Improving supply chain intelligence with AI and web data.
Everywhere we look, supply chains are being transformed. The new kind of globalisation is about security, not efficiency: this means that companies prioritise doing business with suppliers and partners they can rely on, and in countries with friendly governments. Before COVID, however, the focus was on efficiency and optimisation of supply chains. The pandemic has highlighted the importance of robust supply chains across all industries and regions. In their 2020 COVID-19 Special Edition of its Global Manufacturing Outlook report¹, KPMG found that 24% of manufacturers listed the supply chain as the biggest threat to growth. In a most recent report², 68% of manufacturers said that ensuring supply chain resilience was now their most important objective.
As a result of this, companies and governments are increasingly conducting a deeper analysis of supply chains — going beyond tier 1 suppliers, down to levels 3 and 4— and diversifying by adding new locations and sources, including on-shoring more parts of the supply chain. But getting good knowledge about supply chains and the potential disruptions still presents a big challenge for many organisations. At glass.ai we have worked on a number of projects related to understanding different aspects of supply chains. We have used our language understanding technology to read the open web and gather supply chain intelligence at a scale that was not possible before. In this post, we share a few examples and why it matters to our clients.
Supply chain depth and potential diversification
Trying to understand a supply chain below the immediate suppliers can be difficult without very good lines of communication between companies and a willingness to share information. In many cases good information is not available, so we have used our technology to collect signals available on the web to build maps of supply chains. The web is the biggest research resource that has ever existed and many companies regularly disclose supply chain-related information on their websites — often listing customers, partners, joint ventures and suppliers in press releases, case studies, customer pages, or via information published in external news sources and social media. Using our web-scale machine reading technology, we have discovered and analysed these sources, finding and following mentions of companies to iteratively map and monitor the supply chains of individual companies, sectors or geographies. General Motors may not disclose details of its supply chain, but its suppliers leave useful information on the web that can be found and analysed. That’s where we come in.
In other projects, our clients have been interested in understanding the broader supply chain beyond their existing suppliers. The purpose here has been to identify potential alternative suppliers that can potentially make a supply chain more robust. We have worked collaboratively with clients and analysed the bill of materials of the products of interest — then we have mapped those inputs onto a taxonomy of terms that describe the items in the bill of materials. This has allowed our language models to explore the business web to map the companies within each relevant area of the taxonomy, essentially building multiple sector mappings to cover the various parts of a supply chain. With the richness of the data, our clients and partners have been able to understand the range of potential suppliers and identify targets for diversification of their supply chains.
Supply chain support: logistics, import and export
As well as using the web data to discover and analyse existing members (or potential members) of supply chains, we have also worked on projects that aimed to identify logistics businesses that ‘support’ the supply chains of various countries. We recently undertook two major projects where we had to identify different types of customs and logistics companies through the reading and classification of text content on the web. Using our AI to perform deeper dives into the websites of businesses, we were also able to identify those that support international supply chains, the countries where they can transfer goods and the types of goods they are capable of transporting.
On-shoring supply chain research
For a government client we were tasked to identify potential opportunities for on-shoring parts of a supply chain. They were looking for companies that were complementary to the existing large employers in the region. There was a desire to improve the capabilities of the businesses that were already in the region so the aim was to discover companies that could either supply to or consume the output from those existing businesses. By mapping the sectors that were either suppliers or consumers of the companies that already existed, we were able to find companies that could potentially help those regional businesses by moving their supply chain closer to them.
Supply chain risks
Another use case is about early warning of supply chain threats. As well as examining the supply chains for particular companies, sectors or regions, it is also possible to use our AI to read news sources and discover issues that are disrupting supply chains. During the pandemic, a client was interested in identifying supply chain disruptions caused by both COVID and Brexit so we monitored and read articles from 2021 to mid-2022 in order to identify articles that referred to supply chain issues. We then analysed the articles further to identify the causes of disruption. Over the period, as well as spotting disruption caused by COVID or Brexit, the system was also able to identify other impacts, for example rising energy prices, Ukraine and the blockage of the Suez Canal.
By further analysing the articles, we were able to pinpoint the impact on particular sectors from the disruption through either identifying the companies mentioned in the articles and/or by assigning the articles to particular industries. This provided an early warning to other companies in the same industry that potential supply chain disruption was coming.
In summary, there are a large variety of supply chain signals that can be discovered on the web for companies, sectors and geographies. Underpinning these examples is the ability to read and understand the written content on the web at scale. It is now possible to provide a level of analysis of supply chains and potential threats and opportunities that was not possible before. We are only getting started.
KPMG, Global Manufacturing Outlook 2020 https://home.kpmg/xx/en/home/insights/2020/11/global-manufacturing-outlook-2020.html
KPMG, Global manufacturing prospects 2022 https://home.kpmg/xx/en/home/insights/2022/01/global-manufacturing-prospects-2022.html