Tim Tylor-Jones is a professional in the energy industry with over two decades of experience and is currently enrolled as an MSc student in Energy Economics and Finance at the University of Aberdeen.
In 2025, countries that signed the Paris Agreement continue to navigate a difficult trade-off, which is how to reduce emissions at scale without triggering politically or socially unacceptable costs. The policy conversation is shifting away from political grand gestures and pledges to the harder structural realities and practicalities of implementation underneath, especially the emissions that remain hidden in plain sight.
Understanding where emissions come from, not just from direct firm business activities, but across entire value chains, is increasingly central to credible climate action. Demand for transparent reporting and robust data is still growing, driven by investors, regulators, and public scrutiny. Yet the biggest gap remains for Scope 3, the indirect emissions outside a company’s direct control, which are often poorly measured or entirely absent from disclosure.
Scope 1, 2 and 3 emissions now feature consistently and repeatably in discussions about corporate responsibility, sustainability, and climate policy. These discussions are happening both internally at boardroom level and externally in public forums. Scope 1 covers direct emissions from sources owned or controlled by a company (e.g. on-site fuel use or company vehicles). Scope 2 includes indirect emissions from purchased electricity, heat or cooling. Scope 3 captures everything else, the indirect emissions that occur across a company’s value chain, from suppliers to customers and even end-of-life.
The Greenhouse Gas Protocol’s 15-category framework, established in 2011, has become the global benchmark for reporting Scope 3. In many industries, these emissions account for between 70% and 90% of the total footprint, but reporting remains patchy. Estimates often rely on sector averages or input–output models rather than measured data, leaving major blind spots in company disclosures.
Why Is Scope 3 So Tricky?
Supply chains in 2025 are more global, interwoven, and opaque than ever. This complexity is in most cases born out of necessity to stay competitive and secure the suppliers and skills required, a strategic necessity. It’s the trade-off firms make to stay competitive in a high-efficiency, low-margin world shaped by decades of globalisation and digital acceleration. But complexity can’t be treated as an immoveable barrier to better Scope 3 accountability. Companies still design and manage their supply chains, and they remain accountable for the emissions embedded within them, whether or not those emissions are easy to measure.
Even when firms want to report Scope 3 properly, the data often just doesn’t exist. Most companies can’t see much beyond their direct suppliers of goods and services (Tier 1), and even this level can be incomplete. Further upstream or downstream, emissions are tied up in dozens of actors, each with different systems, reporting standards, or incentives to share information.
In the UK, frameworks like the Streamlined Energy and Carbon Reporting (SECR) regime and the Task Force on Climate-related Financial Disclosures (TCFD) both encourage companies to report Scope 3 emissions, but they don’t require it outright. Instead, they follow a “report or explain” model, where firms are expected to disclose if they can, or explain why they haven’t. While the reasoning behind this design is understandable, it makes the Scope 3 reporting landscape inconsistent at best and barren at worst. In the EU, mandatory Scope 3 disclosures were planned to be implemented for 2025 under the Corporate Sustainability Reporting
Directive (CSRD). Recent political headwinds have led to a formal delay of key provisions until 2028 for many companies, a sign of how politically sensitive this space remains.¹
Inference: A New Way to See
Rather than wait around for perfect data, some researchers are challenging the problem proactively. If companies won’t disclose their full supply chain emissions, or simply lack the data control to do so, maybe we can estimate or predict them another way. A growing body of work is doing exactly that, inferring supply chain structures using models that don’t rely on firm-level emissions data. These predictive algorithms use inputs like location, sector, revenue, and public filings. Combined with assumptions about typical supply chain behaviours, they reconstruct the likely connections between firms. Once a supply chain is constructed, emissions can be estimated across the inferred network.
For example, Fessina et al. (2024) showed that predictive supply chain maps can be built using only sector-level input–output data. Mungo (2023) takes it slightly further using machine learning and network reconstruction. His model treats companies as nodes and predicts missing links between them based on ‘dyadic’ traits, i.e. country, industry, and sales, which are then fed into a gradient boosting framework. Even with sparse data, the model achieved high accuracy in identifying likely supply chain relationships.
Often the goal of forward modelling is just to try and replicate a synthetic output that closely replicates the real-world dataset, to validate the accuracy of the model for a specific set of variables. This is essential for model validation, but this approach goes further. The aim is visibility, to pinpoint where emissions actually flow, and who is indirectly responsible. Tabachova (2025), also at INET Oxford, applied these techniques to carbon pricing. Her study used a contagion-style model to simulate how emissions costs ripple through supply networks. The study found that when inference techniques were not used, carbon exposure and risk were systematically underestimated.
More recent work is extending this idea, using inferred networks not just to measure emissions, but to test how different climate policies might affect supply chains and economic systems. These models are starting to offer a way to explore policy impacts, not just in theory, but in terms of where costs and pressures might land in practice.
From Data Gaps to Emissions Mapping
Scope 3 is the part of the iceberg below the surface, massive and mostly hidden from view for emissions models. Active research in inferred supply chain modelling is helping generate vital maps that show where companies should focus their time and resources to be most effective at reducing emissions. These techniques can’t solve everything about Scope 3 emissions, but they are a tool to bring visibility to the problem. That shifts the narrative from companies saying “we don’t know” to “we can estimate, and act.”
As climate policy and capital allocation become ever more data-driven, working with these models, and embedding them into reporting standards, climate risk frameworks, and transition planning, could be transformative. Inference won’t remove all uncertainty from Scope 3 mapping, but it can make it manageable, giving firms and policymakers a clearer picture to act on
Sources and Further Reading
Fessina, A., Cimini, G., & Mazzilli, D. (2024). Supply chain inference and carbon accountability using input–output data.
Mungo, L. (2023). Reconstructing supply chains with machine learning. INET Oxford Working Paper.
Tabachova, I. (2025). Carbon risk contagion in supply networks: An inference-based model. INET Oxford.
Greenhouse Gas Protocol. (2011). Corporate Value Chain (Scope 3) Accounting and Reporting Standard.
EU Commission. (2024). Corporate Sustainability Reporting Directive (CSRD).
UK BEIS. (2022). Streamlined Energy and Carbon Reporting (SECR) Guidance.
TCFD. (2017). Recommendations of the Task Force on Climate-related Financial Disclosures.
Post your comments and questions for the speakers here