Dun & Bradstreet Enlist the Help of AI to Solve a Decades-old Problem
Accurate business function identification remains crucial for many segments and disciplines such as customer segmentation, compliance, and risk management. But the lack of global standard, the sheer volume of volatile variables, and mountains of rapidly accumulating data combine to create a series of problems that ripple throughout the business world. But now, with the help of AI, Dun & Bradstreet is addressing the problem in a new and novel way.
Will the Standards Align?
The Standard Industrial Classification (SIC) code dates to 1930s. Introduced by the United States, this four-digit code gradually gained acceptance by other countries but has failed to assume the position as the global standard. Currently, we have two other systems used throughout the world and this causes issues since these codes do not align. We cannot easily perform one-to-one mapping from one system to another, resulting in a loss of accuracy and reduced granularity, which in turn makes it difficult to share and analyze data effectively.
Dun & Bradstreet provides SIC, NAICS, and NACE throughout its product portfolio. To improve its SIC data, Dun & Bradstreet carried out independent research and added a proprietary extension to the 1987 SIC. Where code cannot be applied locally, D&B applies an industry code model to assign one. Although the modeling of SIC greatly improved the situation, Dun & Bradstreet could not accept the anomalies. So, after careful consideration and consulting with customers, D&B decided to see whether AI could, in fact, improve SIC assignment and precision.
The Brave New World of Neural Modeling
Dun & Bradstreet ran a proof of concept using machine learning Neural Modeling, taking basic Company House ID information, SIC UK2007 descriptions, and data mined from the web. The abundant pieces of information that Dun & Bradstreet has on individual companies, both proprietary and open source, can be read and synthesized by the AI system. After all the evidence has been weighed, the system produces a confidence score which reflects the confidence in each decision.
Dun & Bradstreet utilizes a continuous improvement program and quality assurance tool (QA), to ensure that the process is always learning; inputs help ‘teach’ the machine with “human in the loop” processes. So far, the process has delivered 6.7 mil UK SICs, either newly assigned or verified, and the program is being extended to the US database. The results speak for themselves:
- Improved coverage = 4.3 million additional SICs
- Improved accuracy = 1.2 million verifications
- Improved depth = 1.5 million advancements in precision
As the model evolves there are more implicit datasets (transactional data and social media feeds), and machine learning has enabled identification of industry activity through implied insights such as relationships.
It Works! But Why Is It Important?
Well, there are a lot of excellent examples to answer this question. And we have more details on the process and the promising possibilities regarding this successful application of AI. To learn more, read our whitepaper entitled, Commercial Entity Industry Identification - How AI is improving classification. In this compelling study, our experts tell the story in detail of how AI is driving change – not just for Dun & Bradstreet – but for entire industries.