About
Ferra Solutions started in 1997 as "MG's quantitative and investment services" with a focus on delivering rigorous quantitative analysis to the investment community.
Since its inception, the company has developed and applied models to support strategic and management decision-making. Over time, this work has expanded into a substantial body of analysis for policymakers, including institutions such as Eskom, Telkom, National Treasury and the central bank.
In response to increasingly complex challenges, Ferra Solutions began integrating machine learning techniques into its work in 2004. Through combining these approaches with established econometric and time series methods, Ferra Solutions is able to provide more flexible and robust solutions across a wide range of applications.
Alongside its analytical work, Ferra Solutions has developed specialised modelling tools, including a system that has supported the modelling and reporting of monetary policy decisions since 2012. While software development forms part of its capability, the primary focus remains on building and applying models that address real-world client problems.
Services
Ferra Solutions offers a comprehensive range of modelling services tailored to your needs.
If you’re looking to turn complexity into clarity, you’re in the right place. Whether you need an end-to-end pipeline to predict or analyse key outcomes, expert support applying modern machine learning techniques to your data, or an independent review of your existing models, Ferra Solutions can help.
We recognise that sophisticated models and technical expertise alone don’t solve problems. Real value comes from applying them thoughtfully - delivering insights that are both meaningful and actionable.
Services include
- Strategic inputs to assist in transitioning to a data driven management style
- Assistance in migration to employing machine learning and AI techniques
- Model planning and development
- Model operation and reporting
- Technical guidance and training to the modelling team
- Review of model performance
- Review of the model suite and guidance in modernising processes and pipelines
- Assistance in integrating model pipelines into business processes
- Guidance and assistance in setting up modelling infrastructure
In the News
Our work has been recognised across leading platforms and international collaborations, reflecting both depth of expertise and a consistent ability to perform in highly competitive environments.
Scientific American
Using AI to solve a centuries old murder mystery puzzle
"Cain's Jawbone" was published in 1934 as a murder mystery in which the reader is challenged to correctly order the shuffled pages of the book and, in doing so, to solve a number of murder mysteries. The book's publisher joined forces with the Zindi competition platform to see if AI could solve the mystery. Our model placed first in this competition.
Kaggle
Kaggle Competitions
Kaggle is a platform for predictive modeling and analytics competitions in which data scientists and machine learning practitioners compete to produce the best models for predicting and describing the data. Ferra Solutions has participated in numerous competitions and achieved top rankings.
Zindi Platform
Zindi Competitions
Ferra Solutions has consistently placed in the top positions in various competitions on the Zindi platform, demonstrating our expertise in applying advanced machine learning techniques to complex real-world problems.
AI for Good
AI Expo Africa 2025
Through the AI for Good initiative, led by the International Telecommunication Union, Ferra Solutions was invited to present its work at the AI Expo Africa Summit in 2025, showcasing our solution to one of the Zindi challenges.
Case Studies
Directional trading model
Client
Service
Challenge
- Algorithmic directional trading
- Manage proprietary trading book exposures
Solution
- Directional trading model
- Start by optimising technical indicators and augmenting these using econometric techniques
- When the performance of this approach lacked, neural networks were investigated as an alternative
- Neural networks were able to deliver an edge
- The networks were optimised on a large compute cluster
- This cluster was constructed locally using commodity hardware
- The cluster was augmented with AWS EC2 resources when required
- This was 2004 - long before neural networks was a thing
- The solution had to be implemented from scratch!
- It traded a variety of SAFEX instruments, including ALSI and maize contracts
- It was also briefly used to trade S&P 500 and USD/ZAR instruments
- The model was used by other traders but eventually bought exclusively by a large multinational investment bank
- It was used to run the proprietary trading book successfully from 2006 to 2011
- It was discontinued when risk taking was scaled back across the bank
Impact
- Profits!
