Environmental, Social, and Governance

For most large companies, 70-90% of their total carbon footprint comes from things they don’t directly control. These include suppliers’ factories, consumer use of their products, and their employees’ commute to work. These things make up a category in carbon accounting called Scope 3.¹

Here’s the awkward part about Scope 3. One widely used estimation method, particularly when supplier-specific data is unavailable, is tied to how much a company spends. If a supplier genuinely cuts emissions in half, the buyer’s reported carbon accounting number doesn’t change. But if the buyer just renegotiates a lower price with that supplier, the number drops, even though nothing has changed in the real world. 
That’s a strange way to measure something you’re supposedly trying to manage. And it’s a clean example of a much bigger problem: ESG analytics was built for reporting, but it is now being asked to do decision-support. The two objectives require very different things, and the gap between them is where most ESG programs quietly fail.

What is ESG?

ESG represents a company’s environmental, social, and governance practices. “Environmental” covers emissions, water, and waste. “Social” covers workforce, human rights, supplier labor practices, and community impact. “Governance” covers board structure, executive compensation, ethics, and compliance. A handful of frameworks define what to measure: GRI,² SASB,³ TCFD⁴ (now folded into ISSB⁵), and ESRS⁶ in the European Union. They overlap heavily on carbon metrics but diverge on definitions, especially in the social pillar.

Three Workstreams, One Team

Walk into any large company’s sustainability function and you’ll find three different objectives wearing the same name. 

• Compliance reporting: This is what regulators require. It is backward-looking, audit-grade, and annual. The data must be defensible in front of an auditor.
• External signaling: This is what rating agencies and investors want. It is comparative, ranked, and methodology-dependent. It is different from compliance, because raters apply their own weights and definitions on top of disclosed data.
• Internal decision support: This is what an operations, procurement, or facilities team needs to change behavior. It is granular, real-time, and tied to a specific decision someone has to make this week.

These three objectives need different data architectures, different cadences, and different governance. A compliance number is fine if it lands once a year with full reconciliation. An operational number is useless unless it lands before the decision it’s supposed to inform.

Most ESG programs treat these three objectives as one workstream. The reporting team builds a pipeline optimized for audit defensibility, but then someone asks why operations doesn’t use the data. Operations doesn’t use the data because it shows up too late, at the wrong granularity, in a unit that doesn’t connect to anything operations already cares about. The problem is not necessarily that the reporting metric is wrong. It is that a metric designed for one decision context is being reused in another. 

That’s the translation problem. The Scope 3 example from the opening deserves a closer look, because it’s where this fails most visibly.

The Many Shapes of Scope 3

The estimation method has a name: spend-based environmentally extended input-output (EEIO).⁷ You take how much money you spent with a supplier in a given category, multiply by an industry-average emissions factor for that category, and call it your Scope 3 contribution. It’s the default approach because primary supplier data where every supplier reports their actual emissions to you is rare and expensive to collect.

The shape of Scope 3 varies dramatically by industry. A retailer’s Scope 3 is its suppliers’ factories. A bank’s Scope 3 is its loan book – the emissions from companies that the bank finances. A car company’s Scope 3 is mostly customers driving the cars it has sold. The category is enormous and shaped differently for every business. The problem arises when coarse estimation methods, such as spend-based calculations, are used for decisions that require much more granular information.

The metric is sensitive to spend – not emissions. If you use it to allocate reduction capital, you’ll allocate based on spend changes. You’ll reward suppliers who got cheaper, not suppliers who got cleaner. A 2023 study published in PLOS Climate found that even with machine-learning models trained on the best available Scope 3 data, absolute prediction errors stay high. The data isn’t there to learn from.

This is the canonical translation failure. The metric you have isn’t the metric you need, and no number of better dashboards is going to fix it. What will fix it is investing in primary supplier data, which is slow, expensive, and political.

When Translation Works

Microsoft’s internal carbon fee is the cleanest example I’ve seen of an ESG metric translated into something an operational decision system can consume.

The mechanism is simple. Microsoft allocates emissions to each business unit, then charges that unit a real internal fee for those emissions. The fee shows up as a line in the unit’s P&L. As of FY23, Scope 3 business travel is priced at $100 per ton of CO2 equivalent,⁸ with the proceeds funding sustainable aviation fuel purchases. The rate increases annually through 2030.

What changed downstream is the interesting part. Microsoft’s devices team built a Power BI audit system to track supplier emissions, because it now affects their P&L. The Xbox team redesigned standby power from 15 watts down to under 2 watts. Those decisions weren’t made because someone read the sustainability report. They were made because emissions had been translated into a unit (dollars) that the existing decision system already knew how to consume.

