Well, it boils down to brokers not having the right tools to assess risk properly. They end up passing the whole assessment burden to the financing providers. It's like trying to bake a cake without knowing the ingredients – you're bound to end up with a mess!
But here's the thing: it doesn't have to be this way. We at Invoys believe AI/ML can be a game-changer in solving this lead quality problem. That's why we're pouring our hearts (and brains) into building an AI-broker platform.
Why is Risk Assessment So Critical?
To reduce acquisition costs and improve financing outcomes, we need laser-focused matching between SMEs' needs and available financing offers. This requires precise risk assessment—both of the SME and their buyers. But while every financing provider has its own risk policy, brokers often lack the means to pre-assess this risk effectively. The result? Low-quality leads that are declined.
Enter Machine Learning: Our Secret Weapon
Machine learning frameworks used in the lending industry, like LightGBM, XGBoost, and CatBoost, are very powerful. They can be trained quickly on millions of data points with thousands of features, identifying patterns that humans might overlook.
We're talking about crunching numbers from structured data (invoice amounts, due dates) and even making sense of unstructured data like transaction descriptions or customer reviews. It's like being able to read between the lines of a company's financial story.
Leveraging ML for Risk Assessment
To achieve quality results, millions of records and thousands of features (attributes) might be required for training. These features include:
• Structured data, like invoice amounts, due date, etc., which can be directly obtained via ERP/EDI systems integration or e-invoice networks.
• Unstructured data such as transaction descriptions or even sentiment analysis from customer reviews or call recordings, which can be mined and transformed into meaningful insights.
By analyzing this diverse set of data, ML models can uncover hidden patterns, enabling accurate predictions of future behavior.
Where Do We Get All This Juicy Data?
Great question! We're not just pulling this out of thin air. We're looking at:
1. Cash Flow Statements: Accounts payables, receivables, and transactional data provide crucial insights into an SME’s financial health. ERP integrations and OCR-AI for PDFs help extract this data in real-time.
2. Behavioral Analysis: By identifying patterns from similar companies, we can forecast an SME’s likelihood of being a good borrower.
3. Customer Reviews and Social Media: Public feedback from sources like TrustPilot or social media activity can provide alternative indicators of a company’s credibility.
4. Job Listings: Hiring trends can serve as additional signals of company growth and stability.
The Road Ahead
Now, I won't sugarcoat it - building and training risk models isn’t easy. Building and training these risk models is no walk in the park. It's more like climbing a mountain... in the rain... It requires large datasets, advanced ML techniques, and a highly-experienced team. But most importantly, it requires continuous monitoring to ensure the models perform as expected. The fundamental requirement is obtaining enough training data, which also means building a great user experience that allows potential borrowers to grant access to their systems of record (such as accounting/ERP) and bank transactions in a matter of seconds.
But you know what? We're up for the challenge. At Invoys, we’re committed to overcoming these challenges. We believe that by using a combination of financial and non-financial data, along with advanced ML models like LightGBM, we can significantly improve lead quality, lower acquisition costs, and help SMEs access the financing they need.
In the coming weeks, I'll be sharing more about our data collection techniques and how we're training and evaluating our models.
Stay tuned as we continue to transform the invoice financing landscape.
