Cost Estimating Without Comparables
Cost estimating without comparables, the practical core of what Michael Stern calls the economics of building where precedent is scarce, rests on one rule that the estimating profession took decades to write down: the quality of an estimate is set by how much of the project has been defined when the estimate is made, not by the effort spent, the software used or the time taken. That principle is the foundation of the AACE classification system, first published in 1997 as a recommended practice, which sorts estimates into five classes from Class 5, the least defined, to Class 1, the most.1 The numbered classes replaced older labels that had drifted in meaning across industries, terms like order of magnitude, budget, definitive and detailed, and gave owners, engineers and contractors a shared way to say how reliable a number is expected to be.1
The rule matters most where no twin of the building exists. For a conventional project the easiest route to an early number is to find a finished building that resembles it and scale. For a project without that twin, the estimator has to work from the parts, and the history of the methods explains why.
Two early methods and the one that fails on a new kind of building
The estimating literature divides early work into two families. A cost management software firm compares them: analogous estimating takes the actual cost or duration of a previous, similar project as the basis for a new one and relies on expert judgment, while parametric estimating applies statistical relationships, cost per square meter or labor hours per weld, drawn from a reliable historical database.2 Each has a known weakness. Analogous work is subjective and hard to adjust for large differences in scope or technology; parametric work depends on data quality and is, in the firm’s words, less flexible on highly unique or innovative projects where no comparable data exists.2 The same source notes that analogous estimating is nonetheless the method that fits unique or first of a kind projects at the conceptual stage, and recommends a hybrid: screen with analogy, refine with parameters as definition grows.2
Analogous estimating in six moves
- Define the scopeDeliverables, quality requirements, exclusions and constraints are written down, often as a work breakdown structure.
- Find the past projectsExpert judgment picks the completed jobs whose scope and complexity come closest.
- Collect the dataCosts are gathered by element: site preparation, foundation and structure, materials, labor, equipment, finishes, contingency and overheads.
- NormalizeOld costs are adjusted for inflation and market change between then and now.
- Find the driversThe parameters with the largest effect on cost, typically materials, labor rates and size, are identified.
- ScaleFactors for size and complexity bring the old numbers onto the new project.
A project management software publisher walks through the analogous method in six steps and in doing so shows where it strains. After defining scope and shortlisting similar past projects, the estimator collects historical cost by element, normalizes it for inflation and market change, identifies the parameters that drive cost most, and applies scaling factors for size and complexity.3 Every one of those steps assumes a shortlist exists. The publisher’s own caution is that the method relies heavily on the assumption that the current project will follow the pattern of past ones, which may not hold, and that the more data available the better.3 When the shortlist is empty the method does not refuse to run. It produces a number from the nearest wrong precedent, and nothing in the arithmetic flags how far off the precedent is. That is the gap decomposition fills.
Decomposition: pricing the parts that do have a history
An unprecedented building is almost always made of precedented parts. A concrete core of a given strength, a curtain wall of a given span, an excavation in a known soil, a crane of a known class: each has been priced thousands of times. The databases exist to do exactly this. Autodesk’s guide to RSMeans describes a database of more than 90,000 line items covering materials, labor, transportation and storage, researched for roughly 25,000 hours a year and adjustable for more than 1,000 locations in North America.5 It offers two levels. Unit prices give the user granular control and suit unique projects; assembly prices bundle components into systems and suit larger projects at design development, where they are used to test whether a design is feasible and to compare construction types.5
The square foot number, and why it fails here
The square foot method sits at the other end of the scale and the publisher of the data is candid about what it can and cannot do. RSMeans’ own guide to square foot estimating describes multiplying total area by a predetermined unit cost chosen for building category, type, wall and framing, size and story height, then adjusting for site conditions, design changes, material selections and market movement.4 The guide treats the result as a ballpark figure, a starting point rather than a definitive cost, and warns that location, finishes, unique design features and local regulation introduce variability the method does not see.4 On a building with an unusual proportion or envelope, the unique design features are the building, so the square foot number is the least useful figure in the room and the component list is the most.
Two early methods and where each breaks
| Analogous | Parametric | |
|---|---|---|
| Basis | The actual cost of a similar past project, scaled | Statistical relationships between historical data and measurable variables |
| What it needs | Expert judgment and at least one close precedent | A clean, structured database and a calibrated model |
| Precision | Lower, and dependent on who performs it | Higher when the data behind it is sound |
| Weak point | Hard to adjust for large differences in scope or technology | Less flexible on a highly unique project with no comparable data |
Decomposition also changes where uncertainty lives. A single top down figure carries one unexamined error. A list of priced components carries many small ones, each of which can be argued about with a trade contractor or a supplier who has priced that component before. The pillar on underwriting takes up what happens when those small errors move together.
Time, and why the answer is a range
Two adjustments turn a priced list into an estimate. The first is time. A preconstruction glossary explains escalation as the percentage applied to an estimate for the expected rise in labor, material and equipment prices between the estimate date and the date the work is bought, compounded to the construction midpoint, the point by which roughly half the money has been spent.6 Rates come from published indices, producer price indexes and cost consultants’ forecasts, and the glossary lists the common failures: escalating only to the start of construction, applying one national rate to a market moving differently, and failing to update when the start date slips.6 The cost indices entry covers where those rates come from and what they miss.
The second adjustment is for scope that has not yet been drawn. A firm of architects in Iowa describes design contingency as a slice of the building cost set aside while the architect explores options, falling as unknowns are resolved: in its practice 20 percent at predesign, dropping five points at each later phase until it reaches 5 percent when construction documents are done.7 The firm lists what remains unknown early on, from finishes to ground conditions to where prices are heading, and separates contingency from escalation, which it treats as its own line item predicted to mid construction.7 The contingency entry follows those numbers through a project’s life.
What a square foot model asks for
- new or renovation, commercial or residential
- Category
- apartment building, hospital, school and so on
- Type
- exterior wall construction and framing type
- Wall and frame
- area, number of stories, story height, fee assumptions
- Parameters
- elevators, fire suppression and other optional systems
- Additives
Put the pieces together and the honest output of early estimating on a building of this kind is a range, not a point. The classification system says so by assigning a class rather than a precision; the escalation line says so by compounding an assumed rate to an assumed midpoint; the design contingency says so by carrying a percentage for drawings that do not exist yet. A single figure on the cover sheet is a convenience for the reader, and the people who know what it conceals are the cost consultant who built the component list and the scheduler who fixed the midpoint it was escalated to. Those are the two people a reader of this page would want in the room before the number is repeated.
