Most waste infrastructure models fail in the same place: a single-point forecast. One tipping-fee assumption, one feedstock volume, one throughput rate, multiplied out to a clean internal rate of return that looks defensible until the first assumption misses. Monte Carlo simulation exists precisely because that clean number hides the risk that actually kills deals.
Instead of asking “what return does this facility produce?” a Monte Carlo model asks “across ten thousand plausible futures, what is the distribution of returns, and how often does this deal lose money?” For an investor weighing a landfill, transfer station, MRF, or biogas project, that distribution is the difference between a disciplined underwriting decision and an expensive guess.
This article explains how Monte Carlo simulation works, then walks through three worked examples specific to waste infrastructure: tipping-fee volatility, feedstock supply risk, and throughput uncertainty. The technique is only as good as the input data behind it, so we close on where those distributions actually come from.
What Monte Carlo Simulation Actually Does
Monte Carlo simulation is a method for modeling uncertainty. Rather than assigning a single value to each uncertain input, you assign a probability distribution: a range of possible values and how likely each one is. The model then runs thousands of iterations, each time drawing a random value from every distribution, and records the result. The output is not a number but a curve, a picture of every outcome the deal could produce and the odds of each.
The name comes from the Monte Carlo casino, a nod to the random sampling at the method’s core. But there is nothing casual about it. Monte Carlo methods are standard practice in project finance, insurance underwriting, and reservoir engineering because they surface tail risk that a base-case spreadsheet cannot.
From Point Estimate to Probability Distribution
A conventional pro forma treats a $65-per-ton tipping fee as a fact. A Monte Carlo model treats it as a random variable: perhaps a distribution centered on $65 with a realistic spread reflecting how much fees have actually moved in that market over the past decade. The same treatment applies to every uncertain input, whether that is feedstock volume, operating cost inflation, or the discount rate.
Choosing the right distribution matters. A sensitivity analysis tells you which variables move the outcome most; the distribution tells you how far each one can plausibly move. Common choices include the normal distribution for symmetric variation, the lognormal for values that cannot go negative and skew high (prices, volumes), and the triangular distribution when you only have a low, likely, and high estimate from a market study.
The Simulation Loop
The mechanics are straightforward once the distributions are set:
- Build the model. A standard cash-flow model of the facility: revenue, operating cost, capital expenditure, financing, and the return metric you care about (IRR, equity multiple, DSCR).
- Define the uncertain inputs. For each variable that is genuinely uncertain, assign a distribution grounded in real market data rather than a guess.
- Draw and run. Each iteration draws one random value from every distribution and computes the outcome. Run this ten thousand times or more.
- Read the distribution. The result is a full picture: the median return, the spread, and, critically, the probability the deal falls below your hurdle rate.
That last output, the probability of loss, is the number a straight base case can never give you.
Worked Example One: Tipping-Fee Volatility
Disposal revenue is the largest line in most landfill and transfer-station models, and the tipping fee that drives it is not stable. It moves with regional capacity, competitive entry, and regulation. Modeling it as a fixed $65 per ton assumes away the single biggest source of revenue risk.
A better approach treats the fee as a distribution informed by how fees have actually behaved in that specific market. If regional tipping-fee data shows fees between $52 and $88 over the relevant horizon, that range, not a single midpoint, is what belongs in the model. Run the simulation and the output might show a median IRR of 14% but a 20% chance of falling below a 10% hurdle, almost entirely driven by fee downside in scenarios where a competitor opens nearby airspace. That 20% is the number that should shape the price you pay and the covenants you negotiate.
Worked Example Two: Feedstock Supply Risk
For biogas, composting, and biomass projects, feedstock is the mirror image of tipping-fee risk on the landfill side: if the tons do not show up, the plant does not perform. A dairy digester assumes a herd size and manure yield; a food-waste facility assumes committed generator contracts; a biomass plant assumes residues within haul distance.
Each of those is a random variable. Herd sizes change, contracts churn, and competing facilities bid for the same tons. A Monte Carlo model that draws feedstock volume from a realistic distribution, rather than assuming the contracted maximum every year, will often reveal that the project clears its debt service comfortably in the median case but breaches its coverage ratio in the 15% of scenarios where supply runs light. Knowing that in advance changes how much supply you contract, how much reserve you hold, and whether the deal works at all.
Worked Example Three: Throughput and Utilization
A MRF or transfer station rated for a given tonnage rarely runs at nameplate. Downtime, contamination, seasonal swings, and inbound-volume variability all pull actual throughput below the rated figure. Underwriting to nameplate capacity is one of the most common ways waste deals disappoint.
Modeling throughput as a distribution, informed by realistic utilization rates for comparable facilities, corrects the optimism. When throughput uncertainty is combined with tipping-fee and feedstock distributions in the same simulation, the interactions matter: the bad scenarios cluster, low fees and low volume tend to arrive together in a soft market, and the combined downside is worse than any single variable suggests. This is exactly the risk that scenario comparison across sites is designed to expose before capital is committed.
Ground your distributions in real market data. A Monte Carlo model is only as credible as its inputs. Wastenaut maps facility-level tipping fees, capacity, and feedstock generators across the US. Open Nexus to benchmark the ranges your model needs, or pull the underlying data into your own simulation via Stream API.
Advantages and Limits
The advantage of Monte Carlo simulation is honesty. It replaces a single number that pretends to certainty with a distribution that admits what you do not know, and it quantifies the probability of the outcomes that matter, especially the downside. It also reveals which variables drive the spread, telling you where to spend diligence effort.
The limits are real and worth stating plainly. Garbage in, garbage out: if the input distributions are guesses, the output is a precise-looking guess. The method is also easy to misread; a wide distribution is a signal to gather better data, not a licence to pick whichever tail supports the deal you already want to do. And the interactions between variables (correlation) must be modeled deliberately, because in waste markets the bad cases are correlated.
Used with disciplined inputs and honest interpretation, though, Monte Carlo simulation is the most reliable way to see a waste deal’s real risk before the money moves.
Frequently Asked Questions
How many iterations does a Monte Carlo simulation need?
For most waste infrastructure models, ten thousand iterations is more than enough to produce a stable distribution. The point of diminishing returns comes quickly: once the median, the spread, and the tail probabilities stop shifting as you add iterations, you have run enough. Complex models with many correlated variables may benefit from more, but the bottleneck is almost always input-data quality, not iteration count.
What is the difference between Monte Carlo simulation and sensitivity analysis?
Sensitivity analysis changes one variable at a time to see which inputs move the outcome most. Monte Carlo simulation varies all uncertain inputs simultaneously, according to their probability distributions, to produce a full picture of possible outcomes. The two are complementary: sensitivity analysis tells you which variables to model carefully, and Monte Carlo tells you the combined effect of all their uncertainty at once, including the probability of loss.
Where do the input distributions come from for a waste facility model?
From market data, not intuition. Tipping-fee distributions come from historical and current regional fee data. Feedstock distributions come from generator-level supply data and contract history. Throughput distributions come from utilization rates at comparable facilities. The credibility of the entire simulation rests on these inputs being grounded in verified, facility-level data rather than national averages or a single consultant’s estimate.
Does Monte Carlo simulation work for small waste deals?
Yes. The value of the method does not scale with deal size; it scales with uncertainty. A single transfer-station acquisition with volatile local fees and one dominant customer can carry more concentrated risk than a large diversified portfolio. If the outcome depends on assumptions that could plausibly be wrong, a Monte Carlo model earns its keep regardless of the check size.