Using Text Analytics to Help Reduce the Risk of Modern Day Slavery

Challenge

Modern day slavery (MDS) is less about owning people and more about exploiting others and being completely controlled by someone else without an option to leave. From men and women forced to work in agriculture or construction to children working in apparel or manufacturing sweatshops, MDS comes in many forms, including, among others, forced labor, human trafficking, child slavery, and debt bondage, which is the most widespread form of slavery (forcing someone to repay the debt through required work).

MDS is a worldwide issue for business supply chains. In a 2017 report from the International Labour Organization (ILO), it was estimated that approximately 40.3 million people worldwide are affected by human trafficking, and just under 25 million of those affected are subjected to forced labor. While governments have the primary responsibility to protect human rights, business also plays a positive role. They continue to improve their internal controls to identify and manage suppliers to prevent MDS. Many companies are also part of sector-specific alliances or multi-stakeholder organizations like the Building Responsibly Initiative to share knowledge and ideas that can benefit industries as a whole. Still, there are inherent challenges like how far and deep they can go in their supply chain to address this issue. How can they realistically monitor 50,000 or 100,000 suppliers across multiple geographies where economic development, the rule of law, and respect for human rights can vary starkly? And, even if they could physically monitor all of these suppliers, is that the most efficient way to do it? These are just a few questions that companies grapple with.

Approach

Building a unique, objective analysis and understanding of the risk is one way to improve business efficiencies and decision-making processes on preventing MDS in expansive supply chains. Earlier this year, we shared our approach of using text analytics to develop a business case for the SDGs. Because a high percent of business information is unstructured and mainly in text form, text analytics is a useful approach as it derives high quality information from text using discernible patterns and trends. We applied a similar methodology to help us better understand the potential risk of MDS in the oil and gas sector:

• Formulate a set of specific questions we wanted to answer or generate greater insight.
• Develop an ontology of concepts and issues based on knowledge and assumptions germane to the industry, market, products and services, as well as key risk factors and what drives them.
• Define our key measurements like, for example, “share of voice” (e.g., public information sources that identify or mention industry within the context of MDS issues).
• Use machine learning tools to ingest the ontology, structure the collection data, and generate objective analyses based on millions of sources of relevant, public information.
• Review and interrogate the analysis and conclusions with subject-matter experts. 

Key Learnings

The analysis provided some useful insights. For instance, the likelihood that MDS will be a key human rights issue for the industry over the next several years is high – around 85 percent. The “confidence strength” for the predictor increases when we added manufacturing, shipping, procurement, and construction. This shows the risk is integral in connection to the “whole-of-industry”. Without these additional aspects, the likelihood is relatively low or around 23 percent.

On the margins of the data, we saw a plausible link to the variability in oil prices, which can push suppliers to do more with less revenue, including sourcing illegal labor. This correlation was particularly salient with oil tankers and shipping.

Covering about 90% of MDS, child labor and forced labor were among the top issues relative to the industry.

However, upon examining the data patterns, there was a blur among many of these aspects. Forced labor is correlated with other human rights aspects. This is reflected in broad conversations that gained traction within the public domain, as well as among policymakers and thought leaders. Although child labor, forced labor and human trafficking is likely to be key MDS issues relative to the industry, it is unlikely these issues will be divorced from among other issues like worker welfare or worker rights. Often, media and other stakeholders like business, government, and NGOs may view these issues as connected or one in the same.

Big data and analytics is not a panacea for preventing MDS in business supply chains. For businesses with a vast network of suppliers, it does offer new opportunities to enhance current due diligence processes, as well as monitoring and management of their supply chain. For example, it can complement supplier surveys that can be inherently biased, focus supplier visits or audits on specific suppliers and potential risk areas, or inform strategic interventions to improve supplier performance in preventing MDS issues. More use cases and piloting of methods and machine learning tools should be encouraged, and learnings shared across key supply chain management functions.