Churn modelling: Predict customers at risk of leaving

If your business relies on a subscription model, knowing which customers are likely to cancel can provide you with an important competitive advantage. It is a lot more expensive to acquire a new customer than to keep an existing one. Our predictive churn models deliver high accuracy across industries including energy, telecom, insurance, and fitness.

Client type: insurance, energy, fitness
Technology: Python, SQL, Gradient Boosting, Azure

Recommender systems: Predict products a consumer is likely to buy

When offering a number of different products, a recommender system that predicts which product a specific customer is likely to buy at a certain point in time, is very valuable. Our models can allow you to make your commercial efforts more effective by knowing in advance what and when to offer to your clients.

Client type: insurance
Technology: Python

Customer segmentation: Personalize the customer experience

Go beyond demographics using machine learning to cluster your customers based on hidden patterns, leading to new revenue opportunities and the ability to personalize offerings to customer behaviour and needs.

Client type: insurance, banking
Technology: Python, R, unsupervised learning

Demand prediction: Know what people will want

An efficient supply chain is one that knows what customers will want next. Using internal and external data and advanced predictive models, we can help you reduce uncertainty, cost, and risk while improving the customer experience.

Client type: logistics, retail
Technology: Python, AutoML, XGboost

Sales conversion: Predict promising leads

Enrich your lead database with open data and machine learning to classify prospects by propensity to buy. This helps your sales team spend less time dialling and more time selling, greatly increasing your ROI.

Client type: call centre, sales
Technology: Python, Google BigQuery, Dataflow, GXBoost, Plotly Dash

Employer healthcare management: Illness and injury prediction

Our models score newly reported cases of illness and injury to estimate which will lead to long-term conditions and higher costs. This improves communication and treatment, helping improve care and reduce cost and risk.

Health insurance fraud detection: Stop false claims before you incur costs

Health insurance fraud is expensive and creates regulatory risk. We built sophisticated models to identify anomalies and present the results on an easy-to-understand dashboard, empowering a large insurer to act on potentially fraudulent claims early.

Client type: insurance
Technology: R, Benford’s law, Isolation Forest

Explainable fraud detection: Understand your fraud detection models

The use of fraud detection algorithms by fraud investigators is heavily dependent on the trust they have in the outcome of the model. Using sophisticated explainable AI techniques (e.g. post-hoc methods), combined with meta learning models, we are able to translate the outcome of your model into relevant business insights.

Client type: health insurance
Technology: Python

Message classification: Automate your customer-service workflow

Directing the growing volume of messages your company receives can be a drain on valuable time and resources. Deliver a great customer experience while freeing your staff to focus on value-added tasks with automated message intent analysis. Using natural language processing, our solution can understand and route messages appropriately.

Client type: insurance, legal
Technology: Python, NLP, Dimension Reduction, Decision Tree

Speech-to-text: Understand your call-centre conversations

We use speech-to-text technology to convert call-centre interactions into text files which can then be analysed using natural language processing, providing detailed insights about the customer needs and how well your agents are handling customer interactions. This can lead to improved agent training and customer-handling procedures.

Client type: insurance, call centre
Technology: Python, Google, tf-idf, Dimension Reduction, SAS

Sentiment analysis: Understand how customers really feel

Are your customers angry, frustrated, or happy? Emotions can be early warning signals of issues or new opportunities. We use natural language processing across CRM, call centre, and social media data to help you understand how customers are feeling and respond early to burgeoning problems.

Client type: insurance, call centre
Technology: Python, Google, tf-idf, SAS

Topic Detection: Make knowledge useful

Massive amounts of textual data make it difficult for people to find information. We train algorithms to do it for you, categorizing information and surfacing topic areas to make your organizational knowledge truly useful.

Client type: insurance, legal
Technology: Python, tf-idf, Dimension Reduction

Chatbots: Scale excellent service

Most customer service questions are repetitive and can be answered by chatbots, freeing your agents to focus on more complex cases. We work with your domain experts and identify the most appropriate technology for your needs, helping you bring your customer service to the next level.

Client type: insurance, banking, logistics, health, housing corporations, call centre
Technology: Python, Google, IBM Watson, Luis Azure, BERT (transfer learning)

Caregiver allocation: Long-term care predictive models

Rapidly aging populations and limited caregiver resources place constraints on healthcare systems around the world. Using data engineering and enrichment, along with predictive analytics techniques, we can help organizations plan ahead to allocate resources efficiently.

Client type: health care
Technology: SQL, Azure, Python, Regression

Speech-to-text: Understand your call-centre conversations

We use speech-to-text technology to convert call-centre interactions into text files which can then be analysed using natural language processing, providing detailed insights about the customer needs and how well your agents are handling customer interactions. This can lead to improved agent training and customer-handling procedures.

Client type: insurance, call centre
Technology: Python, Google, tf-idf, Dimension Reduction, SAS

Claims prediction: Improve capital allocation

High and variable costs for bodily injury claims make reserving difficult for casualty insurers. We used machine learning to predict the size of individual claims and enable a large auto insurance provider to improve financial results.

Client type: insurance
Technology: R, Shiny, Decision Trees, Boosting, GLM

Chatbots: Scale excellent service

Most customer service questions are repetitive and can be answered by chatbots, freeing your agents to focus on more complex cases. We work with your domain experts and identify the most appropriate technology for your needs, helping you bring your customer service to the next level.

