KPI Mining Solutions

We’re hitting the road and heading to MINEXCHANGE in Denver & PDAC in Toronto! See you there!

Menu

Cutoff Optimization in KPI-COSMO: Improving Destination Decisions Under Uncertainty

Cutoff Optimization in KPI-COSMO: Improving Destination Decisions Under Uncertainty

Introduction: Why We Plan Mines Strategically and What Cutoff Optimization and Geological Uncertainty Have to Do With that
What is the true purpose of long-term mine planning?
The goal of long-term mine planning is to define a strategy that maximizes the mining asset’s value, traditionally translated by the net present value (NPV), by optimizing the sequence of extraction, the cutoff grade policies and downstream decisions over time. This means that the long-term planning does not seek to determine exactly where every block will go since those decisions will be final once grade control data is available. Rather, it attempts to predict the expected amounts, the resulting forecasts, of ore and waste tonnages, grades, cashflows, and others 1–3.

But these forecasts are only as good as the resources model behind them (and geology is uncertain). Using a single and smooth estimated block model4,5 combined with fixed cutoff values hides this variability 6, leading to mine production plans that are overly optimistic or too conservative, that is, risky with high probability of not being materialized 7. 

KPI-COSMO introduces a different approach: a stochastic optimization framework that directly incorporates uncertainty into the decision-making process. Instead of relying on an estimated block model and fixed cutoff grades, it searches for the optimal cutoff policy across multiple simulated versions of the deposit, aiming to maximize Net Present Value (NPV) while controlling risk 8,9. 

By modeling the probability of each block being ore or waste, the optimizer produces not just a plan, but a probabilistic forecast. This reflects the deposit’s uncertainty and supports more confident strategic decisions. This means better predictions of future production, fewer surprises in execution, and improved alignment between plans and reality. 

In this article, we explore how cutoff optimization under uncertainty works in KPI-COSMO and why embracing geological risk – rather than ignoring it, leads to better mine planning outcomes.  

Cutoff Grade: The Engine Behind Forecasts and Mine Value

In mining, the cutoff grade is the minimum grade that defines whether a block is treated as ore or waste. But in practice, this decision is rarely constant over the life-of-mine, and it plays a central role in strategic mine planning. The choice of cutoff should be optimized since it directly affects the forecasts of ore and waste tonnages, the quality of material feed to the plant, and the Net Present Value (NPV) of the project10–12. 

As mine planning advanced, the concept of cutoff grade shifted from a fixed threshold to a dynamic and optimized value, adapting to market prices, costs, and operational constraints. With the advent of stochastic modeling, this evolution went further: cutoff grades are no longer fixed input parameters, but variable outputs of an integrated optimization process that accounts for geological uncertainty. They emerge as strategic decision variables within the planning process itself. 

At the end of the day, why do we optimize cutoffs? Because they determine how much ore we’ll produce, how much waste we’ll move, and ultimately, how much value we’ll create. A good cutoff policy maximizes NPV not by eliminating uncertainty, but by making smarter decisions under it. 

But how do we do that with 50 simulations, without defining a fixed cutoff beforehand, and without scheduling over pre-defined, non-economically optimized pits and phases? That is the magic of stochastic optimization of mineral complexes. 

Stochastic Optimization: A Smarter Way to Plan Under Uncertainty
Unlike deterministic methods that optimize a single scenario, stochastic optimization in KPI-COSMO aims to maximize Net Present Value (NPV) while minimizing the risk of deviatingfrom production targets across multiple simulated realities and sources of uncertainty.
This is achieved through a simulated annealing algorithm that navigates a multi-dimensional search space using three core decision variables:
  • Block sequence – When each block is mined.
  • Block destination – Where each block is sent.
  • Processing stream – Which flow between destinations to follow.
Figure 1: Components considered in the simultaneous stochastic optimization of mining complexes.
In this framework, cutoff policies are not predefined rules, but they emerge as optimized decision variables. KPI-COSMO dynamically adjusts these thresholds during scheduling to identify strategies that maximize value while respecting blending requirements, operational constraints, and plant capacities of the value chain (e.g., the mine complex).

