Sampling or Modeling Approach: When the ATG Supports Using a Sample

A cost segregation study does not always have to examine every property individually. The 2025 IRS Cost Segregation Audit Technique Guide recognizes sampling and modeling for portfolios containing multiple facilities that are nearly identical in construction, appearance, and use. The approach can reduce the resources and cost required to analyze an entire population. But the value of the result depends on whether the sampled properties actually represent the population. The ATG specifically warns that sampling may not be statistically valid and that even apparently similar facilities can differ because of building codes, geographic location, materials, and labor costs.

Wondering how much tax you could save with cost segregation? Let’s find out. Schedule a Cost Segregation Strategy Call today.

Key Takeaways

• The Sampling or Modeling Approach uses sampled properties to develop a model that can be applied to a broader population of substantially similar facilities.
• The process generally involves defining strata, sampling properties within each stratum, developing a model, and applying the model to the population.
• Sampling can reduce study resources, but statistical validity and sampling error can materially affect the reliability of the result.
• The approach is particularly relevant to portfolios such as retail and food store operations with repeated facility types.
• A quality study should define the population, explain the stratification, identify the sample size, and address sampling error.
• A small difference in a sampled allocation can become a much larger portfolio-wide difference when the result is projected across many properties.
• A sample is only as useful as the population it actually represents.

What Is the Sampling or Modeling Approach?

The Sampling or Modeling Approach is one of the six common approaches identified by the 2025 ATG.

The approach uses a created model or template to analyze multiple facilities that are nearly identical in construction, appearance, and use. The ATG gives fast food chains and retail outlets as examples.

The reason for using sampling is practical.

Studying every property individually can require substantial time and resources when a taxpayer owns a large number of substantially similar facilities.

Instead, engineers can study a representative sample and use the results to develop a model for the broader population.

The ATG describes this as a way to minimize the resources and costs associated with conducting studies on every property.

But the model only works if the properties included in the population are genuinely comparable.

A collection of buildings can look similar from the outside while having meaningful differences in construction, materials, systems, location, or use.

That is why the first engineering question is not:

“How many properties can we sample?”

It is:

“What properties actually belong in the same population?”

That question drives the rest of the methodology.

How Does Sampling or Modeling Work?

The ATG describes a typical sequence for using the Sampling or Modeling Approach.

1. Stratify the properties by facility type.

The population is divided into groups with meaningful similarities. The ATG gives examples such as free-standing facilities, mall locations, leased properties, and owned properties.

2. Perform cost segregation studies on sampled properties.

Properties are selected within each stratum and analyzed through cost segregation studies.

3. Develop a standard model for each facility type.

The sampled results are used to establish a model for the corresponding property group.

4. Apply the model to the population.

The costs derived from the model can then be applied to the broader population on a percentage basis.

For example, the ATG explains that a model might indicate that 10% of project costs are allocable to 5-year property. That percentage could then be applied to each facility within the same stratum.

The model therefore turns detailed analysis of a subset of properties into an allocation across a larger population.

That can be highly efficient.

But it also creates leverage.

If the sample is wrong, the error can be projected across the entire population.

That is why cost segregation audit readiness becomes especially important when sampling is involved.

Why Does Statistical Validity Matter?

The ATG identifies accuracy of the sampling results as a frequent issue.

In some situations, the sampling method may not be statistically valid.

The ATG also states that a population of fewer than 50 properties could limit the accuracy of a sampling technique unless an appropriate sampling error is considered.

That does not mean a population below 50 automatically makes sampling inappropriate.

The ATG's point is that population size and sampling error affect the reliability of the result.

A quality study therefore needs more than a statement that “representative properties were selected.”

The methodology should explain:

• What the population is
• How large the population is
• How the population was divided into strata
• How properties were selected
• How many properties were sampled
• How sampling error was considered
• How the sampled results were projected

The ATG's quality-study guidance specifically identifies population definition, population size, stratification techniques, and consideration of sampling error as factors that should be addressed.

