AI-Driven Property Valuations: How LOOM and FoxyAI Are Transforming South Africa’s Real Estate Market
Revolutionizing Property Valuation Through AI-Powered Quality & Condition Scoring
99.13% Valuation Accuracy
Outperforming traditional Automated Valuations Models by over 2.6%
35% Market Share Target
AI-adjusted valuations projected to represent 35% of all property valuations by the end of 2025, supported by leading bank and origination partnerships.
Faster Loan Approval Process
Reduced the time for mortgage approval by implementing AI-driven insights that streamline the valuation process, enabling straight-through processing for faster decisions.
All Case Studies
LOOM Property Insights is a South African real estate intelligence company providing data-driven property insights with a focus on enhancing property valuation accuracy through advanced analytics and technology.
LOOM partners with two of South Africa’s top four banks—including one of the country’s largest listed institutions—and provides valuation intelligence and data services to support mortgage decisions. The company also works with 21 leading real estate brands.
Real Estate
Property Appraisals Property Valuations Mortgage Approvals
The Company
LOOM Property Insights is a South African real estate intelligence company providing data-driven property insights with a focus on enhancing property valuation accuracy through advanced analytics and technology.
LOOM partners with two of South Africa’s top four banks—including one of the country’s largest listed institutions—and provides valuation intelligence and data services to support mortgage decisions. The company also works with 21 leading real estate brands.
For the first time, banks are getting access to high-quality property images—including photos of distribution boards, water meters, and sewage inspection points—giving them unprecedented visibility into the properties they’re lending against. This will fundamentally change the way valuations have been done for decades. Banks are already asking, ‘How do we work this into our workflow?’. I believe AI-driven valuation models will become the new norm, much like Automated Valuation Models did. Within five years, we expect widespread adoption.
Jacques Rossouw,
CEO, LOOM Property Insights
Executive Summary
LOOM Property Insights, a leading provider of real estate data and insights in South Africa, partnered with FoxyAI to transform the country’s property valuation system. Facing challenges around property valuation accuracy, cost, and speed, particularly in a market dominated by a small number of banks and a heavily tiered valuation process, LOOM sought a solution that could streamline and modernize property assessments.
By integrating FoxyAI’s advanced, image-based property intelligence models, LOOM significantly improved the accuracy and efficiency of their automated property valuations. This collaboration has set a new benchmark in the South African real estate sector, enhancing decision-making capabilities for banks, real estate professionals, and mortgage originators.
Unlike the U.S. real estate market—where property appraisals are independently conducted and there are thousands of banks—South Africa’s valuation system is centrally managed by a small group of financial institutions. With only 30 registered banks (only 13 locally controlled) and a mortgage process where 56% of home loans originate through mortgage origination businesses like BetterBond, ooba, MortgageMax, and EVO, the need for scalable, automated valuation tools was especially critical.
LOOM’s solution aligns with South Africa’s tiered approach to property valuations: first using Automated Valuation Models (AVMs) and desktop valuations to selectively reduce reliance on costly, time-intensive physical valuations. As a result, LOOM has reduced valuation processing time and set a new standard for decision-making in the South African real estate market.
Key Results
99.13% Valuation Accuracy
Achieved 99.13% accuracy on AI-adjusted property valuations using FoxyAI’s models, outperforming traditional Automated Valuation Models (AVMs) by over 2.6%.
35% Market Share Target
AI-adjusted valuations projected to represent 35% of all property valuations by the end of 2025, supported by leading bank and origination partnerships.
Faster Loan Approval Process
Reduced the time for mortgage approval by implementing AI-driven insights that streamline the valuation process, enabling straight-through processing for faster decisions.
Key Differences Between U.S. and South African Real Estate Markets
Property Valuations
U.S. Property appraisals are conducted independently by licensed appraisers. Appraisals are done for lenders. South Africa Property appraisals and valuations are handled by banks. Appraisals are done by lenders.
Valuation Methods
U.S. Traditional appraisals, conducted by a licensed appraiser and including a physical property inspection, are required for mortgage approvals1. South Africa Property appraisals and valuations are handled by banks. Banks use a tiered approach:
- Automated Valuation Model (AVM) based on internal bank data
- Desktop valuation based on external data
- Physical valuation
Mortgage Process
U.S. Homebuyers work directly with a bank or a mortgage broker for home loan approval. The process is managed internally by the bank, not by an intermediary. South Africa Most home loans go through a mortgage origination process where a third-party broker acts as intermediary between borrowers and lenders.
We ran 731 properties through our AI models, and after adjusting for quality and condition scores, we averaged just 0.41% over the actual selling price. Compare that to traditional Automated Valuation Models, which averaged 2.65% over the actual selling price. We achieved a 99.13% accuracy rate—significantly improving on traditional Automated Valuation Models.
Jacques Rossouw,
CEO, LOOM Property Insights
The Challenge
South Africa’s property valuation system faces several challenges:
Limited Insight in Traditional AVMs
Traditional AVMs rely on static data such as property characteristics, location, and sales history, failing to account for the actual quality and condition of a property.
High Dependence on Physical Inspections
Due to AVM limitations, 25-30% of valuations still require physical inspections, adding significant time and cost to the process.
Centralized and Strained Valuation Infrastructure
Valuations are conducted by banks. With only 30 registered banks, the system is highly centralized and prone to bottlenecks, leading to delays in mortgage approvals and inconsistent turnaround times.
Lack of Objective Property Assessments
Real estate professionals often lack access to standardized, objective measures of property condition, making it difficult to confidently price properties, advise clients, or support accurate market comparisons.
The Strategy
To overcome the inefficiencies of traditional AVMs and accelerate the loan approval process, LOOM integrated FoxyAI’s visual property intelligence models into their AVM and desktop valuation processes. The approach was validated through a comprehensive pilot study.
Pilot Study: Real-World Validation
LOOM ran two methodologies to test FoxyAI’s models:
- Initial Test: LOOM partnered with one of their real estate brand partners (RealNet) and used imagery provided for 109 properties.
- Accuracy Test: 731 properties listed in February 2024 (12 months before pilot study) were assessed. AI-adjusted valuation results were compared to actual selling prices for those properties.
FoxyAI’s computer vision models assessed property images, automatically generated quality and condition scores, and identified key property attributes such as distribution boards, water meters, and sewage infrastructure—features that impact valuation but aren’t typically noted on real estate listings.
In the pilot study, LOOM’s AI-adjusted valuations achieved 99.13% accuracy on average, reducing the error margin to just 0.41% over actual selling prices—compared to 2.65% for traditional AVMs.
The Impact
LOOM’s AI-adjusted property valuation system transformed several key aspects of the property valuation and loan approval process, delivering tangible operational benefits.
By increasing the accuracy of automated property valuations—and therefore reducing the need for physical inspections—LOOM addressed one of the most significant pain points of the traditional valuation process. Streamlining this process enables straight-through processing from application to approval, saving time and costs for banks and accelerating loan approval for homebuyers.
Meanwhile, banks and real estate companies gain access to more accurate and objective valuation tools, allowing them to improve property listings and eliminate human bias from the equation.
LOOM’s partnership with FoxyAI demonstrates how AI can transform property valuations—making them more accurate, efficient, and cost-effective, and ensuring the process is objective and free of bias.
By integrating FoxyAI’s property intelligence technology into their processes, LOOM has enhanced decision-making and efficiencies across the entire South African real estate ecosystem.
As AI adoption grows, this collaboration sets a precedent for the future of real estate valuations worldwide.
