What an IRS Engineer Actually Looks For When Auditing a Cost Segregation Study

James C. Peacock spent 38.5 years at the IRS as a General Engineer. He was among the first IRS engineers to examine cost segregation studies. He contributed to the original Cost Segregation Audit Techniques Guide — developed 2000 to 2002, released publicly in 2004 — and every major update since, including the February 2025 revision. Before retiring in September 2025, he trained roughly 200 new-hire IRS engineers on cost segregation methodology.

In other words: he is the person who decided whether your study was good enough. I sat down with him for 90 minutes. This article is what I learned.

Cost Segregation Does Not Trigger Audits

Start here, because most investors have this exactly backward. The fear is that a large depreciation deduction waves a red flag and draws IRS attention. James addressed this directly.

"Cost seg studies don't get discovered until there is already an audit open," he told me. Audits are triggered by DIF scores (the IRS's statistical return-selection model) and specific IRS campaigns — not by the presence of a cost seg study. The study surfaces once an examiner is already inside your return.

He described the overall audit rate for business entities as very low — well under 1% annually. "Very, very, very low," he said. There is no published IRS data showing cost segregation studies raise that probability. The IRS's own audit guide actually says a well-prepared study reduces examination burden, not increases it.

"Two out of a thousand instead of one. It doubled! But it's still very, very, very low."

— James C. Peacock, former IRS Engineer SME

The marginal increase is statistically trivial. What matters is what happens when an audit does occur — and whether your study can survive it. That is the real question, and it is what the rest of this article addresses.

For a deeper look at audit probability and what actually moves the needle on risk, see our article on cost segregation audit risk.

The First Thing an IRS Engineer Checks: Land Allocation

When James opened a cost segregation study, the first thing he looked for was not the 5-year property schedule. It was not the methodology section. It was land.

Land is not depreciable. Every property purchase includes a land component, and that allocation must appear separately in the study. If a property was purchased for one million dollars and the cost segregation components — building, improvements, personal property — added up to one million dollars, there was an immediate problem.

"If a property is purchased for a million dollars and the study adds up to a million dollars, we go — where's the land?" James said. Missing land allocation is not a negotiating point. It is not a question that requires further analysis. It is an automatic adjustment.

This is one of the most common errors James saw, and one of the easiest for the IRS to catch. It requires no engineering judgment. Any examiner, engineer or not, can spot it by comparing the purchase price to the total of the classified components. If they do not reconcile to a number that leaves room for land, the study has a fundamental defect.

RS Means Codes: Vague Support = Immediate IDR

After land, James went to the back of the report. Specifically, he looked at the RS Means codes used to support the cost estimates for classified components.

RS Means is the industry-standard construction cost database. Proper cost segregation studies use RS Means codes to support their per-unit cost estimates — and those codes should be specific. James looked for 12-digit or 16-digit RS Means codes. A code that precise tells the IRS that the engineer looked at the actual assembly, the actual labor, the actual material type, and priced it accordingly for that property.

What he did not want to see: "RS Means mechanical" with no code number attached. That is not a citation. That is a category label. Studies that use category references instead of specific codes tell an IRS engineer that the estimator used averages and did not actually analyze the property. That generates an IDR immediately.

The same logic applies to square footage models — studies that calculate costs by multiplying square footage by an average cost per square foot. James was direct about what that signals: "They used averages and didn't look at the property." Square footage models generate more IDRs because they cannot be tied to actual construction components. The IRS wants to see property-specific analysis, and vague codes are evidence that none occurred.

The LUQ Framework: Large, Unusual, Questionable

Once past the initial checks, IRS engineers use the LUQ framework to decide which line items deserve closer scrutiny. LUQ stands for Large, Unusual, Questionable. Any component that qualifies on one or more of those dimensions gets examined.

The framework is deliberately open-ended. James gave a concrete example: a warehouse where 50% of the total cost is attributed to HVAC. That is a LUQ item. A warehouse is not a refrigerated storage facility. Standard warehouses do not have half their cost in climate control. When an engineer sees that, it flags immediately.

What percentage of HVAC cost is acceptable for a warehouse before it becomes a LUQ item? James said the IRS made a deliberate decision never to publish that answer.

"The IRS deliberately never published percentage thresholds — because as soon as we say X percent is acceptable, everyone's going to claim X minus 1 percent."

— James C. Peacock, former IRS Engineer SME

This is important for investors to understand. There is no published safe harbor percentage. A study that pushes classifications to the limit on every category is not playing a numbers game — it is accumulating LUQ flags across the entire document. The IRS engineer then has a long list of items to examine, each one a potential IDR.

"It's the Support, Not the Report"

James referenced a mantra he heard throughout his career from a colleague: "It's the support, not the report." This is the organizing principle of cost segregation audit defense. The report — the formatted PDF with schedules and conclusions — is not what the IRS is evaluating. The underlying support is.

