For mine superintendents and plant managers, the end-of-month (EOM) reconciliation process is often a high-stakes exercise in frustration. It is the moment when the mine’s claimed production (total ore removed) is compared against the mill’s claimed numbers (usually backed out from recovered metal).
Because these two figures are often derived using different measurement methods at different locations, variance is inevitable.
Most commonly, stockpile inventories are adjusted from the running totals provided by the Fleet Management System (FMS) to match a survey scan, which is often considered the source of truth.
While this adjustment balances the metal inventory on the books, it does nothing to explain why the numbers drifted apart in the first place.
This traditional approach to mine-to-mill reconciliation is fundamentally limited. It balances the ledger, but it does not provide the insights needed to improve operations.
By the time the adjustment is made, the opportunity for Root Cause Analysis (RCA)—where the real value lies—has gone.
The Complexity of Variance
Variance isn’t necessarily the enemy; it is a byproduct of comparing completely different things. You are comparing a blown-up, dug-up, and hauled geological block model estimate against a metallurgical process accounting calculation. Often missing key measurement points along the way.
The real issue is not that variance exists, but that investigations into it are so time-consuming and complicated that they often aren’t worth the effort, or the answers come too late to be useful. When you rely on averages and lagged data, you are operating with a blindfold.
The Data Gap and the “Black Hole”
A significant source of this confusion lies in the data disconnect between the mine and the mill, specifically within the stockpile.
Consider the basic material flow:
- A truck is loaded at the face, and its payload is assigned an estimated grade from the ore block.
- The truck dumps this payload onto a stockpile.
- A front-end loader later reclaims material from that stockpile to feed the crusher.
These steps are standard mining procedure, but from a data perspective, the stockpile often acts as a “black hole.” The discrete truckload data—specific grade, lithology, and hardness—is typically averaged into the wider pile. When the loader reclaims that material, it isn’t picking up a specific truckload; it is picking up a homogenized blend.
This problem is compounded by the fact that mill accounting often focuses on measurement points that can be at a different resolution than the pit, effectively “making up” the difference via back-calculations. If the mill’s head grade samples are widely spaced (or non-existent), they may fail to catch fluctuations that the mine data predicted. This disconnect makes it nearly impossible to pinpoint whether a variance was caused by dilution, modeling errors, or processing inefficiencies.
Moving from “Rear-View” to “Near Real-Time”
Traditional reconciliation is, by definition, backward-facing. However, waiting for a month-end survey to balance the books is simply too slow for operational improvement. The solution is to move toward near-real-time material tracking.
Using FMS data (often exported with a frequency of ~1 hour), 3D stockpile modeling, and plant digital twins, mines can track material as discrete “parcels” rather than generic averages.
Here’s how this approach changes the workflow:
- Pit-to-Pile Traceability: Every truckload’s data is digitally tracked to its dump location on the stockpile model.
- Dynamic Stockpile Modeling: The stockpile becomes a dynamic 3D model that builds and depletes with FMS events, continuously representing the spatial distribution of the grade.
- Predicted Feed: When a loader reclaims from the stockpile, the system calculates the grade being reclaimed, reporting all ore properties fed to the crusher based on the specific parcels being fed.
Reducing Variance to Expectation
Utilizing a near-real-time, data-driven comparison of the material grade entering the mill with the mill’s actual performance allows for much faster analysis.
If the plant performance deviates significantly, you don’t have to wait 30 days to find out. You get an alert while the material is still relevant. This allows for immediate investigation and faster Root Cause Analysis. Was the dilution higher than expected? Did the mill sensors drift?
This approach shifts the focus from simply balancing the books. Instead of writing off variance as a cost of doing business, you can use it as a tool to tighten your process, improve recovery planning, and ensure that what you think you are mining is actually what you are milling.
