Definition
Systematic directional error in AI-estimated Scope 3 supplier emissions by procurement category, measured against suppliers who provide verified emission data.
The Measurement Gap
There is no metric that validates AI Scope 3 emission estimates against supplier-verified data by procurement category, which means reduction strategies may target the wrong supplier categories based on systematically biased estimates.
Scope 3 estimation platforms use industry-average emission factors and spend-based or activity-based models to attribute emissions across the supply chain. These estimates drive procurement decisions: which suppliers to engage for reduction commitments, which categories to target for substitution, and how to report progress on science-based targets. But the estimates are rarely compared against verified supplier data: CDP disclosures, audited sustainability reports, or supplier-specific emission factors.
The risk is systematic bias by category. An AI that overestimates emissions for HVAC equipment suppliers and underestimates emissions for cleaning chemical suppliers will direct reduction effort toward the wrong categories. The overestimated category receives scrutiny it doesn't warrant while the underestimated category avoids attention it needs.
Decision Supported
Are Scope 3 estimates systematically biased by procurement category in ways that misguide reduction strategy, or do the estimates track verified supplier data closely enough to support category-level procurement decisions?
Formula
Plain Language
For suppliers with verified emissions data (CDP disclosures, audited reports), compare AI estimates to actuals. Group by procurement category. Consistent directional error within a category indicates systematic bias.
Notation
S3BI_c = (1/N_c) x Sigma (Estimated_i - Verified_i) / Verified_i
Variables
| Variable | Definition | Unit |
|---|---|---|
| c | Procurement category | Category |
| Estimated_i | AI-estimated emissions for verified supplier i in category c | tCO2e |
| Verified_i | Supplier-reported verified emissions for supplier i | tCO2e |
| N_c | Number of verified suppliers in category c | Count |
Units: Signed percentage. Positive values indicate systematic overestimation. Negative values indicate systematic underestimation. The sign matters more than the magnitude: directional bias misguides strategy even at moderate levels.
Proposed formula, pending validation pilots.
At a Glance
Decision Supported
Are Scope 3 estimates systematically biased in ways that misguide reduction strategy?