What does support look like in practice? James described what he typically requested in the first or second IDR: the contractor's final application for payment and the plans and specifications. The final application for payment is the document that shows what the contractor actually billed for — line by line, by CSI division. Plans and specs confirm what was built and where specific systems were installed.

A study that classified a dedicated electrical panel as 5-year property, for example, needs to be supported by showing that the panel was dedicated exclusively to a specific piece of equipment. James referenced the Scott Paper case (covered in Chapter 8 of the ATG) as the legal basis for that classification. The study can make the claim. But the support has to be there: wiring diagrams, panel schedules, equipment specs.

When that support exists, the IDR gets answered cleanly and the audit moves on. When it does not, the IDR generates a follow-up IDR, and the examination extends.

The IDR Process: How an Examination Actually Works

The IRS does not audit a cost segregation study by reading the report and issuing a verdict. It works through a formal process of Information Document Requests. An IDR is a specific written request for documents or information, issued on an IRS form. The taxpayer responds, the examiner reviews the response, and either the issue is resolved or another IDR follows.

The first IDR in a cost segregation examination typically asks for two things: the purchase agreement and the cost segregation study itself. This is the starting point. The purchase agreement establishes what was purchased, at what price, with what allocations. The study establishes what classifications were made.

From there, the number of IDRs that follow is a direct function of study quality. James was explicit about this: "The least amount of IDRs, the easier the audit goes." A study with complete support, specific RS Means codes, clean land allocation, and well-documented LUQ items tends to generate few follow-up requests. A study with vague support generates a cascade.

If IDRs are not resolved satisfactorily, the process moves toward a draft report and eventually a proposed adjustment. A proposed adjustment means the IRS engineer has determined that a classification is incorrect and is recommending a change to the depreciation schedule. That triggers a formal response process, and potentially appeals.

James noted that IRS revenue agents — the generalist examiners who typically open and manage audits — do not have cost segregation expertise. For larger cases, engineers are brought in separately. "With most of the cases I saw, they were at the smallest, either on the large side of what they call small business self-employed division, and then LB&I." LB&I is Large Business and International — the IRS division that handles larger corporations and high-complexity returns.

The High-Rise Example: What Gets Automatically Adjusted

James described reviewing a cost segregation study for a high-rise building. The 5-year property (Section 1245) schedule looked reasonable on its face. But when he got to the 39-year structural components (Section 1250), something was wrong. There was no steel framework allocation. No concrete slabs. The structural skeleton of the building was either missing or buried in a category that did not identify it correctly.

"Tens of millions of dollars just gone. That was an automatic adjustment right there."

This is the mirror image of the over-claiming problem. Some studies are so focused on finding 5-year personal property that they shortchange the 1250 structural components — the elements that belong in the 39-year schedule. James made the point that AI-generated studies face a similar problem: "We would get cost seg studies where there was so much emphasis on getting the 1245 property estimated that the 1250 property was an afterthought."

A complete study accounts for all cost components. The IRS is not just checking whether you claimed too much in 5-year property. It is checking whether the entire schedule reconciles correctly to the purchase price and whether every major structural element is accounted for.

Court Cases That Shaped Classification Decisions

James discussed several court cases that directly inform how IRS engineers evaluate specific classifications. Understanding these cases helps investors know which property types are likely to face scrutiny — and what documentation is required to defend them.

Kitchen Cabinets: Amerisouth v. Commissioner (2012)

In residential properties, kitchen cabinets, counters, and sinks are a common target for 5-year personal property classification. The Tax Court addressed this directly in Amerisouth v. Commissioner (2012). The ruling: these items are 1250 property — either 27.5-year residential or 39-year non-residential — because they serve the operation and maintenance of the building.

To reclassify kitchen cabinets as 5-year Section 1245 personal property, a study must document two things: (a) the cabinets were actually removed from the property and (b) they were actually reused elsewhere or disposed of. Theoretical removability is not sufficient. The Whiteco test, as James described it, requires that personal property be movable and in fact moved — not just capable of being moved. The Whiteco case involved roadside billboards that were physically dismantled and relocated. That is the standard.

For a deeper look at how court cases like Amerisouth shape current classification practice, see our article on significant court cases in cost segregation.

Self-Storage Metal Partitions

Self-storage facilities are a common cost segregation property type. The metal partitions that divide individual storage units are often classified as 5-year personal property. James confirmed this can be defensible — but only if the owner can prove the partitions are actually removed and moved between units or facilities. If the partitions are screwed into walls and have never been relocated, the classification does not hold. Documentation of actual removal and relocation is required.

Short-Term Rentals and the 39-Year Rule

James flagged a classification issue that affects many real estate investors pursuing the short-term rental strategy. Properties with average guest stays under 30 days are classified as 39-year non-residential property — not 27.5-year residential. This matters for cost segregation because the depreciation schedules and available bonus depreciation categories differ.

