It is four in the afternoon. The rooms are still a long way apart, one lawyer has a flight, and the person who would have to approve anything beyond the last number is not in the building and is not answering his phone. Someone finally says the word impasse. Everyone is a little relieved, and nobody admits it.

The day did not fail at four o’clock. It failed weeks earlier, and most of the people in the building could have told you so.

I have written that the most expensive mediation is the one that fails. It is worth being specific about how that happens, because failure is rarely a mystery and almost never bad luck. When a mediation collapses, the cause is usually structural, and the structure was set long before anyone reserved the conference room.

If the mediator’s first real contact with the case is the morning of the session, the parties spend the most valuable hours of the day educating the person who was hired to guide them. In an estate or trust dispute, with instruments to construe and accountings to follow, an unprepared neutral is not a discount. He is a delay with a fee.

A mediation where the real authority is a committee that meets Thursday, an absent spouse whose blessing is required, or an adjuster reachable only by phone is a mediation conducted by messenger. Every proposal travels to someone who did not hear the reasoning behind it, and proposals stripped of their reasoning are easy to refuse. Before the session, someone must ask the blunt question: who, exactly, can say yes, and will that person be in the building?

I speak with every lawyer in the case before the session, and those conversations regularly change how the day is designed: who should not be in a joint session, which issue is really driving the fight, where the last negotiation broke down. A mediator who walks in without that map spends the morning discovering what he could have known on Tuesday.

A party who arrives believing the case is worth the whole estate has been done no favors. The first realistic number lands as an insult, and the day becomes about absorbing shock instead of making decisions. The kindest thing counsel can do for a client happens before the mediation: a candid conversation about ranges, risk, and what the fight is costing. The mediator will have that conversation too, but it goes better as a second opinion than as breaking news.

Court-ordered mediations conducted only to satisfy the order fail at a rate that should surprise no one. The signals are visible from the front of the room. A half day is booked for a matter that plainly needs a full one. A client arrives having never discussed a number with his own lawyer. An opening position is read aloud that is identical to the one in a pleading filed a year earlier. The parties can tell when the process is ceremonial, and so can the mediator. A court can order attendance. It cannot order the willingness that makes the day work, and a party who decides in advance that nothing will happen is usually proved right.

These cases do not resolve on schedule. They resolve when the grievances have been heard, the arithmetic has been done, and the parties are tired enough to be honest, which is often well into the evening. The hours that feel least productive, the long middle of the afternoon when nothing appears to move, are frequently the hours doing the most work. A mediation with a hard stop at four o’clock gives up precisely the hours in which most of these matters settle. Book the whole day, and tell the clients to arrange their evening around it. The party who has to leave early is negotiating on behalf of everyone who stays.

Set against that list, the prepared mediation is an unglamorous thing. The mediator has read the instrument as well as the submissions, and has spoken with every lawyer. The people who can say yes are present, and so are the people who could talk them out of it at home. Each client has already had the uncomfortable conversation about range and risk with his own counsel, in private, before the day rather than during it. The calendar is clear. Nothing on that list is difficult. It is only work, and it is done in the weeks nobody thinks of as part of the mediation.

Every one of these failures is structural, and every one is preventable. That is the encouraging part. The discouraging part is that the cost of a failed mediation is never limited to the wasted day: the parties leave convinced that settlement is impossible, positions harden around that conviction, and the litigation resumes with new energy.

I will not pretend the inverse promise: that a well-prepared mediation always settles. No honest mediator will. Some matters need a ruling, and a small number truly need a trial. Nothing here is legal advice for any particular case, and no two disputes carry the same risk. But in thirty years of practice and ten on the bench, I saw far more cases that failed to settle for the reasons above than cases that genuinely could not settle at all.

Prepare the day the way you would prepare the trial you are trying to avoid. Put the people who can decide in the room, and keep them there past the hour when it stops being comfortable.

The difference between a mediation that ends a dispute and one that merely postpones it is built in the weeks before anyone walks in.