Every mined block carries two independent numbers: what the geological block model predicted before mining, and what was actually measured afterward by truck scales and plant assays. Reconciliation tracks the ratio between them:
F_tonnes = Σ T_measured / Σ T_model × 100%
F_grade = Σ(T_measured·G_measured) / Σ T_measured
÷ [Σ(T_model·G_model) / Σ T_model] × 100%
F_metal = Σ(T_measured·G_measured) / Σ(T_model·G_model) × 100%
T_measured = T_model · (1 − oreLoss) · (1 + dilution)
G_measured = G_true / (1 + dilution) + assay noise
G_model = G0 · (1 + bias) + estimation noise
The true in-situ grade of each block (G_true) varies naturally around the deposit's mean grade — that natural variability is hidden from the model, which only ever sees its own noisy, possibly biased estimate. When you mine a block, dilution mixes in barren waste rock (grade drops, tonnage rises), ore loss leaves some ore unrecovered (tonnage falls), and the plant's own assay adds a little measurement noise. A perfect operation reconciles at 100% on all three factors; positive dilution pulls the grade factor below 100%, ore loss pulls the tonnage factor below 100%, and a biased block model pushes reconciliation away from 100% in whichever direction the bias points — exactly the discrepancies a mine's reconciliation team investigates every reporting period.
- Start mining — mines one block automatically every ~0.6s; Mine one block steps through manually.
- Dilution / ore loss / bias — change how the next mined blocks diverge from the model; effects compound in the cumulative reconciliation factors.
- Averaging period — how many blocks are grouped into one point on the charts; a longer period smooths block-to-block noise so the underlying trend is easier to read.