How a Cost Segregation Study Works
The steps of a typical study, from gathering cost records and inspecting the property to classifying components, reconciling costs, and handing a report to your CPA. Publication 5653 describes six approaches; using actual cost records is the most accurate. The guide is examiner guidance, not a required format.
Educational information, not tax advice.
Updated 2026-10-14
Key takeaways
- Studies start from documents: closing statements, construction contracts, pay applications, invoices, and drawings.
- Each component is assigned a cost and a recovery class, and the total must reconcile to the property's basis.
- The IRS audit guide describes six study approaches; using actual cost records is the most accurate.
- A field inspection is recommended for quality studies because classification turns on attachment, function, and use.
- The guide is examiner guidance, not a required format and not legal authority.
A cost segregation study answers two questions for every part of a property: what did it cost, and which recovery class does it belong in? How a study gets those answers depends on what records exist.
The IRS Cost Segregation Audit Techniques Guide describes six approaches. The most detailed uses actual construction cost records; the least reliable applies rule-of-thumb percentages with little documentation. In between are engineering cost estimates, contractor surveys, residual estimation, and sampling models for near-identical facilities.
The Service has not established requirements or standards for preparing studies. Substantiation still matters. The guide treats actual costs as more accurate than estimates, and it treats a uniform step-up that forces estimates to match a purchase price as unacceptable.
Publication 5653 also lists thirteen elements of a quality study and nine elements of a quality report. Those lists are what examiners are told to look for. They are not a certification program, and they do not make any report “IRS-approved.”
CostSeg SI’s proposed Super Intelligence workflow is an explanatory product description of how extraction, classification, and review could be organized. It is not an operational claim that those steps are running on live studies. Educational information, not tax advice.
1. Gather documents
For new construction, the best records are the construction contract, AIA-style pay applications, change orders, invoices, and as-built drawings. For a purchase, the closing statement, purchase agreement, any appraisal, and available drawings are the starting point.
Indirect costs travel with the project: design fees, permits, construction-period interest, and similar items that belong in basis. A quality study allocates those costs across components instead of leaving them in a single “soft cost” bucket attached only to the building.
Missing records do not stop a study, but they change the method. When contemporaneous contractor costs are gone, the work shifts to documented estimates and a closer inspection of what is actually installed.
2. Inspect the property
The audit guide recommends a site inspection for all quality studies. Photographs and notes record what is actually installed, how it is attached, and what it serves — facts that drive classification.
Drawings show design intent. The inspection shows what was built. Carpet that is tacked rather than glued, a circuit that serves one machine rather than general outlets, and lighting that is decorative rather than the room’s primary source are facts that rarely appear on a closing statement.
For new construction, before-and-after photographs of site work help support how grading and landscaping were split among land, the building, and depreciable land improvements.
3. Identify, quantify, and price components
Components are taken off the drawings and inspection notes and priced, either from actual contractor costs or from documented estimates using published cost data. Indirect costs such as architect fees and permits are allocated across components.
The guide names contractor data and reliable published sources, including R.S. Means and Marshall Valuation Service, as examples of estimate sources. Those databases are estimating tools. They are not tax authority, and an estimate still has to be adjusted for location, date, and the quantities actually observed.
Mixing techniques — detailed pricing for short-lived items and a residual for the building — can tilt the result toward shorter lives. The guide treats that mix as a weakness.
4. Classify and reconcile
Each component is placed in a class — for example 5-year personal property, 15-year land improvement, or 27.5- or 39-year building — with the legal reasoning. The audit guide stresses that total allocated costs should reconcile to total actual costs, and that forcing estimates to match a purchase price with a pro-rata step-up is not acceptable.
Classification cites the regulations and cases that apply to the facts, not a vendor name or a local fixture statute. When an industry matrix in the audit guide is used, the report should say so and should not export that row to a different industry.
Differences between estimated replacement cost and the price paid need an explanation, such as obsolescence or intangibles in the price. Scaling every line by the same factor to force a match is the method the guide rejects.
5. Report and hand off
The report typically includes a narrative, the methodology, a schedule of assets, cost schedules, assumptions, and exhibits. Your CPA then uses it to set up depreciation, and decides on elections such as bonus depreciation. For property already in service, the CPA usually implements the study through an accounting method change on Form 3115.
A current-year placement usually needs no Form 3115 because the classification appears on the first return. A look-back study of a building already depreciated as one asset is generally a method change. Those filing decisions belong to the taxpayer’s CPA.
Sources
- 1IRS Publication 5653 — Cost Segregation Audit Techniques Guide (rev. 2-2025)(opens in a new tab)
Written for IRS examiners. The guide states it is not an official pronouncement of law and cannot be cited as authority. It predates the July 2025 law.
- 2
- 3
- 4