The favorable news for STR investors: interior improvements to a short-term rental qualify as Qualified Improvement Property (QIP) — a 15-year category eligible for 100% bonus depreciation. Land improvements like pools, landscaping, and parking also qualify as 15-year property regardless of the rental type. These categories still capture significant first-year deductions even under the 39-year building classification.

Contingency Fees and What They Signal to Examiners

The Cost Segregation ATG instructs examiners to "closely scrutinize" studies prepared on a contingency fee basis — where the provider's compensation is tied to the size of the deductions claimed. James said the engagement agreement was typically requested in the first IDR.

He described the concern plainly: when a provider promises to "save you X and take a percentage," the incentive structure raises questions about whether the classifications reflect genuine engineering analysis or optimistic assumptions designed to maximize the fee. That is not an automatic disqualification, but it puts the study under heavier scrutiny from the start.

This is one of several red flags that indicate whether a provider is oriented toward defensibility or toward maximizing claimed deductions. The two are not always the same thing.

What About AI-Generated Cost Seg Studies?

AI-generated cost segregation studies — reports produced without a physical property inspection — are increasingly being marketed as low-cost alternatives to engineering-based studies. James was direct about how the IRS views them.

"It goes against every IRS rule," he said, because the IRS has consistently required that cost segregation studies be based on physical inspection of the property and review of construction documentation. A report generated from public data or satellite imagery without an engineer walking the property cannot establish the property-specific facts that survive IDR scrutiny.

That said, James saw a legitimate use for AI as a completeness check. The problem he described — studies where the 1250 structural components were an afterthought — is exactly the kind of omission an AI tool could flag. Using AI to verify that all categories are accounted for, without using it to make classification decisions, is a reasonable application. The classification work still requires an engineer and a site visit.

The Adjustment Factor Red Flag

James described one more common issue that triggers engineer scrutiny: the adjustment factor. In cost segregation, engineers estimate the cost of each component using construction cost data, then apply a factor to reconcile those estimates to the actual purchase price of the property.

A small adjustment factor — say, 1.05 — is unremarkable. It means the estimated costs were close to actual. But when the estimated costs need to be multiplied by 10 to match the purchase price, something is wrong. Either major cost components are missing from the estimate, or the methodology was not property-specific to begin with.

"There's usually something missing in the estimate if the factors are that bad," James said. A large adjustment factor tells an IRS engineer that the underlying estimate was not built from the property — it was built from averages and then scaled up to fit. That is a square footage model problem in disguise, and it generates the same IDR response.

What Makes a Study Survive Examination

After 38.5 years examining cost segregation studies, James's summary is simple. The studies that survived examination shared specific characteristics:

• 1 Clean land allocation. The purchase price reconciliation accounts for land separately. There is no ambiguity about what portion of the price is non-depreciable.
• 2 Specific RS Means codes. Every cost estimate ties to a 12-digit or 16-digit code. The engineer can trace every line item to a specific assembly in the database.
• 3 Complete structural accounting. The 1250 schedule accounts for all major structural components — foundation, frame, mechanical, electrical rough-in — with no unexplained gaps.
• 4 Proportionate LUQ items. No single category dominates in a way that is inconsistent with the property type. HVAC for a warehouse does not approach 50% of cost.
• 5 Supporting documentation on file. Contractor's final application for payment, plans and specs, and any property-specific evidence for unusual classifications are retained and available for IDR response.
• 6 Reasonable adjustment factor. The ratio between estimated component costs and actual purchase price is in a normal range, indicating the estimate was built from the property rather than from averages.

"It's the support, not the report."

— IRS colleague's mantra, cited by James C. Peacock throughout his 38-year career

The report is what you deliver to your CPA. The support is what survives an examination. The two need to be built together, not as an afterthought.

A Note on New IRS Engineering Capacity

Before retiring, James trained roughly 200 new-hire IRS engineers on cost segregation and 179D. The program was active through 2025. He noted that even as AI tools begin to assist IRS agents with return selection, the underlying examination capacity for cost seg studies has been replenished. "The audit rate might not go down... if you get caught in a trap, it's still a trap."

The standards have not softened. The ATG has been updated through February 2025. The engineers being trained are using the same evaluation framework James helped build over two decades. A defensible study today requires the same quality it required in 2004.

What This Means for Real Estate Investors

The practical takeaway is not complicated. Cost segregation is a legitimate, IRS-recognized strategy. The IRS published an entire guide explaining how to evaluate studies correctly. The risk is not in using cost segregation. The risk is in choosing a provider who optimizes for the largest possible deduction rather than the most defensible one.

A study with specific RS Means codes, clean land allocation, proportionate categories, and retained supporting documentation will generate minimal IDRs if examined. A study built on averages, vague citations, and aggressive categories will generate a long list — and the audit will be long.

Before choosing a provider, ask how they handle RS Means citations in their reports. Ask what supporting documentation they retain after delivery. Ask whether they have performed a physical site inspection. Those three questions reveal more about study quality than any marketing claim.