iz v0 · 2026-05 · open under Apache-2.0
tCO₂ = capacity × route-EF × capacity-factor
Three numbers your own annual report already publishes. No machine learning, no satellites, no proprietary data. We've validated this formula against 20 audit-grade Turkish facilities and shown it beats the EU CBAM default by 85.9% in leave-one-disclosure-out cross-validation across all four CBAM scopes.
The EU's CBAM default overcharges most Turkish cement, steel, aluminum, and fertilizer facilities by 2-10×. We've built a public per-facility benchmark with audit-grade Scope 1 cited to operator-published source PDFs, and a one-line formula that recovers it within ±15% for big emitters.
Formula vs EU CBAM default
−85.9%
cap × route-EF × cf on 20 audit-grade Turkish facilities, all 4 CBAM scopes.
Closed-form, no learned parameters. This is what we ship.
2-layer NN (validates the ceiling)
−84.5% ± 0.3%
95% CI across 5 outer × 3 inner LODO runs. Statistically tied with the formula. We trained a NN to prove the formula is at the data-scale ceiling, not to replace it. Verify in your browser →
When a Turkish operator ships cement, steel, aluminum or fertilizer into the EU, they owe CBAM. If they haven't filed verified emissions, the EU charges them against a default value that's deliberately punitive — designed to push operators toward expensive MRV audits. For most plants, the default overstates real emissions by 2-10×.
Reality is more granular than the EU's default categories. A downstream aluminum roller isn't a primary smelter (23× different EF). A fertilizer plant with an N₂O catalyst on its nitric acid line isn't a plain integrated NH₃ producer (10× different EF). And every plant actually runs at less than nameplate capacity — numbers the operator publishes themselves on a separate page of the same annual report as Scope 1.
iz reads those reports. We've built a public bench of 59 Turkish CBAM-scope facilities,
hand-curated 20 audit-grade strong labels across all four CBAM scopes, and shown that a
one-line formula cap × EF × cf beats the EU default by 85.9% in
leave-one-disclosure-out cross-validation.
A per-facility CBAM-grade Scope 1 number, sourced and defensible. For each plant we hand you:
Two-week turnaround. If your real emissions are below the EU default (they almost certainly are), a single shipment's CBAM savings cover the audit fee. We do not need verifier accreditation to hand you a defensible analysis — we make your verifier's job 10× faster.
Email hi@barisgunaydin.com →Leave-one-out: each plant's audited Scope 1 was hidden during prediction. Ratio = iz-1 prediction ÷ audited truth. EU default = capacity × EU CBAM default EF. Full table at /bench.
| Facility | Sector / route | Truth tCO₂ | iz-1 | Ratio | EU default |
|---|---|---|---|---|---|
| Erdemir Karadeniz Ereğli | steel · BF/BOF | 6,673,266 | 7,217,342 | 1.08× | 7,600,000 |
| İsdemir İskenderun | steel · BF/BOF | 10,663,364 | 12,286,677 | 1.15× | 10,450,000 |
| Kardemir Karabük | steel · BF/BOF | 5,650,626 | 5,486,735 | 0.97× | 6,650,000 |
| Akçansa Çanakkale | cement | 3,466,000 | 3,732,549 | 1.08× | 9,504,000 |
| Akçansa Büyükçekmece | cement | 1,514,000 | 1,541,976 | 1.02× | 7,128,000 |
| Batısöke Söke | cement | 1,577,926 | 1,506,845 | 0.95× | 6,336,000 |
| Çolakoğlu Dilovası | steel · EAF | 566,519 | 593,295 | 1.05× | 5,700,000 |
| İzdemir Aliağa | steel · EAF | 271,123 | 310,105 | 1.14× | 2,850,000 |
| Assan Tuzla | aluminum · downstream | 108,500 | 81,564 | 0.75× | 450,000 |
| Gübretaş Yarımca | fertilizer · blender | 13,281 | 12,996 | 0.98× | 1,200,000 |
Don't take our word for it. The bench loads into a 2-layer LoRA-shaped MLP that trains in your browser via WebGPU in 3 seconds — no backend, no telemetry, no account. Click Train and watch it converge on the same leave-one-out set we used. If we'd fudged the bench, you'd see it here.
Verify in your browser →5 outer LODO runs × 3 inner seeds. Mean ± 2σ. Wider bands = smaller in-stratum n.
| Sector / route | n | Reduction | ±2σ | Range |
|---|---|---|---|---|
| Steel · EAF | 3 | +98.3% | ±2.2% | 97.0 – 100.0 |
| Fertilizer · blender | 1 | +99.3% | ±1.1% | 98.3 – 99.8 |
| Aluminum · downstream | 2 | +86.0% | ±5.1% | 82.8 – 89.8 |
| Fertilizer · N₂O-controlled | 1 | +84.7% | ±2.2% | 83.5 – 86.5 |
| Fertilizer · integrated | 3 | +79.3% | ±8.4% | 75.5 – 85.8 |
| Cement | 7 | +73.8% | ±1.7% | 72.9 – 74.6 |
| Steel · BF/BOF | 3 | −7.3% | ±71.0% | −51 to +48 |
| Overall | 20 | +84.5% | ±0.3% | 84.4 – 84.8 |
BF/BOF stratum is wide because TR has only 3 BF/BOF integrated mills (Erdemir, İsdemir, Kardemir) and the EU CBAM default 1.9 t/t happens to be within ±15% of TR audited reality (1.97-2.40 t/t) on these mills. iz adds value in cement / EAF / aluminum / fertilizer, not on big integrated steel.