One of the first steps in open pit mine design is the completion of a pit optimization analysis. Pit optimisation is used to define the most profitable pit shell (or nested pit shells) for a given set of economic parameters. The economic parameters generally include the metal prices, process recoveries, and operating costs. Normally when optimization is done, a range of metal prices or revenue factors is used to develop a series of nested shells to understand how the mine will expand or shrink with increasing or decreasing metal value.
Once optimization is complete, the mine engineer will then design a pit with benches and ramps that mimics the optimized shell as closely as possible. Depending on the slope angles used in the original optimization and where the mine engineer positions the haul ramps, the pit design may or may not very closely resemble the original shell. So the actual tonnage mined will likely be different that the tonnage defined by the optimizer.
Various experts in pit optimization will use approaches of differing complexity. Some may apply variable mining costs with pit depth; apply variable recoveries link to head grade; apply variable pit slopes around the pit or vertically by elevation; etc. One can make the pit optimization step as simple or as complex as one wants it. The question is whether complex pit optimization is warranted. My personal view is that overly detailed pit optimization is likely not required, other than if someone simply wants to test the impacts of parameter variability.
Some of the uncertainties related to the optimization process are described below:

Pit optimization can generate large pits that would ultimately have a long life. However one doesn’t know what the metal prices will be in the future, so be careful in worrying about a high level of accuracy on the initial optimization.

Operating costs will also change in the future, and so again the optimization is just a snapshot using current information.

Some of the nested smaller pits, if developed, would likely be smaller operations and therefore would have different operating costs than assumed in the original optimisation. Similarly some of the larger pits would have different sized throughputs and hence have different operating costs than assumed in the optimisation.

The ore and waste tonnages reported within the pit will be based on a specific lifeofmine cutoff grade, which again has the metal price and operating cost assumptions factored in.

Overall pit wall slopes may differ for shallow or high pit walls, or above the groundwater table and below it, but in many instances during optimization the pit wall angles are maintained the same regardless of the pit depth.

Dilution may be applied the same everywhere during pit optimization unless one is working with a diluted block model. In reality, dilution may be different in different parts of the ore body, but that may not be considered in the optimization stage, thereby adding another uncertainty to the final result.
I agree wholeheartedly with your bottom line but would point out one additional necessary factor, which is a discount rate. Without it you won’t see much in terms of optimal cash flow.
Over time I’ve come to realize that revenue trumps cost by a factor of about 2 to 1, so improving grade or recovery will have more impact on pit size than decreasing waste.
It’s also important to note that pit optimization uses the resource model to develop pit shells, and the accuracy of the resource model must be defined before one spends a lot of time on varying bench costs or other bells and whistles. The ultimate shell will not be more precise than the model allows.
I agree that the resource model is another issue to consider in optimization. If your resource model has a significant proportion of inferred ore, particularly towards the base of the block model where the pit bottom is being optimized, then there is another level of uncertainty. For a PFS or FS, inferred ore is ignored but any uncertainty in the model itself is still a factor in deciding how accurate the optimization result really is.
I don’t ignore inferred when developing the pit shell but I don’t include it when calculating the reserve.
That’s an interesting approach. I don’t recall seeing any PFS or FS where inferred was included in the optimization but then not included when reporting the reserve inside the pit. Usually the inferred is just flagged as waste for both optimization and reporting. Your approach does have merit. At a minimum one should run the optimization with and without inferred to see the impact that it has.
It’s more common in Oz. If you have time, check out Altona. (ASX) Their DFS included work done by a group called Optiro. That’s how they did it.