At the end of the optimization process, an optimal year-by-year cutoff policy is selected for each material type at each mine, so as to maximize value and manage risk for that specific context. 

Figure 2: Year-by-year optimized cut-off policy generated taking into account multiple simulated resources model on KPI-COSMO.
This means that a block might be classified as ore in some simulations and waste in others, depending on its characteristics and the optimized cutoff policy applied to its origin and period. As a result, KPI-COSMO can quantify how likely each block is to be economically processable and generate probabilistic forecasts that reflect both value and uncertainty.
While not the goal of the optimization, these processing likelihoods (whether a block is sent to the plant, to the stockpile, or to the dump) are a powerful consequence. They help planners identify high-risk zones, support dynamic decision-making, and ultimately create risk resilient mine plans under uncertainty.
Figure 3: Out of multiple scenarios, KPI-COSMO generates a single extraction sequence. The destination of the scheduled blocks will vary, though, based on the probabilities coming from the simulations.
But this goes beyond individual blocks. Because each block contributes probabilistically to different destinations across simulations, the aggregated production tonnages for each year become distributions – not single values. By applying the optimized sequencing and cutoff policy to each simulated realization of the deposit, KPI-COSMO generates multiple production forecasts for each year. These forecasts (tonnages, grades, economics) are then summarized into probabilistic risk profiles, typically using percentiles such as P10, P50, and P90. These profiles are a key output of the software, allowing decision-makers to visualize and compare different planning strategies, identify riskier periods, and balance long-term value with operational reliability.
Figure 4: Associating a single risk-resilient sequence and cut-off grade policy with the probabilistic block destinations, enables the generation of probabilistic production forecasts for an improved decision-making process.
How KPI-COSMO Chooses the Optimal Cutoff Policy
Unlike traditional mine planning, where cutoff grades are predefined by the user, KPI-COSMO derives the optimal cutoff policy during optimization. Its stochastic engine, the Stochastic Integer Programming (SIP) model, dynamically tests and adjusts cutoff thresholds to maximize NPV while managing risk.
This tuning process considers the full range of block grades across all simulations and adapts the policy to each mine, material type, and year. The resulting cutoff policy defines the economic routing of blocks: whether they are sent to the plant, the stockpile, or the dump.
In the example below, 3 simulations generate a set of histograms of the grades. The optimizer applies cutoff policies (0.7%, 0.3% or 0.2%) to the simulations. The image shows the destination in the colors: red for the blocks that go to the waste rock dump (waste blocks); and green for the blocks that go to the plant (ore blocks).
Figure 5: Single optimized cut-off grade policy across multiple scenarios may seem counter-intuitive. Here is an example on how KPI-COSMO optimized cut-offs.
The optimizer iterates thousands of times to find out the solution that – together with the other 2 decisions variables (sequence and processing streams), maximizes the NPV, minimizes the deviations from production targets and creates an operational sequence. In the simplified example above:
  • Cutoff policy #1 is not selected because does not guarantee maximum NPV (few blocks to the plant).
  • Cutoff policy #3 is not selected because, although it guarantees better NPV, it is generating products with low grade and exceeding the plant capacity.
  • Cutoff policy #2 is the one which maximizes NPV, generating a product with desired quality and respecting the plant capacity.
Once the optimized cutoff policy defined by the stochastic optimization is selected, each block is now assigned a probability distribution of destinations, derived from its simulated outcomes. From the same example, we can create a “hot map” of probabilities of the block going to the plant.
Figure 6: As a result of the cut-off grade optimization on KPI-COSMO, blocks with higher variability will be evaluated in different destinations, while blocks with lower variability will be consistently sent to the same optimized destination.
  • The block in purple, in the top-left corner, has 0% chance of going to the plant – all simulations are going to the waste rock dump. This is a low-risk block.
  • The block in red, in the bottom-right corner, has 100% chance of going to the plant – all simulations confirm it as ore. This is a low-risk block.
  • The block in green, in the center, has a 66% chance of going to the plant under the optimized cutoff policy (0.3%) – 2 out of 3 simulations are going to the plant, and 1 to the waste rock dump. This is a high-risk block.
  • The block in light blue, in the bottom-left corner, has a 33% chance of going to the plant under the optimized cutoff policy (0.3%) – only 1 of the 3 simulations sends it to the plant, while 2 send it to the dump. This is a high-risk block.
Importantly, the algorithm doesn’t eliminate geological risk – it manages it. By assessing the probability of each block going to a particular destination across simulations, the optimizer identifies blocks with high classification uncertainty and tends to delay them in the mining sequence. This not only reduces the risk of early misallocation but also takes advantage of discounting: postponing high-risk decisions generally results in better NPV outcomes.
How an Optimized Cutoff Policy Transforms Mine Planning Decisions
An optimized cutoff policy, supported by stochastic optimization, provides valuable insight into the probability of each block being routed to different processing paths under geological uncertainty. This empowers mine planners to:
  • Identify high-risk blocks near cutoff thresholds, especially in transition zones.
  • Prioritize high-confidence areas, reducing reliance on stockpiles and rehandling.
  • Improve plan adherence, by aligning destination choices with actual block variability.
  • Support strategic decisions, such as evaluating the economic viability of processing routes and avoiding unnecessary CAPEX.
  • And ultimately, proactively manage geological risk, turning uncertainty into a tool for value optimization.
By integrating destination probabilities directly into the optimization process, mine planners can create more resilient and flexible mine plans, especially when multiple processing options or plant investments are under consideration.
Real-World Example: Unlocking Smarter CAPEX Decisions with Optimized Cutoff Policies
We present hereby an example from a real-world mining complex with one existing high-grade product plant, and one proposed low-grade concentration plant. The focus was a specific material type that had previously been classified as waste by the deterministic approach.
The central investment questions – under geological uncertainty, were:
  • Should the company expand its high-grade processing plant capacity to accept this material?
  • Is the low-grade concentration plant truly viable for processing it?

