Risk assessment should be adapted to the business and the risk profile you choose to have, but there are a few simple principles that apply across the board.
The simplest of these is the quality of the information you use. The better the information you have, the better you can estimate the risk. You rarely get everything you want in this world, so the more up to date the information is, the less precise it usually is. One example is news articles compared with accounting figures. The first category holds more recent information, retold by someone who may not have a full overview of the company. The second is very thorough, with last year's accounts, which are already out of date.
This is how we are building our new risk model for companies
This is what we emphasise:
- Machine learning for continuous learning
- Public registries for context and tracking of change
- News for events that provide timeliness
- Sector-specific sources for in-depth understanding of the sector
- Industry analyses to understand trends
- Scraping and text analysis of websites for unique insight into the individual company
- Ownership structures to understand the wider picture
- Customer-specific data for experience
These go through refinement processes in which we have built a range of analyses that extract even more insight from the various sources. The processes also combine the sources so that the model can make sense of them.
What sets the model apart from traditional company risk assessment
For many people, risk assessment of a company means a model based on accounting data and payment remarks, often helped along by a few rules that adjust it afterwards. These use only very precise data, but often data that is out of date. That does not make them useless. It is a solid basis to build on, but we believe you have to do more.
In our models we combine both structured and unstructured data, so that we get a more precise understanding of the company. It is updated regularly as new information becomes available, whether that comes from public registries, from the news, or from new financial statements. By using machine learning, that is, models that learn relationships on their own, you can draw on vast amounts of real-world experience to predict what lies ahead.
The weighting of data relevance sits in the refinement process
The "magic" in what we are building lies in the refinement processes themselves: the more we can help the model, the better the predictions it can make. How much weight the model gives each source and each data point is something it learns for itself over time. Some of the sources are more useful for human interpretation of the result but not useful to the model, and those are used only to show context around the prediction. Others are hard for people to understand but important to the model, and those are shown to the model only. We integrate new sources continuously, and we bring them in quickly whenever they give us deeper insight.
The end result is a model that we believe hits the mark better than the traditional ones. It is perhaps a little harder to understand, since it takes in thousands of factors where traditional models take in dozens, but it gives us a very current picture of the state of a company. We believe that is decisive for making choices based on data.
Guidelines for using data for credit information
We want to stress how important it is to distinguish between risk assessment and credit information. Risk assessment is used for acquisitions, for example, and when looking into collaborations and partnerships. When creditworthiness is assessed, the sources and methods are subject to licensing and will therefore differ from the model described above.
Sole proprietorships that are not registered in the Register of Business Enterprises are also treated as private individuals, with specific requirements for the processing of personal data. Those companies, and the private individuals themselves, are not covered by Enin's risk models. There are separate guidelines that must be followed for information about key people connected to businesses as well.
If you would like to take part in developing our new risk assessment model for companies, or have input we ought to know about, we would really appreciate it if you get in touch with us here.
