What is the most defensible cost segregation methodology?

The detailed engineering approach from actual cost records is the most accurate and defensible methodology per the IRS Cost Segregation Audit Techniques Guide, because every reclassified dollar traces to a contractor payment or invoice. The detailed engineering cost estimate approach is nearly as strong for acquired buildings where cost records do not exist and components must be estimated from published cost data.

Does the IRS require a specific cost segregation method?

No. Neither the Code nor the regulations prescribe a methodology, and the Audit Techniques Guide says so explicitly. But the ATG also tells examiners that studies based on actual cost records or detailed engineering estimates deserve more deference, while rule-of-thumb and abbreviated residual approaches invite closer scrutiny and larger adjustments on exam.

Is a site visit required for a cost segregation study?

No rule requires one, but the Audit Techniques Guide lists physical inspection among the principal elements of a quality study. A site visit verifies that components claimed on paper actually exist, documents their use — which drives Section 1245 classification — and supports the study on exam. Studies performed entirely from a rent roll and a purchase agreement are materially weaker.

Are software or DIY cost segregation studies defensible?

Sometimes, for small and simple properties. Software studies typically apply statistical cost models rather than property-specific engineering, which the ATG treats as a modeling approach requiring careful review. For a small residential rental they may be proportionate; for a larger or specialized commercial property, the absence of engineering detail, inspection, and documented cost sourcing leaves the allocation hard to defend.

What does a quality cost segregation study cost?

Full engineered studies on commercial property commonly run from roughly $5,000 to $15,000-plus depending on size and complexity, residual or abbreviated studies less, and software-driven reports a few hundred to a couple of thousand dollars. The fee should be weighed against both the present-value benefit and the cost of defending a weak report on examination.

Cost segregation study quality: engineered, residual, and DIY approaches compared

A cost segregation study is only as good as the method behind it. The IRS's Cost Segregation Audit Techniques Guide (ATG) — the manual its own examiners use — recognizes six approaches and ranks them bluntly: studies built on actual cost records and detailed engineering are "the most methodical and accurate," while rule-of-thumb allocations "lack sufficient documentation" and merit little deference. No statute prescribes a method, so the market sells everything from $500 software reports to six-figure engineering engagements, and the buyer is left to judge whether a given report will survive contact with an examiner.

That judgment matters because the deduction is only provisionally yours. A study that reclassifies 30 percent of basis into short lives but cannot show where the numbers came from does not just risk losing the excess — it risks accuracy-related penalties and a soured exam.

Why methodology matters at all

A study's legal job is to allocate a lump-sum basis between Section 1250 real property (27.5- or 39-year) and Section 1245 personal property and 15-year land improvements, applying the case law that runs from Hospital Corporation of America back through the old investment tax credit rules. The classification questions are legal; the dollar amounts attached to each classification are engineering. Get the law right and the quantities wrong, and the study still fails.

The ATG instructs examiners to test both: whether the components claimed as 5-, 7-, or 15-year property qualify under the case law, and whether the costs assigned to them are supportable. Methodology is the second test. Pub 946 supplies the recovery periods once classification is settled; it does not tell you how many dollars belong in each bucket. That is the study's whole value, and the whole exam exposure.

What are the six approaches in the ATG?

The ATG describes six methodologies, in descending order of rigor:

The hierarchy, condensed.

Approach	How costs are determined	Typical use	ATG posture
Detailed engineering from actual cost records	Contractor payment applications, invoices, change orders traced component by component	New construction, major renovations	Most accurate and defensible
Detailed engineering cost estimate	Quantity take-offs priced from published cost data (e.g., estimating manuals), reconciled to total basis	Acquisitions with no cost records	Strong, if estimates are documented and reconciled
Survey / letter	Contractors asked to state what they charged for specific components	Supplement where records are partial	Acceptable for the costs it covers
Residual estimation	Short-life items costed directly; everything left is dumped into the building	Mid-market acquisitions	Faster, less accurate; the residual hides errors
Sampling / modeling	A model built from a sample of properties applied across a portfolio	Chains with many similar units	Acceptable if the sample is valid and reapplied honestly
Rule of thumb	Percentages by property type, no property-specific work	Marketing teasers, some cheap reports	Little to no evidentiary weight

The first two are "engineered" studies in market parlance. The last is what the ATG warns against: an allocation "based on a preparer's experience" with nothing behind it.

What does an engineered study actually involve?

A full detailed-engineering study on new construction starts from the complete cost record — the general contractor's schedule of values, pay applications, subcontractor invoices, and soft costs — and assigns each cost to a component, then each component to an asset class. Indirect costs (architect fees, general conditions, permits) are allocated across classes on a defensible ratio. On an acquisition, the cost-estimate variant substitutes a quantity take-off: the engineer measures or counts components from drawings and inspection, prices them from published estimating data, adjusts to the placed-in-service date and locale, and reconciles the total back to purchase price so the estimate cannot exceed actual basis.