Traditionally, these questions would have been tackled using a single estimated block model and fixed cutoff values, potentially masking critical risk factors. Instead, the team deployed a fully stochastic framework that embraced uncertainty: 20 geological and grade simulations were combined with a unified optimization process, where cutoff grades were not fixed by rules but discovered as optimized outputs of the optimization.

The images below show the probability maps for the evaluated material type, illustrating its probability of being routed to either the high-grade or low-grade plant. 

Figure 7: Probability map outputted by KPI-COSMO. It is possible to identify blocks with low and high variability and their associated chances of being sent to two different processing plants: high-grade (left) and low-grade (right).
The stochastic optimization and the optimized cutoff grade policy unlocked key insights:

1. No Need to Expand the High-Grade Product Plant

The Plant was originally considered for expansion to process a high-grade material characterized by high-grade specs, strong market demand, but significant silica variability and recovery uncertainty.
The stochastic optimization revealed that:
  • Although the mine has high probability of sending blocks to this plant, the risk profiles of feed tonnages dropped significantly after the initial years in many simulations. Several scenarios resulted in early shutdown of the plant, making its expansion economically unjustifiable.
  • On the other hand, as can be seen in the image, the majority of the blocks have high probability of being sent to the low-grade product concentration plant, resulting in higher overall project value.
  • The optimizer considered the synergy of the plants and also the simulations of the main metal and contaminants during optimization, showing that the high-grade product plant expansion is not necessary due to quality risks and declining availability.
  • The probabilistic model demonstrated that expanding this plant would lead to poor returns under uncertainty.
2. Viability of Low-Grade Product Concentration Plant
In contrast, the low-grade product plant proved to be resilient and economically attractive:
  • As can be seen in the images, there are many blocks of the mine that have high probability of being consistently routed to these plants across all simulations.
  • Probabilistic forecasts (P10, P50 and P90) showed stable and sustained throughput over the entire life-of-mine.
  • Strict contaminant constraints were respected.
  • Final product grades consistently met or exceeded commercial specifications.
These results confirmed that lower-grade, lower-variability routes are not only viable but more probable for long-term value creation, especially in uncertain geological.

