Most disallowed cost segregation studies do not fail because the law is unclear — they fail because something obvious is missing or wrong. Five patterns account for the bulk of examiner adjustments. Each maps directly to a section of the IRS Audit Techniques Guide (Pub 5653, the 2025 edition).
The fastest path to disallowance. The ATG names site inspection as a primary quality indicator (Pub 5653, §2.02). A study with no photographs, no take-offs, and no record of a walkthrough is making claims about a property the preparer has not seen.
The IRS does not require an in-person visit for every study. What it requires is documentation that the preparer actually examined the property — through photographs of the components being classified, dimensioned floorplans, fixture counts, or a structured photo manifest. A residential workflow built around investor-supplied photos and floorplan documents satisfies the requirement when the manifest is comprehensive and time-stamped. A "desk study" generated from a Zillow listing does not.
The failure mode looks the same in both cases: the examiner asks "where are the photos of the components you reclassified" and the preparer cannot produce them. Once that conversation happens, every classification in the study is now an unsupported assertion. Examiners typically resolve the entire study by reverting all components to 27.5- or 39-year MACRS — eliminating the year-one benefit and triggering interest and accuracy-related penalties under §6662.
"We used 20% land" is the second-most-common reason studies fail. The ATG instructs examiners to scrutinize the land allocation first because the impact is large and the documentation requirement is concrete (Pub 5653, §3.04).
The defensible methods, in order of strength: (1) the county tax assessor's land-to-improvement ratio at the time of purchase, (2) a contemporaneous appraisal, (3) a current appraisal with statistical adjustment, (4) Form 8825 historical allocations, and finally (5) a published metro-level land ratio applied to a price-per-square-foot adjusted basis. Methods 4 and 5 are weaker but defensible when documented; methods 1 and 2 are the standard.
What fails: any flat percentage applied without sourcing. "We used 20% because that's typical" is not a methodology — it is a claim about the property that the examiner can falsify in three minutes by pulling the assessor card. When the actual assessor ratio shows 32% land, the examiner re-runs the entire study at the corrected ratio. Year-one bonus depreciation drops by whatever proportion the land was understated — frequently $30,000 to $80,000 of disallowed deduction on a single-family rental.
For investors who suspect the assessor's ratio is unrealistic in their market — common in coastal California, parts of New York, and dense urban infill — the right answer is a damped statistical override with explicit reliability-gate logic, not a different rule of thumb. Document the reasoning. The ATG accepts overrides that are sourced and reasoned; it rejects overrides that are convenient.
RSMeans 2024 (or current year) with local jurisdiction adjustment is the standard the ATG names directly, alongside Marshall & Swift (Pub 5653, §4.03). A study built on internal spreadsheet averages, "national" cost data, or stale RSMeans editions has no replicable source for the examiner to verify.
RSMeans publishes ~970 city cost indices that adjust national base costs to local labor and material rates. A 5-ton residential HVAC unit installed in San Francisco genuinely costs ~1.4× what the same unit costs in Mobile, Alabama. Studies that ignore this are wrong on the absolute dollars, and the examiner can prove it with the same database the study should have used.
The other common failure: using last year's RSMeans (or worse, a 2018 edition) when costs have moved 20-40% since. Construction costs are not stable. The Bureau of Labor Statistics PPI for construction inputs has compounded materially since 2020. A 2018 RSMeans figure applied to a 2025 placed-in-service property is a documented error. Examiners increasingly verify the cost-data vintage in the study and adjust accordingly.
Marshall & Swift, ENR Building Cost Index, and contractor invoices for the specific property are also acceptable. Generic "industry averages" without a published source are not.
Component-to-MACRS-class mapping is not arbitrary. The framework is Whiteco Industries v. Commissioner, 65 T.C. 664 (1975) — the six-factor permanence test — extended to building components by Hospital Corp. of America v. Commissioner, 109 T.C. 21 (1997). Asset class tables come from Rev. Proc. 87-56 (now reproduced in Pub 946 Table B). A study that assigns components to 5-year, 7-year, or 15-year property without applying the Whiteco factors and citing the Rev. Proc. 87-56 asset class is making unsupported claims.
The Whiteco factors, in plain language: (1) is the component movable, (2) is it designed to be moved, (3) how is it attached, (4) what is its expected economic life, (5) is it permanent in the structural sense, (6) what damage results from removal? A bathroom vanity in a residential rental scores high on movability and low on structural permanence — it classifies as 5-year personal property. The wall behind it scores high on permanence — it stays in 27.5-year. Carpet, cabinetry, trim, decorative lighting, appliances: 5-year. Land improvements (landscaping, fencing, paving, exterior signage, exterior lighting): 15-year. Site utilities, foundations, framing, roofing structure, HVAC ducting integral to the building: 27.5- or 39-year.
What fails: a study that simply lists "carpet — 5-year" without articulating the Whiteco logic, or that classifies decorative lighting as 5-year while putting equivalent fixtures into 39-year on a different page. Inconsistency within the same study is a flag the examiner notices in the first read-through. The fix is documentation: a brief reasoning paragraph for each component class, citing the Whiteco factor that controls and the Rev. Proc. 87-56 asset class that maps to the chosen MACRS life.
For property types that are case-law unsettled (med-office tenant build-out, restaurant kitchen equipment versus structural plumbing, MRI-shielded radiology rooms), the study needs to disclose the ambiguity and pick a defensible lane. Hospital Corp of America is the controlling case for healthcare facilities; AmeriSouth XXXII Ltd v. Commissioner, T.C. Memo 2012-67 limits aggressive multifamily classifications. The study should cite these where they apply.
The arithmetic check the examiner runs in the first five minutes. Total reclassified components plus residual real property must exactly equal the depreciable basis: purchase price minus land minus any capitalized acquisition costs. When the totals don't tie, the rest of the study is provisional.
The reconciliation has three line items the examiner verifies against the closing statement (HUD-1 or ALTA): (1) purchase price, (2) capitalized closing costs (title insurance, recording fees, transfer taxes, certain loan costs that capitalize rather than amortize), and (3) the land allocation derived in section 02 above. The remainder — the depreciable basis — must equal the sum of every component in the cost segregation tables.
Off-by-thousands errors are common in studies built from spreadsheets. Component totals add up to $487,200 against a depreciable basis of $492,800. Where did the missing $5,600 go? The study cannot answer. The examiner now has a wedge: either the components are understated (which implies the 27.5-year residual is overstated, hurting year-one benefit) or the basis is misstated (which questions the whole foundation). Either way, the study comes back for revision.
The fix is a single reconciliation table at the front of the report. Purchase price, line-item closing-cost adjustments, land allocation with method cited, depreciable basis. Then a one-page summary showing the sum of all class-life components matching that basis to the dollar. A study that does this proactively rarely sees the issue come up at audit. A study that doesn't is making the examiner's job into the examiner's argument.
Before you sign off on a cost segregation study you received, run this checklist. If any answer is "no" or "I don't see it," send the study back for revision before filing.
Cost Seg Smart's standard residential and commercial reports include all five elements above by default — site-visit photo manifest, county-assessor land allocation with statistical override logic, RSMeans 2024 with city cost indices, Rev. Proc. 87-56 asset-class citations on every component, and a closing-statement reconciliation table at the front. Browse a real 40-page sample report and a separate methodology document.
View the sample report → Read the methodology →