Leveraging Data

Data has become far more abundant and accessible in the modern world. Whether this data is from instruments in an industrial plant, remote sensors, or customer transactions, it can contain a wealth of information for a business. As well, statistical methods and Artificial Intelligence (AI) algorithms have seen a period of rapid advancement resulting in a huge variety of data analytics and predictive analysis methodologies becoming available. We work with our clients to help them use state-of-the-art tools to analyze their data to improve existing processes or to develop new ones.

Relevant experience, trusted capabilities

A central part of the problem-solving process developed and applied at Coanda is the gathering, handling, and interpretation of data through which the fundamentals underpinning the problem can be uncovered. To support this approach, Coanda has made significant investments in both technical personnel and hardware, employing PhD-level mathematicians and physicists that are versed in the most recent Big Data Analytics and Machine Learning techniques in and operating a 800+ core High Performance Computational (HPC) cluster.

Leveraging Coanda’s past experience and significant infrastructure allows our clients from a variety of industries to generate insights, identify patterns and extract correlations from large data sets to gain new levels of understanding:

  • A team of PhD-level data scientists, familiar with state-of-the-art methodologies
  • Extensive computing capabilities and infrastructure
  • Experience with handling sensitive customer data
  • Proven results with implementation of Machine Learning models for clients
  • A client-centric approach to delivering custom solutions

In the last year alone, Coanda’s Advanced Instrumentation and Data Science team has completed several successful programs:

Empowering small banks

More recently Coanda has been working within the banking sector to develop algorithms to enable smaller banks to remain competitive, especially against their larger counterparts.

Banking data is typically managed by third parties. The costs for gaining access to this data can be prohibitive for smaller banks, limiting their ability to leverage the benefits of Big Data and Machine Learning.  This gives larger banks a competitive advantage.

We provide low-cost ways to circumvent this barrier, with the ultimate goal of enabling smaller banks to apply big data analysis and machine learning to their benefit.

Some Big Data and Machine Learning options requires complete restructuring of data collection and management strategies.  We instead provide focused, application-specific solutions, collecting data from existing access points or historic statements, and delivering simple solutions that provide immediate returns.

This can include data analysis and models to achieve substantial improvements to:

  • Customer attrition prediction
  • Risk modelling for lending
  • Product recommendations and other “know your customer” applications

We perform/provide the following services:

  1. Data collection
  2. Machine learning model development and optimization
  3. Software integration with existing systems and practices
  4. Continuing model support and software updates

Case Study

Account Closure Prediction Algorithm – “Unchurn”

“We have developed our algorithms to enable small banks to better leverage their data and to remain competitive, especially against their larger counterparts. Our trained models can analyse large data sets in seconds to identify complex patterns in behaviour and predict valuable outcomes.”
Dr Neville Dubash,
Data Scientist

Machine Learning models greatly improved the accuracy for targeted actions to avoid loss of clients (churners).

We worked with a community bank in the US to enable them “scrape” data from their historic bank statements, and then develop state-of-the-art Machine Learning models to identify customers who were about to close their accounts.  Compared to basic strategies looking at changes in account balance, the Machine Learning models had roughly a 20-fold improvement in prediction accuracy for account closures.

Download our Machine Learning Case Study