Rank targets, not intercepts
Three targets. A $2M program. One season.
Every exploration manager has been in this room. The geology team has three areas they believe in, the treasury supports maybe eighteen months of drilling, and someone has to decide where the rigs go. The decision gets made, the program runs, and eighteen months later the company either has something worth financing or it has a lot of very good geological data about a deposit that will never be mined.
What decides that outcome is not usually the quality of the geology. It is the quality of the ranking.
How the decision actually gets made
In practice, targets get ranked on geological excitement. Best intercept. Most continuous mineralization. The section that looks most like a deposit. Sometimes the loudest advocate in the room, sometimes the target that already has sunk cost behind it and therefore momentum.
None of these are bad instincts. They are just answering a different question than the one that matters. Grade tells you whether there is metal. It does not tell you whether the metal can be extracted at a profit, and those two questions can point in opposite directions.
The industry's standard sequence puts the economic question last: geology, then drilling, then a resource estimate, then a PEA. By the time the first pit shell gets run, most of the exploration budget on that project is already committed. The shell does not inform the program. It grades it.
That ordering is not a scientific necessity. It is an artifact of when engineering studies were slow and expensive enough that you only ran one, and you ran it when you had enough data to justify the cost. That constraint is gone. The sequence stayed.
What early simulation can and cannot tell you
Here is the objection, and it is a fair one:
You cannot run a pit optimization on ten drill holes. That is false precision.
Correct. And that is the wrong use of it.
Terrabright does not produce an NPV you would put in a news release. It produces a relative ordering, and relative orderings are far more robust to input uncertainty than absolute values are.
The distinction matters. If your tonnage estimate is wrong by 40%, your NPV is wrong by a lot. But the same error applies to every target you are comparing, and the ranking usually survives it. What you are testing is not "how much is this worth." It is "does this one have a shell at all, and does the other one have a better one."
Consider two targets. Target B has the better intercept: 3.4 g/t against A's 2.1 g/t. Target B is also deeper, with a worse strip ratio and a geometry that pushes cost per tonne up. Sweep the tonnage estimate across a ±50% error band and watch what happens to both.
[CHART: NPV vs tonnage estimation error. Two lines, A above B across the entire range, B crossing into negative territory on the downside. Caption: Indicative NPV for two targets as the tonnage estimate varies. Target A: shallow, low strip, 2.1 g/t. Target B: deeper, high strip, 3.4 g/t.]
The lines never cross. Target B loses at every point on the sweep, and on the downside it does not have a viable shell at all. The higher-grade target is the worse investment, and no reasonable amount of estimation error changes that conclusion.
That is what early simulation is for. Not a valuation. A verdict on ordering.
Two honest caveats. This holds cleanly when estimation error is correlated across targets, which it usually is, because the same team, the same methods, and the same assumptions produce all of the estimates. Where errors are genuinely independent and large, the gap between targets has to exceed the combined uncertainty before the ranking is safe. An early shell is exactly how you find out whether it does. And the failure mode being prevented here is not mis-estimating NPV. It is spending an entire program on the target that was never going to have a shell.
Why we built it
Terrabright was Minebright's first project, four years ago. It came from watching geologists make portfolio allocation decisions with the one input that has almost no bearing on the outcome, not because they did not understand mine economics, but because there was no tool that would give them an economic read at the stage where the decision was actually being made.
Four years in, what we have learned is mostly about where the approach breaks. It is weakest when metallurgy is the swing variable, because early metallurgical uncertainty is not correlated across targets in the way tonnage and grade uncertainty tends to be. It is weakest when the deposit type does not lend itself to open pit at all, where the geometry question changes shape entirely. And it is useless if the underlying geological model is wrong, because ranking bad models against each other produces a confident answer to the wrong question.
What it does well is catch the target that has no economic geometry, early, while there is still budget left to redirect.
What this changes on Monday
You do not need a resource estimate to run this. You need a geological model, rough continuity assumptions, and honest bounds on what you do not know.
Three questions worth asking before the next program is finalized:
- If you ranked your targets by pit shell outcome instead of best intercept, would the order change?
- For each target, what tonnage would it need to have a shell at all, and is that number plausible?
- Which target is in the program because of its economics, and which is in it because of momentum?
The third one is usually the expensive one.
---
Terrabright is Minebright's early-stage pit optimization and mine design tool, built for target ranking before a resource estimate exists. [Get in touch] if you want to talk through how it would apply to your program.