The ATG's list of principal elements of a quality study is effectively a specification: preparation by someone with expertise and experience, a report describing the methodology, an inspection of the property, photographs, identification of the legal authority for each classification, reconciliation of allocated costs to total basis, and treatment of indirect costs. A report missing several of these is, by the government's own manual, not a quality study.

The site visit

No regulation requires an inspection, but its absence is the single most common weakness in cheap reports. Classification under Section 1245 frequently turns on use — whether electrical circuits serve equipment or general building load, whether plumbing serves a process or a restroom — and use is observed, not assumed. The visit also catches assets that were removed, never installed, or shared with another parcel. On exam, "our engineer walked the property on this date, and here are the photographs" is worth a great deal.

Sampling for portfolios

For an owner with forty near-identical drive-through restaurants, engineering each one is wasteful. The ATG accepts statistical sampling: engineer a valid sample fully, then extrapolate. The traps are non-representative samples (flagship stores studied, tired ones extrapolated), stale models applied to later-built units with different specifications, and extrapolation across genuinely different prototypes. A sampling study should document the sampling plan, not just the result.

What do residual and DIY studies deliver?

The residual approach costs out the short-life property directly — often from a walk-through and estimating data — and assigns everything remaining to the building. It is faster and cheaper, and the ATG accepts that it can produce reasonable results. Its structural flaw is asymmetry: every error in the short-life estimate flows silently into (or out of) the 39-year residual, and nobody ever audits the residual for reasonableness. Residual studies also tend to skip indirect-cost allocation, understating the short-life benefit or, done carelessly, overstating it.

Software-driven "DIY" studies — the taxpayer answers a questionnaire, an algorithm allocates — are essentially modeling approaches without a property-specific sample. For a $400,000 single-family rental, where an engineered fee would consume much of the benefit, a conservative software allocation may be a proportionate answer; the dollars at stake are small and the property is generic. For a $6 million medical office building, the same product is an invitation to exam trouble: no inspection, no cost sourcing, no component-level legal analysis, and an allocation the taxpayer cannot explain because the taxpayer did not make it.

What does each approach cost, and when is each defensible?

Cost versus stakes, illustratively. Fees vary by market and property; the pattern is what matters.

Study type	Typical fee	Sensible when	Hard to defend when
Detailed engineering (actual costs)	$8,000–$20,000+	New construction or renovation over roughly $1M; records exist	Rarely — this is the benchmark
Detailed engineering (estimate)	$5,000–$15,000	Acquisitions; look-back studies with a Form 3115	Estimates undocumented or unreconciled to basis
Residual	$3,000–$8,000	Mid-size acquisitions, straightforward property types	Complex or specialized buildings; heavy indirect costs
Sampling / modeling	Varies by portfolio	Large fleets of near-identical units	Sample unrepresentative or model applied beyond its scope
Software / DIY	$400–$2,500	Small residential rentals; benefit under ~$50K	Larger commercial property; anything specialized
Rule of thumb	~Free	Feasibility screening only	As support for an actual return position

Two calibration points. First, the exposure scales with the reclassification, not the fee — saving $5,000 on the study for a $10 million building protects nothing. Second, defensibility compounds with other filings: a look-back study implemented through a Form 3115 method change puts the study's numbers directly into a Section 481(a) adjustment, where sloppy work is easiest to see. If the study feeds a 3115, buy the engineered version.

What are the red flags in a cheap report?

No site visit and no photographs, or photographs that are obviously stock or from a listing.
Round numbers: 5-year property at exactly 20.0 percent of basis suggests a percentage was applied, not a property studied.
No reconciliation of the allocated components back to total depreciable basis — or components that sum to more than the basis.
No land allocation discussion. A study that segregates a purchased building without first carving out non-depreciable land has inflated every bucket.
No legal citations per component — just asset lists with lives, no authority for why the carpet or the parking lot qualifies.
Aggressive outliers without explanation: reclassification percentages far above the norms for the property type, unexplained.
Boilerplate identical across properties: the same "engineering narrative" for a warehouse and a hotel.
Contingent fees or "audit protection" sold separately — defense of the report should be part of the report.

A report with two or more of these should be priced as what it is: a deduction the taxpayer may have to give back.

When is the expensive study not worth it?

Neutrality cuts both ways. An engineered study is overkill — or the whole exercise is — when the depreciable basis is small, the holding period is short, the owner's losses would be suspended under the passive activity rules anyway, or a near-term sale would convert the acceleration into recapture. Those thresholds are the subject of when cost segregation doesn't make sense; the study-quality question only arises after the study-at-all question is answered yes. And on small properties where the answer is a marginal yes, a conservative residual or software study, honestly labeled and modestly claimed, can be the right-sized tool — provided the preparer's numbers would survive the taxpayer explaining them out loud.

The mechanics of what the study feeds — MACRS lives, conventions, and bonus eligibility and the interaction with 100 percent bonus depreciation — are covered separately. Whatever the method, the output lands on Form 4562 the same way; the difference is what happens if anyone ever asks how the numbers got there.