- Successful proprietary trading throughout the GFC
Funding strategy optimizer
Client
Service
- Funding strategy optimiser
Challenge
- Optimise funding strategy across currencies and the yield curve
Solution
- Integrated cost-at-risk and bond calculator
- Supporting econometric model
- South-African core with foreign economy (USA, Japan and UK) satellites
- Optimise funding strategy using monte carlo simulations
Impact
- Multiple quantitative targets (e.g. optimal currency mix, best duration) to aim for in order to optimise funding
- Insight into how to optimally fund fiscal spending requirements
- Understanding through measurement of the cost of deviation from this optimum
Financial sector assessment programme
Client
- Financial stability department of central bank
Service
- Financial sector assessment programme
Challenge
- Perform model-based assessment of South-Africa's financial sector
Solution
- Integrate big bank income statement and balance sheet models and internal econometric model
- Multiple univariate econometric models to estimate default and recovery rates per asset category per bank
- Develop and simulate stress-testing scenarios and evaluate financial statements under these scenarios
Impact
- Assessment of financial sector health under various stress scenarios
- This was tested in practise during Nene-gate
Business strategy
Client
- Capricorn Financial Group, Namibia
Service
Challenge
- Business strategy to
- Survive the pandemic!
- Expand into neighbouring countries
Solution
- Analysis and measurement of developments in external environment
- Large array of economic indicator models
- Grouped into PESTLE categories
- Modelling internal environment
- Default rate model
- Quantitative inputs into strategy
- The shape of the future!
- These inputs sketched important emerging trends and changes that provided a view of the landscape of the future and that allowed the bank to strategically prepare for these
Impact
- Survival!
- Support growth and expansion into other countries
Spot price models
Client
Service
Challenge
- Accurate forecasts of near-term spot prices
Solution
- Suite of machine learning models
Impact
- Robust and growing collection of forecasts used to calibrate assumptions and downstream model starting points
Econometric model
Client
Service
Challenge
- The modelling team required assistance in estimating and running a new kind of forward looking (SSEM) econometric model
Solution
- Technical assistance and equation estimation
- Develop EViews routines to support and automate SSEM equation estimation
Impact
- Ability to replace core with SSEM model
- Training of staff
Capacity utilisation model
Client
Service
- Server capacity management
Challenge
Solution
- Parse system logs and use naive model to extrapolate usage trends
Impact
- Informed capacity planning and system upgrades
- Effective budgeting
- Proper but easy management of resources
Modelling infrastructure
Client
Service
Challenge
- Modernise modelling processes
Solution
- Deploy shared Python and Jupyter-lab environments across an array of Linux servers
Impact
- Productivity gains
- Modernisation of modelling processes
- Effective collaboration
Business valuation
Client
Service
Challenge
- Estimate value of key customer business to support splitting of business into separate entities
Solution
- Econometric model
- Contract valuation calculator under different economic scenarios
Impact
- Business evaluation
- Strategic inputs
- Quantify required capital allocation or fair selling price of business entities
Modelling
Client
Service
- Bursar and graduate programmes
Challenge
- Training projects for annual bursar and graduate programmes
Solution
- Multiple projects spanning a diverse range of machine learning and econometric topics
- Guidance and assistance to implemented NLP, computer vision, econometric, time series and machine learning models
Impact
- Rapid development of essential skills
- Job creation
- Integration of completed models into business processes to provide downstream value
- Multiple derived projects
Integrated collaborative modelling platform
Client
Service
- Integrated collaborative modelling platform
Challenge
- Collaborate securely
- Manage repository of past results
- Integrate diverse econometric models
- Publish results
- Including press kits
- Performance (error) analysis
Solution
- Integrated collaborative modelling platform
- This was delivered as a COTS solution for 15 years, after which the bank acquired the IP
- It has advanced time series data functionality
- It supports a large number of uni- and multi-variate time series and econometric models
- The platform was developed in Java and has its own NOSQL high-performance storage solution
Impact
- Years of successful inter-team collaboration and model integration
- Years of policy-decision making based on data, models and reports produced by the platform
- Publication and dissemination of model results
- Multiple derived projects
Market risk management
Client
Service
Challenge
Solution
Impact
- Ability to effectively manage market risk
Grain market model
Client
- Ferra Solutions (internal training)
Service
Challenge
- Model the international and local grain markets
- Simulate the impact of the war
- Dynamic and interactive presentation of results
Solution
- Hybrid (econometric, time series and machine learning) model
- Alternative scenarios
- Interactive website with results
Impact
- Upskilling of interns
- Insight into dynamics of grain market
- Understanding the impact of the war
