top of page

Machine Learning

Modern organisations or data mature organisations rely more and more on predictive analytics to advance their goals and automate where applicable. Machine learning, AI or predictive analytics all mean the same with different applications, utilising an algorithm based on historical data to predict some event or future event.

machine_learning.jpg

ML and AI is a journey not a end goal

Use cases stretch far and wide, young and old such as credit risk scores to chatbots. Our customers typically engage with us from a project inception phase until a project monitoring phase to ensure project success. Machine learning projects can seem daunting to organizations new to this field, we work with our clients to ensure as much transparency and assistance in order to empower them for best in class results.

 

We develop bespoke solutions and tailor them according to the specific use case, this means that we develop transparent models if applicable or highly complex models for maximum accuracy. We typically deliver these models in API form either in real time or batch response times.

We have a combined experience of 16 years in the area of machine learning and data engineering, ranging from use cases in the finance sector, e-commerce, insurance, IOT, retail and travel. Due to the complexity of machine learning solutions, when needed, we tailor the end product in a way to deliver and allow for configuration by extracting the complexities away from the end user. 

 

We specialize in ensuring your machine learning project sees the light of day in a production environment.​

Improve your business with machine learning

Machine Learning can be complex but in fact machine learning is very demanding in regards to data quality and size. We have a pragmatic approach to ML and would not advice our clients to go this route unless we certain you are ready. 

We have deployed with several projects in Machine Learning and predictive analytics. Below you can see some of the business problems we have experience with

​​

  • Scorecards for bad debt
    Developed a predictable model to provide an estimate of the probability of recovering 80% of a consumer's outstanding debt.

  • Bad Debt Pricing Model Developed a predictable model to provide an estimated value of a bad debt for debt collection companies or companies to bid for a reasonable selling price of their bad debt

  • Location and route algorithm 
    Developed a custom algorithm using telematics data and predicting a vehicle's home and workplaces. 
    Used to determine how far vehicles drive from origin to points of interest, cost reduction and risk

  • Prediction of car theft
    Developed a predictable model using telematics data to predict whether a vehicle has been stolen or not.
    Used in the center to prioritize contact and work

  • Churn predictions
    Developed a model to predict whether a customer will terminate his contract with a company in the near future
    Used to focus efforts on retaining customers at risk to end the relationship

  • Customer segmentation
    Analyzed customer and transaction data to create customer segments to identify additional sales and cross sales opportunities, as well as understand different customer segments.

  • Financial forecast
    Developed an economic forecasting system, used for budgets, enabling input from sales managers and senior executives along with external factors. Further included forecast models on outstanding balance, opening volume and sales as well as outstanding debt.

  • Recommendation System
    Developed an offline recommendation system that was used to send recommended products to customers via email to drive traffic to the e-store.

  • Customer Lifetime Value (CLV)customer lifetime
    Developed a model used to determine a long-term value for each customer.
    Applications Personalized Marketing and Customer Retention

  • Strategies Data Cleaning Prediction
    Model Developed a predictable model to predict the right brand and model and use output as input for reporting dashboards across the enterprise, as the source data contained too many errors to effectively report on.

  • Customer attractiveness
    valuation method of customers' transaction data to derive segments of customers that can indicate the valuable customer segment.

bottom of page