Underwriting a Building With No Comparable

Underwriting a building with no comparable, the question at the center of the last of Michael Stern’s five subjects, the economics of rare buildings, is how a project is evaluated when every method of evaluation assumes a similar project has already been built and observed. A project controls software firm’s guide to early estimating describes the analogous method, which takes the cost of a past project of similar scope as the basis for a new one, as quick, dependent on expert judgment and, by its own account, a poor fit where scope, technology or market conditions diverge sharply from the precedent.² This page is about what replaces the relative: pricing the parts, treating the result as a class and not as a single number, watching how the parts move together, and choosing the order in which the unknowns are resolved.

What a precedent supplies, and what disappears with it

A comparable building gives an evaluator two things. The obvious one is a number; the other is a record of how far similar numbers have been wrong before, and that second gift is what an unprecedented project loses. The AACE classes exist to make that loss visible. They tie each estimate to the maturity of the project definition behind it, from Class 5 at the sketchiest end to Class 1 at the fullest, on the principle that definition rather than effort, software or time determines how good an estimate can be.¹ A Class 5 estimate of a conventional building and a Class 5 estimate of an unusual one carry the same label. Behind the second there is no body of finished buildings to say how wide the band around the number should be, so the band has to be built from something else.

Figure

A point estimate against a classed estimate

	The older labels	The AACE classes
Names	Loose terms such as budget estimate or detailed estimate	Class 5 through Class 1
What fixed the label	Habit, which varied by industry and by firm	The maturity of the project definition when the estimate was made
What it did not depend on	Nothing was stated	Effort spent, software used or time taken
Who could read it	Those inside one organisation	Owners, designers and builders sharing one vocabulary

Pricing the parts

That something else is the components. A building without precedent is assembled from pieces that have plenty: a core of a known concrete strength, a dig through a familiar soil, a hoist of a familiar size, trades whose rates are published. Autodesk’s description of RSMeans gives the scale of the record available for such pieces: more than 90,000 priced items, adjustable for over a thousand locations, offered as single units, which the guide says suit unique projects, and as assemblies, whole systems priced as one.³ Pricing by component moves the uncertainty from one place to many. One figure handed down from the top conceals one big mistake; a priced breakdown spreads the risk across many small mistakes, each open to argument by a subcontractor or supplier with experience of that piece. The estimating entry lays out the methods.

A class, not a number

A priced breakdown still adds up to a total, and the temptation is to read the total as the answer. A preconstruction glossary’s treatment of contingency explains why that reading is wrong even on an ordinary job: an estimate carries several contingencies at once: one for design, covering scope the drawings have not yet shown; one for construction, covering coordination misses and small surprises on site; and one held by the owner beyond the construction contract, each set by risk and expected to shrink as design matures, so that an estimate omitting them runs low rather than gaining precision.⁴ The glossary’s list of common failures reads as a description of a band mistaken for a point: the same percentage kept through every phase, risk counted twice because an allowance already covers it, contingency spent on work the estimator overlooked.⁴ The honest total is the component sum, the contingencies that belong to it and a stated class. The page on contingency traces each of those through a project.

When the parts move together

Decomposition has one failure of its own, and it is the most important idea here. The parts of a difficult building do not go wrong independently. The approvals that run long are the same approvals that push the start date; a later start meets a different market; a different market changes which trades are available and at what rate. A construction economist’s inflation note states the mechanism from the side of the market: any project delayed or put on hold must have its inflation worked out again to the new mid construction date, and how busy a region’s builders are foretells that inflation better than most things, since firms short of work bid low and firms with full books bid high.⁶ Time and price are one risk seen from two sides.

Figure

How a delay reprices a building

1. 01
   Budget at current costAn estimate is prepared from prices known today.
2. 02
   Carry inflation forwardHalf the spending falls before mid construction and half after, so the factor is applied to that date.
3. 03
   The date movesA hold or a delay shifts that date, and the inflation has to be recalculated from the new one.
4. 04
   The market moves with itLocal activity predicts inflation better than most indicators, so a later start meets different bids, not merely later ones.

A cost consultancy’s account of Monte Carlo simulation describes the tool built for this: a risk register in which each risk carries a sum and a chance of occurring, is shaped into a distribution and is sampled thousands of times, so that the output is a spread of outcomes with their odds, not a total, with the caveat that the result is only as good as its inputs.⁵ A register that treats the approval risk and the market risk as separate draws will understate the total spread, and understate it in the comfortable direction. One scheduling practitioner calls interfering float the slack one activity shares with those behind it, so that spending it narrows everyone else’s room, and watches how quickly float disappears as the earliest sign that the whole program is becoming critical.⁷ Shared float is correlation written on a calendar.

The order in which answers are bought

On a routine project the order in which uncertainties are resolved barely matters, because none is large. On an unprecedented one the order is most of the discipline, because some questions can only be answered by spending money that is gone if the answer is bad. An engineer’s article in STRUCTURE magazine on peer review wants the reviewer in the room during concept design, while the frame type and its layout are still open, because advice at that point alters the design at little cost, whereas a reviewer arriving late may send the team back to redesign; run well, he says, the exercise adds little to the budget.⁹ The value engineering standard says the same about function. ASTM E1699 asks for a value planning exercise at predesign, before drawings exist, so that owner, designer and builder agree what the project is for in one vocabulary.¹⁰ The rule that falls out is easy to say and hard to follow: resolve the uncertainty that could end the project ahead of every other, and among those, the cheapest to resolve goes ahead of the rest.

Building the unknown part twice

The clearest purchase of an early answer is a mockup. A magazine for the glazing trade explains that standard wall systems carry a history of satisfactory performance and seldom need testing for one job, but that an unproven design with new features may need a preconstruction mockup as the only way to verify it, and that the specimen must be a faithful full size piece of the actual design.⁸

Figure

Choosing what to build twice

1. 01
   A single bay, one story highToo small to stand for the wall.
2. 02
   Two stories by two baysThe customary specimen, taking in a stack joint and the seams between units.
3. 03
   A low corner addedAn improvement, since two conditions meet there.
4. 04
   A top cornerThe most telling and the costliest, because wind loads peak there.

The magazine then prices the choice: the common specimen covers two stories and two bays; adding a corner from low on the building improves it; a corner taken from the top, where the wind loads peak, is the most informative and the costliest; and a specimen of one story and one bay tells too little.⁸ The part of the building that has never been built is built once in a laboratory, and the money spent there is accepted as gone in exchange for an answer before fabrication. The entry on bid dispersion describes what that answer does to the prices that follow.

The margin and the second margin

An unusual building tends to carry two margins for one uncertainty. The first is the one the methods above produce: a class, a set of contingencies, a register with correlated draws. The second is added by people who have no reference point for the building and widen the margin to cover their own unfamiliarity rather than the project’s risk. The glossary’s warning against a percentage that never changes and against counting a risk twice is the professional form of the same point: a margin should be traceable to a named risk and should fall when that risk is retired.⁴ Whether the second margin is prudence or penalty cannot be settled before the building exists, since once it exists it becomes a comparable. What can be done beforehand is to keep the two margins apart on the page.

Who holds the question

The entry on indices covers what the published series track and what they miss; the schedule entry treats time as the twin of money; value engineering and independent review cover the two disciplines that keep a saving from becoming a defect; and skyscraper economics traces the height premium. On a project that has reached the point this page describes, the question is held by a cost consultant who prices parts and declines to price the whole, and by a scheduler who marks which durations on the program are guesses.