Almost every technique for evaluating a proposed development rests on the same quiet assumption, which is that something sufficiently similar has already been built and observed. Take that assumption away and the machinery does not break loudly. It keeps producing numbers, and the numbers look exactly as confident as they did before.

This is the pillar I am least comfortable writing, because the honest treatment of it runs straight into subjects I do not discuss publicly. So I am going to stay on the mechanics of the reasoning and leave the arithmetic alone entirely. No figures appear here and none are coming.

The mechanism first. A comparable does two jobs at once and they are usually confused. It supplies an expected outcome, and separately it supplies a measure of how wrong that expectation has historically been. The second job is the one that matters for an unusual building, and it is the one nobody notices is missing, because a spreadsheet will happily accept a single point estimate without any sense of its own dispersion.

When no comparable exists, what replaces it is decomposition. You stop trying to price the building and start pricing its parts, each against something that does have a history. Structure of a given type. Facade of a given complexity. A review process of a given length. Trades of a given scarcity. The building is unprecedented; almost none of its components are. Uncertainty then attaches to the parts, where it can be reasoned about, rather than to a single figure at the end where it cannot.

That decomposition has a failure mode worth naming. Correlation. The parts of a difficult building do not go wrong independently. The same conditions that extend a review also extend a schedule, and a longer schedule changes what the trades cost and who is still available. Treating the components as independent produces a total uncertainty that is too small, and too small in the direction that feels reassuring.

The second mechanism is sequencing what you learn. On an ordinary project the order in which uncertainty is resolved barely matters because nothing is very uncertain. On an unprecedented one the order is most of the discipline. Some questions can only be answered by spending real money, and some of that money is unrecoverable if the answer comes back wrong. So you buy the answers that could kill the project first, cheapest first within that, and you accept that a portion of that spending is simply lost. I regard that as the entry fee.

Here is my arguable claim. An unprecedented building is penalized twice for the same uncertainty, and the second penalty is not rational. Once correctly, because genuine uncertainty deserves a margin. Then again because the people evaluating it have no reference point and reach for a larger margin to cover their own discomfort rather than the project's actual risk. Those are different things. A competent professional will tell me the second margin is prudence and that the distinction is academic when the money is real, and that argument is not weak.

The honest limit is that I cannot prove the distinction in advance. It resolves only after the building exists, at which point it has become a comparable and the question no longer applies to it. That is genuinely unsatisfying and I do not have a better answer.

The buildings on this list that had no useful precedent when they were evaluated are the ones I learned this on. Where a project reaches this point, the people to retain are a cost consultant who will price components rather than buildings, and a scheduler willing to state what they do not know.