3. Operational and Economic Benefits Beyond Processing

The benefits extended to mine operations as well:
  • A significant reduction in waste movement and in stripping ratio, over 20%, compared to the deterministic plan.
  • An NPV similar to the deterministic benchmark, but without additional CAPEX and with significantly lower operational costs.
  • All operational constraints — such as sinking rate and minimum mining width — were respected, confirming the plan’s practical feasibility.
Why It Matters
By optimizing cutoff grades rather than using a fixed rule, the model enabled smarter trade-offs between profit, risk, and capital and choose the best decisions for the plants. Instead of committing millions in CAPEX to chase uncertain returns from the expansion of the high-grade product plant with unstable quality and production, the company gained confidence to:
  • Prioritize stable, low-grade product production routes with consistent product performance.
  • Avoid unnecessary investments in high-risk expansion projects.
  • Use risk-informed schedules to support strategic decisions under uncertainty.
This real-world case illustrates the core value of probabilistic block destinations: rather than asking “what can this block be?”, the algorithm asks “what should we do with this block and where should we send it – under uncertainty, to optimize value?”.
That is cutoff optimization. That is stochastic mine planning.

Conclusion: A Shift Toward Resilient Planning 

Cutoff optimization under uncertainty, as implemented in KPI-COSMO, represents a meaningful step toward more resilient and realistic mine planning. It acknowledges geological variability, models it explicitly, and uses it to drive better decisions at every level—from strategic design to daily dispatch. 

When combined with KPI-COSMO’s stochastic optimization engine, cutoff policies are no longer assumed or fixed, but they are discovered, validated, and tailored to risk. This transforms uncertainty from a planning challenge into a strategic advantage. 

References

  1. King, B. Transparency in cut-off grade optimisation ‘Clear-cut ’. in Strategic Mine Planning Conference 26–29 (Whittle Programming Pty Ltd, 2001).
  2. King, B. Optimal mining principles. in Advances in applied strategic mine planning (ed. Dimitrakopoulos, R.) vol. 17 7–11 (The Australasian Institute of Mining and Metallurgy, 2018).
  3. Hall, B. & Stewart, C. Optimising the strategic mine plan – Methodologies, findings, successes and failures. Australas. Inst. Min. Metall. Publ. Ser. 14, 301–307 (2007).
  4. Goovaerts, P. Geostatistics for natural resources evaluation. (Oxford University Press, 1997).
  5. Journel, A. G. & Huijbregts, C. J. Mining geostatistics. (Academic Press, 1978).
  6. Lane, K. F. Choosing the optimum cut-off grade, Colorado School of Mines Quarterly. 59(4). (1964).
  7. Dimitrakopoulos, R., Farrelly, C. T. & Godoy, M. Moving forward from traditional optimization: grade uncertainty and risk effects in open-pit design. Min. Technol. 111, 82–88 (2002).
  8. Dimitrakopoulos, R. & Lamghari, A. Simultaneous stochastic optimization of mining complexes – mineral value chains: An overview of concepts, examples, and comparisons. Int. J. Mining, Reclam. Environ. April, 1–18 (2022).
  9. Goodfellow, R. C. & Dimitrakopoulos, R. Global optimization of open pit mining complexes with uncertainty. Appl. Soft Comput. J. 40, 292–304 (2016).
  10. Lane, K. F. The economic definition of ore – Cut-off grades in theory and practice. (COMET Strategy Pty Ltd, 1988).
  11. Dagdelen, K. An NPV Maximization Algorithm for Open Pit Mine Design. in Application of Computers and Operations Research in the Mineral Industry XXiV 257–263 (Canadian Institute of Mining Metallurgy and Petroleum, 1993).
  12. Rendu, J. M. An Introduction to Cut-Off Grade Estimation. (Society for Mining, Metallurgy and Exploration (SME), 2014).

Douglas Alegre

Douglas is a Mine Planning Consultant with extensive experience in strategic and operational planning, economic modeling, and project evaluation. He has worked across a wide range of commodities—including gold, iron ore, copper, and rare earths—supporting major mining companies throughout Brazil. He focuses on consulting, implementation, and training to help clients adopt advanced, simulation-based mine planning tools. Connect with Douglas on Linkedin.

Request a demo.

See for yourself why KPI-COSMO Stochastic Mining Optimizer is considered the paradigm shift of software for the mining industry. Fill out the form below to schedule a free demo, get answers to your unique questions and discover how our software can help optimize your mining operations.