That is what separates a documented sampling methodology from simply picking a few properties that appear convenient.

Where Is Sampling Most Useful?

The ATG specifically points to situations involving large numbers of substantially similar properties, including retail and food store operations.

Consider a restaurant operator with 150 locations.

The portfolio might contain:

• Free-standing restaurants
• Shopping-center locations
• Leased locations
• Owned locations
• Different construction generations
• Different geographic regions

Those properties may share a common brand and similar operating functions.

That does not automatically make them one homogeneous population.

A free-standing restaurant may have different site improvements and utility infrastructure from a mall location.

A newer prototype may have different construction specifications from an older prototype.

A leased property may have a different scope of tenant improvements from an owned facility.

The ATG's modeling discussion recognizes this issue by describing stratification according to facility type.

The objective is not simply to find similarities.

The objective is to identify differences that could materially affect the cost allocation.

This is consistent with the broader engineering cost segregation process.

The engineer is not just looking at whether two buildings look alike.

The engineer is evaluating how they were built, what they contain, how they are used, and what systems support their operations.

How Should Investors Evaluate a Sampled Study?

Investors with large property portfolios should ask several practical questions.

What is the population?

Before accepting a projected percentage, understand which properties were included in the population.

How were the properties grouped?

Ask why certain facilities were placed in the same stratum.

Which properties were sampled?

Understand whether the sample selection was statistically designed or based on judgment.

How large was the sample?

A small sample can create greater uncertainty, particularly when the population itself is limited.

How was sampling error addressed?

The ATG specifically identifies sampling error as an issue that should be considered in a quality study.

The ATG also distinguishes statistical sampling from judgmental sampling.

A judgment sample is typically selected based on perceived similarities and is not statistically valid. The ATG says judgmental sampling carries a higher level of risk because of subjectivity, although it may be appropriate in certain limited circumstances when the basis for selecting the sample is rational and supported by adequate data.

That distinction matters.

“These buildings looked similar” is not the same statement as “these properties were selected through a documented sampling methodology designed to represent the defined population.”

The methodology should make the difference clear.

How Can a Small Sample Difference Affect a Large Portfolio?

Illustrative example only. Figures shown are estimated for demonstrative purposes only. Actual classifications, costs, depreciation deductions, and tax results depend on the specific properties, supporting documentation, engineering analysis, applicable tax authority, and taxpayer circumstances.

Assume an investor owns 100 substantially similar commercial facilities.

Each facility has $2 million of depreciable project cost.

The total population therefore contains:

100 × $2,000,000 = $200,000,000

Suppose the sampled properties produce a model allocating 10% of project cost to 5-year property.

The projected 5-year property allocation would be:

$200,000,000 × 10% = $20,000,000

Now suppose a different, properly supported sample produces an 11% allocation.

The projected allocation becomes:

$200,000,000 × 11% = $22,000,000

The one-percentage-point difference produces a:

$2,000,000

difference in the projected 5-year property allocation across the population.

The example demonstrates why sampling methodology matters.

A small change at the sample level can become a large change when projected across a large portfolio.

That does not mean the higher or lower result is automatically correct.

It means the sample needs to be designed and supported carefully because its conclusions may affect every property in the population.

What Should Investors Remember About Sampling?

Sampling can be a powerful tool for large portfolios.

It can reduce the resources required to analyze substantially similar properties.

But it does not eliminate the need for engineering analysis.

The engineer still has to understand the population, determine meaningful strata, evaluate the sampled properties, develop the model, and determine whether the results can reasonably be projected to the broader population.

The ATG recognizes that statistical sampling can be reliable when conducted properly. It also warns that improper sampling can produce a result that does not accurately reflect a valid estimate.

Modeling presents a similar issue.

A model can be reasonably accurate when properly analyzed, but the ATG notes that defining appropriate strata can be difficult and can become an area of controversy.

For CostSegRx engineers, the principle is straightforward:

A sample is only as useful as the population it actually represents.