Why Your Supplier Carbon Data Is Wrong (And How to Fix It)

The Scope 3 Accuracy Problem

You invited 50 suppliers to submit emissions data. Here's what you got:

Supplier	Status	Issues
GlobalTech	Submitted	✓ Complete
SupplyCo	Submitted	Emissions 5x normal (typo in zeros?)
LocalVendor	Submitted	40% fields blank (missing production data)
TierTwo	Submitted	Date is in the future (2025)
Subcontractor	Submitted	Same data as last quarter (copy-paste?)
MissingCo	No response	—

Result: Of 50 suppliers, 2 are usable, 3 are suspect, 45 unreliable or missing.

Your Scope 3 estimate is now based on industry averages for 90% of your spend. This defeats the purpose of supplier data collection.

Common Supplier Data Quality Issues

Issue 1: Duplicate Invoices

Supplier invoices same shipment twice:

• Invoice #12345: $50,000 (2024-09-15)
• Invoice #12346: $50,000 (2024-09-16, probably a resubmission)

Impact: Scope 3 spend inflated 2x for this supplier.

Issue 2: Missing Data

Supplier submits incomplete form:

• Production volume: [BLANK]
• Emissions intensity: 0.45 kg CO₂e/unit
• Time period: Q3 2024

Impact: Can't calculate emissions without production volume.

Issue 3: Unit Errors

Supplier reports "500" but is it:

• 500 tonnes?
• 500 kg?
• 500 lbs?

Result: 2,200x difference in calculated emissions.

Issue 4: Out-of-Range Values

Supplier reports emissions 10x their historical average with no explanation.

Impact: Auditor asks "Did this supplier's emissions really spike 10x?"

Issue 5: Suspicious Patterns

• Same data submitted for 4 consecutive quarters (clearly copy-pasted)
• Data always rounded to "1000" or "50000" (suspiciously round)
• Trend opposite to industry (supplier claims -50% emissions while industry grew +5%)

The Solution: 8-Rule Anomaly Detection

CarbonSite automatically detects supplier data quality issues:

Rule 1: Duplicate Invoice Detection

if (count(invoices) > 1 
    AND vendor_id=same 
    AND amount=same 
    AND date_within_7_days)
  → Flag: Likely duplicate
  → Severity: Critical
  → Action: Ask supplier to confirm

Rule 2: Completeness Check

if (required_fields < 80%)
  → Flag: Incomplete submission
  → Severity: Warning
  → Missing fields: [list]
  → Action: Request update

Rule 3: Unit Consistency

if (category="energy" AND unit="tonnes")
  → Flag: Possible unit error
  → Severity: Warning
  → Likely unit: kWh or MWh
  → Action: Confirm with supplier

Rule 4: Statistical Outlier

if (current_value > supplier_baseline * 2.5)
  → Flag: Value 2.5x above supplier average
  → Severity: Info/Warning
  → Action: Request explanation or manual review

Rule 5: Duplicate Submission

if (this_quarter_data == last_quarter_data
    AND this_quarter_data != industry_trend)
  → Flag: Likely copy-paste
  → Severity: Critical
  → Action: Ask for updated data

Rule 6: Date Validation

if (submission_date < reporting_period_start 
    OR submission_date > today)
  → Flag: Date inconsistency
  → Severity: Info
  → Action: Confirm period or corrected date

Rule 7: Cross-Invoice Reconciliation

if (purchase_order_qty != invoice_qty)
  → Flag: Received qty ≠ Invoiced qty
  → Severity: Warning
  → Action: Investigate discrepancy

Rule 8: Price Spike Detection

if (unit_price > supplier_baseline * 1.2
    AND no_market_explanation)
  → Flag: Unit price 20%+ above baseline
  → Severity: Info
  → Action: Optional: investigate

---

Real-World Example: Reconstructing Supplier A's Data

Supplier submitted:

Q3 2024 Emissions: 1,250 tonnes CO₂e
Production volume: [BLANK]
Unit: Tonnes
Timestamp: 2024-10-01

Anomalies detected:

🚨 Critical: Completeness score 60%
   └─ Missing: production_volume, emissions_intensity
   
⚠️ Warning: Value 2.5x above supplier average (baseline: 500 tonnes)
   └─ Likely explanation: Seasonal spike or new production line
   └─ Action: Request confirmation

Supplier confirms:

"Yes, we ran a special production run in September. 
New equipment added 50% capacity. Emissions will return 
to ~500 tonnes in Q4."

Resolution: ✅ Approved: 1,250 tonnes (legitimate, explained) ✅ Note recorded: "Special run Q3, back to 500/quarter baseline Q4" ✅ Emissions used: 1,250 tonnes (not rejected, explanation documented)

---

Supplier Performance Scoring

After each submission, supplier gets a quality score:

SupplyCo Ltd
├── Data Completeness: 85% ✓ (good)
├── Accuracy Trend: Improving ✓ (was 72%, now 85%)
├── Responsiveness: On-time ✓ (3/3 quarters on time)
├── Flagged Issues: 1 warning ⚠️ (production volume spike)
└── Overall Quality Score: 78% (good, trending up)

Scores enable:

• Peer benchmarking: Which suppliers perform best?
• Segment management: Treat top suppliers differently from laggards
• Trend tracking: Is this supplier improving or declining?

---

Scope 3 ML Estimation Fallback

For suppliers who don't respond (17% of original 50), use ML:

Known:
  - Supplier industry: Manufacturing
  - Employee count: 500
  - Annual spend from us: $2.5M
  
Estimated:
  - Baseline emissions: 600 tonnes CO₂e
  - Confidence: 78% (based on similar suppliers)
  - Range: 450–750 tonnes

ML model trained on:

• Responsive suppliers' data
• Historical emissions intensity by industry
• Employee count + facility size

Accuracy: ±15% vs. actual supplier data (when eventually received).

---

From Suspect Data to Trustworthy Scope 3

Stage	Status	Action	Owner
Submitted	50 suppliers	Auto-detect anomalies	System
Flagged	3 suspicious, 2 incomplete	Request corrections	Procurement
Corrected	+5 suppliers now usable	Accept corrected data	System
Unresponsive	45 suppliers (original 50)	Use ML estimation	System
Final	10 actual + 40 estimated	Publish with transparency	Report

Result: Scope 3 you can defend in external audits.

---

Next Steps

1. Invite suppliers → CarbonSite portal with no-login link
2. Collect data → Anomaly detection flags issues automatically
3. Resolve quality issues → Request corrections or use ML estimates
4. Publish with confidence → Defend Scope 3 in audits

---

Related reading:

• Scope 3 Emissions: From Supplier Silence to Collaborative Data
• Anomaly Detection in Carbon Data: Catching the Outliers
• From Field to Finance: The Complete Emissions Data Journey