Client type: insurance, banking, logistics, health, housing corporations, call centre
Technology: Python, Google, IBM Watson, Luis Azure, BERT (transfer learning)

Caregiver allocation: Long-term care predictive models

Rapidly aging populations and limited caregiver resources place constraints on healthcare systems around the world. Using data engineering and enrichment, along with predictive analytics techniques, we can help organizations plan ahead to allocate resources efficiently.

Client type: health care
Technology: SQL, Azure, Python, Regression

Health insurance fraud detection: Stop false claims before you incur costs

Health insurance fraud is expensive and creates regulatory risk. We built sophisticated models to identify anomalies and present the results on an easy-to-understand dashboard, empowering a large insurer to act on potentially fraudulent claims early.

Client type: insurance
Technology: R, Benford’s law, Isolation Forest

Customer segmentation: Personalize the customer experience

Go beyond demographics using machine learning to cluster your customers based on hidden patterns, leading to new revenue opportunities and the ability to personalize offerings to customer behaviour and needs.

Client type: insurance, banking
Technology: Python, R, unsupervised learning

Sales conversion: Predict promising leads

Enrich your lead database with open data and machine learning to classify prospects by propensity to buy. This helps your sales team spend less time dialling and more time selling, greatly increasing your ROI.

Client type: call centre, sales
Technology: Python, Google BigQuery, Dataflow, GXBoost, Plotly Dash

Demand prediction: Know what people will want

An efficient supply chain is one that knows what customers will want next. Using internal and external data and advanced predictive models, we can help you reduce uncertainty, cost, and risk while improving the customer experience.

Client type: logistics, retail
Technology: Python, AutoML, XGboost

Claims prediction: Improve capital allocation

High and variable costs for bodily injury claims make reserving difficult for casualty insurers. We used machine learning to predict the size of individual claims and enable a large auto insurance provider to improve financial results.

Client type: insurance
Technology: R, Shiny, Decision Trees, Boosting, GLM

Claims handling: Automate assessments

Property damage claims today typically require personal visits from assessors and long wait times for the customer. Using deep learning, these requests can be automatically interpreted using image recognition and other AI technologies to speed solutions for your customers, reduce costs for your business, and improve your risk profile.

Client type: insurance, housing corporations
Technology: Python, Google

Sentiment analysis: Understand how customers really feel

Are your customers angry, frustrated, or happy? Emotions can be early warning signals of issues or new opportunities. We use natural language processing across CRM, call centre, and social media data to help you understand how customers are feeling and respond early to burgeoning problems.

Client type: insurance, call centre
Technology: Python, Google, tf-idf, SAS

Chatbots: Scale excellent service

Most customer service questions are repetitive and can be answered by chatbots, freeing your agents to focus on more complex cases. We work with your domain experts and identify the most appropriate technology for your needs, helping you bring your customer service to the next level.

Client type: insurance, banking, logistics, health, housing corporations, call centre
Technology: Python, Google, IBM Watson, Luis Azure, BERT (transfer learning)

Underwriting: Improve risk selection

Improve the speed and accuracy of your underwriting decisions, freeing human reviewers to focus on more complex cases and reducing risk. We help you find the relevant data sources and combine them with your internal data to build superior underwriting models.

Client type: insurance
Technology: R, Azure, Power BI, web scraping

Pricing & Underwriting: Find relevant information by webscraping

Webscraping can be an effective tool for finding public information that is not captured in external data sources. Finding and collecting this information can bring a competitive advantage in insurance pricing and underwriting.

Client type: insurance
Technology: R, Python, web scraping

Underwriting: Detecting reed roofs using image recognition

Using image recognition on aerial photographs in combination with external data sources, enables automatic detection of roof characteristics of buildings. E.g. the detection of fire prone materials, like reed, or the use of solar panels.

Client type: insurance
Technology: Python, Azure

Claims prediction: Improve capital allocation

High and variable costs for bodily injury claims make reserving difficult for casualty insurers. We used machine learning to predict the size of individual claims and enable a large auto insurance provider to improve financial results.

Client type: insurance
Technology: R, Shiny, Decision Trees, Boosting, GLM

Employer healthcare management: Illness and injury prediction

Our models score newly reported cases of illness and injury to estimate which will lead to long-term conditions and higher costs. This improves communication and treatment, helping improve care and reduce cost and risk.

Client type: insurance, benefits
Technology: R, Azure, Regression

Retail trends: Discover insights in your point-of-sale data

Retailers can use product, category, and store benchmarks to understand and address sales performance. For a large retailer, we analysed almost 900 million cash-register transactions using algorithmic clustering that grouped similar stores together automatically. Anomaly detection then surfaced revenue development issues the company could take action to address.

Client type: retail, sales
Technology: R, unsupervised learning, anomaly detection

Demand prediction: Know what people will want

An efficient supply chain is one that knows what customers will want next. Using internal and external data and advanced predictive models, we can help you reduce uncertainty, cost, and risk while improving the customer experience.

Client type: logistics, retail
Technology: Python, AutoML, XGboost

Data quality assessment: Make your data ready to analyze

For data driven decision making or training AI models, high quality data is essential. With our domain knowledge we are able to assess which of your data are key data elements and what quality requirements and tolerances are acceptable. We can provide you with an interactive data quality dashboard that shows you all relevant insights.

Client type: insurance, pension funds
Technology: Python, Azure, Power BI

Algorithm validation: Challenge your modelling approach

To make sure that your machine learning models are state-of-the art, well-governed, and in line with regulatory requirements, we have a structured validation process. In this 7-step process we assess the objective of the model, the training data, the modelling technique and the governance processes that are implemented. Our findings will support you in improving your models and defining your development roadmap.

Client type: insurance
Technology: Python