- Diagnosing model prerequisites: contractual or non-contractual types of customer relations and purchase frequency.
- Identifying business areas capturing customer definition (revenue and cost streams).
- Maximizing return over the customer’s lifecycle.
Part 2: Data preparation
- Data integration (verifying the usefulness of customer behavior modeling, cleaning and organizing data).
- Feature engineering techniques for valuable algorithm inputs.
- Working with a dataset: splitting the dataset, normalizing data.
- Data exploration, measures of success.
Part 3: Hands-on coding: Modeling Customer Lifetime Value
- Building a CLV model.
- Tuning your solution to optimize results.
- Verifying a CLV model with customer characteristics.
- CLV reporting – what’s most important?
- Applying the results: future evaluation for existing and potential customers, identifying potential churn segment customers.
The propensity to buy: cross-selling & up-selling
Overview
Boosting customer engagement can be achieved by offering new products for customers, which can be reached by proposing higher-priced products (up-selling) or complementary ones (cross-selling).
Machine learning techniques support the automation of offers prepared for customers. This may include planning the optimal size of target groups, selecting the most interested customers only or proposing the optimal channel for marketing communications.
Your team will learn
- How to agree on common goals with marketing
- How to define data-driven marketing campaign target groups
- How to develop good predictive models
Related workshops
Customer segmentation, Churn prediction
Detailed agenda (1 day)
Part 1: How analytics supports business needs
- Determining common goals with marketing teams.
- Applying predictive modelling in direct marketing.
- Marketing automation support.
- Up-selling data-driven strategy.
Part 2: Data preparation
- Defining target variables.
- Feature engineering techniques into valuable algorithm inputs.
- Working with a dataset: splitting the dataset and normalizing the data.
- Data exploration, measures of success
Part 3: Hands-on coding: Modeling the propensity to buy
- Working with a dataset: data exploration, measures of success.
- Baseline prediction: linear regression, logistic regression.
- Decision trees for the propensity to buy.
- Boosting model results: decision trees, XGBoost, LightGBM, ensemble.
- Applying the results: supporting telemarketing sales.
Fraud detection
Overview
Customer fraud usually boils down to methods used to deceive. Identifying suspicious operations takes time, labor and no small number of mistakes. Regardless of how it is carried out, however, fraud can be costly.
Machine learning algorithms can be used to both improve the detection of fraudulent behavior and obtain fewer false alarms.
Your team will learn
- How to define fraudulent behavior
- How to analyze data to obtain different types of fraud
- How to apply machine learning techniques to improve the precision of anomalous behavior detection
Related workshops
Detailed agenda (1 day)
Part 1: Business aspects of fraud
- Defining fraud: anomalous behavior, type of frauds.
- Working with a dataset: identifying fraud cycles.
- The challenges fraud presents.
Part 2: Data preprocessing for fraud detection
- Data preprocessing (missing value handling, inconsistency identification).
- Description statistics.
- Data transformation, setting metrics, building rankings.
Part 3: Hands-on coding: Fraud detection
- Detecting anomalies.
- Clustering algorithms among groups of data.
- Classification methods for identifying and uncovering patterns.
Churn prediction
Overview
Customer churn prediction tells us how likely a customer is to discontinue a contract or stop using a service. Identifying such customers and preventing attrition is a challenging task simple business heuristics often fall short on.
During the workshop we lay out a complete path–from setting the assumptions of the definition of churn to be employed to data wrangling and modelling.
Your team will learn
- How to define churn
- How to prepare features for churn modeling
- How to get a baseline model and boost results
Related workshops
Customer segmentation, Customer Lifetime Value
Detailed agenda (1 day)
Part 1: Business aspects of defining churn
- Not so simple – defining types of churn: quantity or quality churn/partial or total churn.
- Working with a dataset: setting the frequency of interaction.
- Customer personalization.
Part 2: Feature engineering for churn modeling
- Data exploration and preprocessing.
- Feature engineering techniques into valuable algorithm inputs.
- Working with a dataset: splitting the dataset, normalizing data.
Part 3: Hands-on coding: Churn modeling
- Baseline model: Logistic regression.
- Decision trees for churn prediction.
- Boosting results: XGBoost, LightGBM.
Let’s discuss your needs
Leave us a message and we will contact you shortly.
Ready to train your team?
LEARN
how to apply AI to your solutions
START
building machine learning skills
ACCELERATE
your data analytics with Apache Spark