Most people budget for litigation the way they'd budget for a home renovation: estimate a total, divide by expected months, done. But litigation cost doesn't accrue evenly. It arrives in lumps tied to procedural phases, and each phase tends to cost more than the last because it builds on unresolved uncertainty from the one before.

A motion to dismiss is relatively cheap. Discovery is where costs typically compound, because you're now paying to gather, review, and fight over facts rather than just arguing law. Every phase you enter without resolving the case is a phase where you're paying to reduce uncertainty that the previous phase left standing.

The reason cost compounds rather than simply accumulates is that later phases inherit the complexity of earlier ones. If discovery produces a messy factual record, summary judgment briefing has to work harder to make sense of it. If summary judgment is denied, trial prep has to prepare for every issue the court left open rather than a narrowed set.

This is one reason litigation risk analysis treats a case as a sequence of decision points rather than a single upfront estimate. At each phase, you're not just paying a fee, you're buying (or failing to buy) a narrowing of the range of possible outcomes. A decision tree analysis makes this explicit: every branch has both a probability and a cost, and the expected value of a lawsuit changes as you move through branches, not just at the end.

Cost compounding matters strategically because it changes each side's incentives at different points. Early in a case, a party with a weak claim may have little to lose by pushing forward, since motion practice is comparatively cheap. By the time discovery costs are sunk, the calculus shifts, sometimes toward settlement, sometimes toward doubling down to protect the investment already made.

This dynamic is a version of the prisoners' dilemma settlement problem: both sides might be better off resolving early, but neither wants to be the one who moves first and signals weakness. Modeling this as a game, rather than a single negotiation, helps reveal where one side's strategy is exploitable, for example, a party that reliably escalates costs to force capitulation rather than to develop the record on the merits.

Because litigation is a game played over time, the value of settling isn't static. Settlement value shifts as new information arrives through discovery, as bayesian updating would predict, and as sunk costs accumulate on both sides. A number that looked unreasonable in month two can look prudent in month fourteen, not because the underlying facts changed but because the cost of continuing changed.

This is why sensitivity analysis is useful even for cost projections alone, separate from probability-of-outcome modeling. Small shifts in how long discovery takes, how many depositions are contested, or whether summary judgment is fully briefed can swing total cost substantially. Treating your cost estimate as a single number rather than a range tends to understate the risk of a long, contested case.

Simulation-based approaches, including running a matter through many seeded variations the way monte carlo simulation litigation methods do, let you see cost not as a fixed figure but as a distribution. Some paths through the case resolve at summary judgment; others go to trial. Each path has its own cost profile and its own probability of resolving favorably.

Combining a cost distribution with an outcome distribution is how you get a genuine expected value of a lawsuit rather than a rough guess. It also clarifies your BATNA at each decision point: the choice to keep litigating is only rational when the expected marginal value of continuing outweighs the marginal cost of the next phase, not just the total cost already sunk.

None of this replaces case-specific counsel, and every matter has its own cost drivers depending on jurisdiction, complexity, and the parties involved. But a few general habits tend to produce more realistic budgets.

Building phase-by-phase estimates, revisiting them after each major event (a ruling on a motion to dismiss, the close of discovery, a summary judgment decision), and treating each re-estimate as new information rather than a failure of the original budget, keeps planning grounded in where the case actually is rather than where it was assumed to be at filing.