September 7, 2026 · 5 min read
What actually makes a cost segregation study defensible
People assume cost segregation is a grey area. It is not. The IRS publishes an audit techniques guide on the subject — a document written for its own examiners, describing how these studies work and what separates a good one from a bad one. It is public, and it is the closest thing the field has to a rulebook.
Reading it, a pattern emerges quickly. Almost nothing in it is about whether a particular item belongs in a five-year class or a fifteen-year one. Most of it is about whether the study shows its work.
Where each number came from
A study that says a property contains $4,000 of electrical fittings has said very little. A study that says the same thing and can point to the specific items, the source of the unit costs, and the reasoning behind the classification has said something checkable.
This is the single most useful thing to understand about the whole exercise. The strength of a study is not the size of the number it produces. It is whether somebody who did not prepare it can follow it back to its sources and arrive at the same place.
Costs that have a source
Construction cost books exist precisely so that estimators are not inventing prices. They publish unit costs by item, by region, by edition. A study line that cites the book, the edition and the specific line it was priced from can be verified by anybody holding the same book. A study line with a plausible round number and no reference cannot be verified by anybody at all.
- Which cost book, and which edition — costs move year to year, so a citation without an edition is half a citation.
- The specific line, not just the category. "Appliances" is not a source; a line identifier is.
- Where regional adjustment was applied, and by how much, kept separate from the national figure.
The judgement calls, named out loud
Every study contains at least one significant judgement call, and the largest is usually invisible to the person receiving it. A cost book tells you what it would cost to build the property today. You paid a different amount, on a different day. Reconciling those two figures — deciding how the difference is spread across the components — is a choice, and different defensible choices produce materially different results from identical inputs.
A study that names the reconciliation method it used has disclosed something a reader needs. A study that quietly applies one and presents the result as arithmetic has not.
What this means when you are choosing
If you are comparing providers, the useful questions are not about price or turnaround. Ask what evidence sits behind each line, whether you will be given it, whether the costs cite a published source, and what happens to the items the study could not price — printed as outstanding, or quietly dropped from the schedule?