Two experts, one schedule. Where is the settlement zone?

What it does

Read this first: what comes out is a range of compensable days under two experts' readings of the same record, not a finding on the claim. A fit-out contract finished 71 working days late, and the windows analysis has already settled how many days each of nine events put on the critical path: 12 for late site possession, 9 for the steel connection redesign, 10 for the steel fabricator, 8 for the rock in the basement dig, and so on down to 5 for the fire-stopping rework.

Nobody is going to move those numbers. What the two experts disagree about is who owns each event, and that is the uncertainty this model draws. Each event carries four probabilities that say how strongly the record supports each reading: owner caused, which is time and money for the contractor; concurrent, which is time and no money; excusable but not compensable, such as the storm week, which is also time and no money; and contractor caused, which is nothing, and liquidated damages exposure.

Late site possession reads 80 percent owner in the claimant's block because the handover letter is dated. The rock in the basement dig reads 45 percent owner, 25 percent excusable and 20 percent contractor because the site investigation clause cuts both ways. The steel fabricator's late delivery reads 70 percent contractor because the fabricator was the contractor's own subcontractor.

The sheet opens on each expert's most likely reading of every event: 44 compensable days, 21 contractor-caused days and 6 days of extension without money for the claimant's expert, and 21, 28 and 22 for the respondent's. Those two opening stories are 23 days apart and neither of them is what the model expects, because a most likely reading of nine separate events is not the most likely total.

Click Run on the 20,000 trials the template loads with. The seed is set to 7 so your first run reproduces these figures exactly. For the claimant's expert, compensable days come back with a mean of 28.7 and a median of 29, a tenth percentile of 16 and a ninetieth of 42. Forty-four compensable days or more, the story the sheet opened on, happens in 7 in 100 runs.

The chance of clearing 30 compensable days is 40 in 100. Contractor-caused days on the same reading average 20.2 with a median of 21, and the days that earn an extension of time but no money average 22.1. For the respondent's expert, compensable days average 14.37 with a median of 12, a tenth percentile of 0 and a ninetieth of 28, and 15 in 100 runs give the contractor no compensable days at all.

Their own opening story of 21 days or more happens in 26 in 100 of their runs. Now put the two distributions side by side, which is the reason for the second block. The claimant's tenth percentile is 16 days and the respondent's ninetieth is 28, so the two readings of one record overlap between 16 and 28 compensable days. That is the settlement zone: the claimant's own model puts 91 in 100 runs at 16 or more, and the respondent's puts 91 in 100 at 28 or fewer, so a number in that zone is one that neither side's expert can call implausible on their own figures.

The probability-weighted lines at the foot of the sheet, 28.7 and 14.37, are the two means and they sit inside the zone too, but a mean of nine draws is not a day count anybody will ever see. The tornado is where the model earns its place, because it ranks the judgment calls rather than the events. For the claimant's compensable days it puts late site possession first, then the steel connection redesign, then the late finishes selection, then the rock in the basement dig.

The correlations are negative, at -0.45, -0.42, -0.38 and -0.37, because a higher code is always a worse reading for the contractor. Those four judgments carry 26, 22, 18 and 17 percent of the spread, and the storm week comes last at -0.06 because both sides already agree it is excusable. For contractor-caused days the ranking flips: the steel fabricator's late delivery is first at 0.53 and carries 43 percent of that spread on its own.

For the respondent's compensable days late site possession carries 53 percent of the spread by itself. Whichever side you are on, that one event is where the record needs to be read most carefully, and the tornado tells you so before anybody has argued about it. Second run, and it is the one that shows what a concession is worth. In the Risk Analysis panel, change the respondent's late site possession draw to the claimant's own 0.80, 0.10, 0.05 and 0.05, and rerun.

The respondent's compensable days move from a mean of 14.37 to 18.57 and from a median of 12 to 20, the runs with no compensable days fall from 15 in 100 to 5 in 100, and the tenth percentile rises from 0 to 8. The ninetieth percentile barely moves, from 28 to 29. Conceding the biggest event moves what the respondent's expert expects by eight days and the top of the zone by one: a concession moves the center of the argument, not its ceiling.

What the model cannot tell you. It draws the nine judgments independently, and a tribunal that reads late site possession against the owner will often read the late finishes selection the same way, so the real spread is wider at both ends than the percentiles here. Each event is wholly one reading in each trial, so an apportionment of one event between the parties only appears as an average across trials, never inside one.

It says nothing about the merits, the notice provisions, the daily rate or the liquidated damages clause, and it assumes every one of the 71 days sat on the critical path, which is the windows analysis's job, not this sheet's. And the overlap is a zone where the two readings agree, not the chance the parties settle. To make it yours, retype the events and their days from your own analysis, put each side's four probabilities beside each event, and set the same probabilities on the matching draws in the Risk Analysis panel, because the probability columns on the sheet are documentation and the panel draws from its own rows. The probabilities total column should read 1.00 on every event before you run.

Next question

• Neither side wants to blink. What happens?Deadline Standoff: Hold Firm or Concede?
• Should you take the settlement or go to trial?Settle or Go to Trial?
• How much should we reserve for this lawsuit?What Could the Lawsuit Cost Us?
• What range can a lost profits opinion actually support?How Much Are the Lost Profits Worth?
• Whose cost to complete should you believe?Is the Cost to Complete Claim Credible?
• Did that delay actually push the finish date?

Every model like this one, and the method behind them: Monte Carlo simulation.