Supply Trace: Fighting forced labour with machine learning

Dr Shawn Bhimani, assistant professor of supply chain management at Northeastern University, the lead involved in developing Supply Trace’s platform which seeks to – at its pilot phase – uncover any potential links between shipments and forced labour occurrences in the Uyghur Autonomous Region of China, explains the global crackdown on forced labour remains front and centre of conversations about apparel supply chains at the moment because according to estimates, it is a problem that has gotten worse with time.

“Right now we know there are 27m people in forced labour situations which is a subset of almost 50m people that are in larger conditions of modern slavery. I’m just talking about adults not even children. That number is the highest at any point in human history and it continues to rise with every estimate.” 

He says to some degree it was exacerbated by the pandemic with people having “more vulnerabilities come to them.” 

“But in reality it’s a system that thrives because it lives in a hidden part of society.  

"Slavery was previously known to everyone. Now, forced labour, as it exists as a form of human trafficking, is harder to detect and remediate. It continues to grow rampant in countries like India and China where you have state-imposed forced labour. That’s not to say those two countries are the only ones, many estimates report it is prevalent in every country, even the United States.” 

The International Labour Organization estimates there was 27.6m people engaged in forced labour on any given day in 2021, equivalent to 3.5 people for every 1,000 people in the world. Between 2016 and 2021, the total number of people in forced labour increased by 2.7m. 

But how much of the spike in forced labour is being prompted by an increased appetite for cheap, fast fashion? 

Bhimani concedes to some degree this is the case, together with companies pressuring suppliers for lower-cost production. 

“I’d say it relates to both sides of the equation. It’s important to remember labour relies on links to corporate supply chains to survive. If these systems were not able to make money, there would be no economic reason for people to put others in forced labour." 

Supply Trace is an open-access platform developed between Northeastern and Sheffield Hallam Universities to allow US companies, trade professionals, law enforcement agencies, and civil society to identify potential exposure to forced labour risks within supply chains. 

Currently in its pilot phase, the platform which offers unrestricted access to its data, seeks to uncover any potential links between shipments and forced labour occurrences in the Uyghur Autonomous Region of China. 

It leverages a comprehensive dataset that includes import, export, shipping, and customs data, complemented by publicly available trade records such as shipping and bill of lading data as well as other supply chain information as its baseline. 

Employing machine learning algorithms, Supply Trace automates the process of analysing this data, tracing intricate relationships between entities. Additionally, the platform integrates risk intelligence, informed by research conducted by a team of Uyghur nationals with specialised expertise in forced labour issues within the Uyghur Autonomous Region of 

“Every time we push an update to supply trace, you see more and more facilities that are using forced labour and our risk spotlight will change because we show the most recent ones. 

“The first thing it does is allow for anyone but particularly companies in the apparel trade to start their due diligence process, especially if they didn’t have access before, and understand that some of their suppliers are connected to these facilities. 

“The second thing it does is automate the scanning of import records into a country where consumers are buying products.” 

This, he says, provides visibility into facilities where risk lies – a task that is near-impossible for a human being to do. 

“It’s automating a task that would’ve taken researchers in the past thousands of hours per supplier.” 

The data can then be sorted and viewed in chart form or maps for a visual representation of the at-risk areas in the supply chain. Bhimani punches in search terms for Xinjiang cotton at which point an array of spots across the US light up. Once a spot is clicked on, more details are given about the company and supplier profiles to allow the user to learn more. 

What will be particularly useful is using the platform to justify concerns of third-country pass throughs. 

Bhimani points to a retailer using a Sri Lankan garment supplier before demonstrating the point of origin for those textiles is actually China. 

But, he asserts, the platform isn’t intended to name and shame. 

“This platform is for people I wish the best for – a provision of information so they can make changes in their supply chain. This is not a naming and shaming platform and nowhere on this platform will you notice anything that says these companies are using forced labour. 

The platform is filtered to only include apparel companies but it is designed to cover all industries. 

“We’re doing one sector at a time and being very strategic about how we do it. We want it done well. We want to start in a place we know it is needed.” 

Bhimani explains access at a very large scale to international apparel data has “really been missing from this space.” 

“And now it’s come into fruition alongside machine learning that can actually read all of that; massive amounts of data and machine learning that can actually crunch it. 

“I don’t think companies can do what is required of them in a meaningful and robust way; there's no other tool out there to help them. We’ve created this platform because it needs to exist. It has to exist to create change.”

What differentiates us to many global gold producers is the high-grade nature of our mine. We run an average grade of around 11 grams per tonne of gold, which is exceptionally high and as such we generate more revenues per ounce produced than a lower grade mine would. Roughly, we are producing around 1,000 ounces of gold per month and our objective, in line with our mining plan, is to reach phase two level of around 2,000 ounces per month.

The Australia-headquartered company produced 2,004 ounces of gold and 10,265 ounces of silver in the third quarter to end-September 2022. It said it expects to produce 3,000 to 3,900 ounces of gold in the fourth quarter, following a record production of 3,531 ounces of gold in the second quarter.

Here, Phillip Day, Scotgold’s managing director and chief executive officer, discusses how the company became Scotland’s first commercial gold producer, how it plans to start further drilling in the future, and how it has become one of the highest-margin, lowest-cost gold producers in the world.

The strategy of the company has been to achieve phase one and then to self-fund phase two. We aim to achieve phase two in the first half 2023 with around 23,500 ounces plus run rate of gold a year.

The financing of the project has been a combination of debt and equity. A lot of the costs associated with achieving phase two have come from self-funding, from the profits from the business since it has been operating. The main shareholders are our directors, led by Nathaniel le Roux, who has around 42% of the equity.