How to Run a Sensitivity Analysis on Your Case's Expected Value
A method for finding which assumptions move your case's expected value the most, so you know where to focus diligence.
A case's expected value is built by combining several judgment calls: probability of prevailing on each element, a damages estimate, litigation cost, and time. Every one of those inputs is a dial you set, not a measured constant. Sensitivity analysis asks a simple question about each dial: if this number were meaningfully higher or lower, how much would the case's expected value actually move?
The answer is often uneven. Some dials barely matter — the expected value looks almost the same across the whole plausible range. Others swing the number substantially. Knowing which is which turns a vague sense of 'this case is uncertain' into a concrete list of what to investigate next, and it keeps you from either over-investing in a low-impact dispute or under-investing in the one fact that actually decides the number.
Start by listing every dial that feeds the number
Before running anything, write down every input that goes into your expected-value calculation: probability of prevailing on liability, probability on each contested element, the damages components that make up the total ask, litigation cost to each stage, expected time to resolution, and any discount you apply for delay.
This list is your baseline case. Sensitivity analysis only makes sense relative to a stated starting point — you are asking how far the output moves when one input moves, not producing a number in a vacuum.
Vary one dial at a time (one-way sensitivity)
The simplest form holds every dial fixed at its baseline value except one, moves that one dial across its plausible range, and records the resulting expected value at each point.
- Run this for each dial separately, then rank them by how much the expected value swings from low to high — this is the basis of a tornado chart, where the widest bar sits on top.
- The dial that produces the widest swing is your dominant driver — the one where being wrong matters most to the bottom line.
- A dial that barely moves the output across its entire realistic range is one you can stop agonizing over — refining it further has limited payoff.
Then test dials in combination (two-way and scenario sensitivity)
One-way sensitivity assumes every dial moves independently, which is often unrealistic. If a key witness gets discredited, both liability probability and the credibility-dependent damages components may move together. Two-way sensitivity tests a pair of dials moving in combination; scenario sensitivity instead defines a small set of coherent stories — a best case, a base case, and a worst case — and computes the expected value under each as a package rather than assuming independence.
Use the results to prioritize investigation, not to fake precision
The point of the exercise is not to arrive at one more-precise number. It is to know where the real uncertainty in your estimate lives.
- Direct discovery, expert, and diligence budget toward the one or two dials the analysis flags as dominant, rather than spreading effort evenly.
- Treat a wide, flat swing across a dominant dial as a signal that the case value is genuinely uncertain right now — that is useful information for pacing a settlement conversation, not a flaw in the model.
- Use the tightened range you get after investigating the dominant dial to set a more defensible demand or offer bracket, rather than anchoring to the original point estimate.
Where sensitivity analysis breaks down
Smooth dial sweeps handle continuous uncertainty well but can misrepresent threshold effects — an element is either satisfied or it is not, and a probability dial approximates that binary reality rather than describing it exactly. Sensitivity analysis also only shows you how the output reacts to the dials you built into the model; it says nothing about whether the dials themselves are well-calibrated. It is a tool for understanding your own estimate, not an external check on it.
- What's the difference between sensitivity analysis and Monte Carlo simulation?
- Sensitivity analysis moves one or two dials at a time to see how far the output swings. Monte Carlo simulation samples many dials simultaneously across their full distributions to produce a range of possible outcomes. They are complementary — sensitivity analysis tells you what to focus on, Monte Carlo shows you the resulting outcome distribution.
- How many dials should I actually test?
- Test everything that feeds the expected-value calculation, but expect the list to be short in practice — usually liability probability, one or two key damages components, and litigation cost to the next major stage.
- Can sensitivity analysis tell me whether I should settle?
- No. It tells you where the uncertainty in your estimate is concentrated so you can decide with better information. The settle-or-litigate decision still depends on your own risk tolerance and BATNA.
This page is an educational explainer, not legal advice, and creates no attorney–client relationship. Juricratic is a simulation engine: every probability-like figure is a dial you set, not a calibrated prediction. Verify every rule, deadline, and figure against the authorities and orders that govern your matter.
Stop estimating one number at a time.
Juricratic models the whole matter as a solvable game and runs it thousands of times — so the settlement value, the risk, and the optimal line all move together when the facts do.
Request access →