The honest caveat: Microsoft missed near-term operational targets in FY24, mostly because of data center buildout for AI workloads. The carbon fee creates pressure but doesn’t override growth. That’s not a failure of the mechanism; it’s a useful reminder that translation is necessary, but not sufficient. The data has to land in the decision system, but the decision system also has to weight it appropriately against everything else.

Most companies haven’t gotten to step one. Microsoft’s fee is notable because it’s one of the few examples where ESG data has been bolted directly into the financial planning system rather than living in a parallel reporting stack.

The Signal-to-Noise Problem With External Ratings

A quick aside on a related issue: ESG ratings.

If you’ve used MSCI, Sustainalytics, or S&P Global9 ESG scores in any analysis, the MIT Sloan paper “Aggregate Confusion: The Divergence of ESG Ratings” is worth reading. The researchers compared six major ESG rating agencies and found their scores correlate at about 0.61 on average. For comparison, credit ratings from Moody’s and S&P correlate at 0.99.

The researchers decomposed the disagreement into three sources: measurement (56%), scope (38%), and weight (6%). The dominant problem isn’t that raters care about different things; it’s that they measure the same things differently. One rater scores labor practices via employee turnover, another counts labor lawsuits. Both call it labor practices.

For an analytics team, this is a familiar problem in unfamiliar packaging. Your training labels are noisy. Any model or screening rule built on a single rater’s score is overfitting to that rater’s idiosyncrasies. The MIT authors suggest using multiple raters and treating divergence itself as a signal, which is reasonable, but rarely done in practice.

Building ESG Analytics

There are three shifts that matter more than any specific framework choice:

• Separate the workstreams: Compliance reporting, external signaling, and internal decision-support need different pipelines. Trying to make one system serve all three produces a system that does none of them well. 
• Instrument for operational cadence: If your ESG metrics arrive annually, but your decisions happen weekly, you don’t have decision-support. You have a history book. Real-time grid carbon intensity, supplier-level data feeds, and IoT telemetry on facilities are all more useful for operational decision-making than a more polished annual report.
• Translate ESG metrics into units the existing decision system already consumes: These may be a dollar charge, a constraint in an optimization model, or a risk score in a procurement dashboard. ESG data sitting in a separate sustainability portal will not change behavior. ESG data that shows up as a P&L line will.

The major frameworks are increasingly converging and becoming more interoperable. ISSB is consolidating SASB and TCFD. ESRS now interoperates with ISSB. The bottleneck is no longer determining which framework to pick. It’s whether the internal data architecture can support decisions, and that’s an analytics problem more than a sustainability problem.

Which is good news, in a way. It means the people who can fix it are the same people reading this magazine.

References

1. Scope 1, 2, and 3 emissions are a classification from the GHG Protocol. Scope 1 is direct emissions from sources the company owns or controls. Scope 2 is indirect emissions from purchased electricity, heating, or cooling. Scope 3 is all other indirect emissions across the value chain, including suppliers, business travel, and product use by customers.
2. GRI (Global Reporting Initiative) develops sustainability reporting standards focused on organizations' impacts on the economy, environment, and people.
3. SASB (Sustainability Accounting Standards Board) developed industry-specific sustainability disclosure standards designed primarily for investor-relevant information. The SASB Standards 
   are now maintained by the ISSB.
4. TCFD (Task Force on Climate-related Financial Disclosures) developed recommendations for reporting climate-related financial risks and opportunities. Responsibility for monitoring companies' progress on TCFD-aligned disclosures has transferred to the ISSB.
5. ISSB (International Sustainability Standards Board) is the body issuing IFRS S1 and S2, the global investor-focused baseline for sustainability disclosure. The ISSB has absorbed SASB and TCFD. 
6. ESRS (European Sustainability Reporting Standards) are sustainability reporting standards used under the European Union’s Corporate Sustainability Reporting Directive (CSRD). They cover environmental, social, and governance topics and use a “double materiality” approach, considering both how sustainability issues affect a company and how the company affects people and the environment.
7. EEIO (Environmentally Extended Input-Output) is an estimation method that multiplies financial spend in a category by an average emissions factor for that category. EEIO is useful for rough totals, but weak for decision-making, since it’s sensitive to spend changes rather than actual emissions changes.
8. CO2e (carbon dioxide equivalent) is a common unit used to express the climate impact of different greenhouse gases by converting them into the equivalent amount of carbon dioxide based on their global warming potential.
9. MSCI, Sustainalytics, and S&P Global are major providers of ESG ratings and assessments that evaluate companies using their respective methodologies and sustainability-related data.