Risk, Uncertainty, and BDI in Brazil

1. Introduction

BDI (the Brazilian acronym for Benefits and Indirect Expenses) is one of the most debated elements in the pricing of construction and engineering services in Brazil.

Much of the discussion focuses on the percentage: which BDI to use, which benchmarks to adopt, and whether a given value is within the commonly accepted ranges.

But there is a previous and more important question:

What risk is being priced in the BDI?

The question sounds simple, but it involves different concepts that are often treated as if they are the same thing.

A construction project has uncertainties. Those uncertainties give rise to risks. Risks are allocated between the parties, and that allocation can produce economic consequences that must be incorporated into the price.

Therefore, before discussing percentages, it is necessary to separate three dimensions:

Nature of uncertainty
Where does it come from?

Risk allocation
Who assumes its consequences?

Economic treatment
How will this exposure be priced?

This distinction became even more relevant with Brazil's Law No. 14,133/2021, which placed risk allocation and quantification at the center of public procurement structuring.

2. What does risk mean in BDI?

BDI should not be understood as a percentage arbitrarily added to a project's direct cost.

Its composition includes different economic portions.

Article 9 of Decree No. 7,983/2013 establishes that the overall reference price for construction and engineering services consists of the overall reference cost plus BDI.

The same provision requires the BDI composition to disclose, at a minimum:

• the central administration overhead allocation rate;
• percentages of taxes levied on the price;
• the project's risk, insurance, and performance-bond rates;
• the profit margin.

This separation is important.

Risk is not profit.

Profit represents corporate compensation. The risk portion seeks to remunerate certain economic exposures assumed by the contractor.

So the correct question is not:

Is there risk in the construction project?

Virtually every construction project has risks.

The question should be:

What risks have been assumed by the contractor and need to be considered in forming its price?

To answer, it is first necessary to understand the nature of the existing uncertainties.

3. Random uncertainty: the variability that remains

Imagine a construction project with a sufficiently developed design, known construction methods, experienced teams, and adequate information.

Still, actual productivity will never be exactly the same every day.

Material consumption will show some variation.

Equipment will not deliver exactly the same output.

Minor rework will take place.

This variability does not necessarily exist because someone made a mistake.

It is part of the real world.

This is what we call random uncertainty.

From a technical point of view, it represents the natural variability of possible outcomes.

It tends to be:

• inherent to the processes;
• not completely eliminable with new information;
• observable in similar activities;
• subject to statistical treatment.

Examples include:

• normal variation of productivity;
• usual losses of materials;
• small rework inherent to the execution;
• operational variability of the teams;
• effective performance of equipment;
• small performance fluctuations within the expected conditions.

Even in mature projects, some variability will continue to exist.

When this exposure is borne by the contractor and is not remunerated in another component of the budget, there is an economic basis for its consideration in the BDI risk portion.

So:

Random uncertainty forms one of the main technical bases of ordinary business risk associated with execution.

But it is not the only form of uncertainty in a construction project.

4. Epistemic uncertainty: what we do not yet know

Now consider another situation.

The project has limited site investigation data.

Utilities and other site interferences have not yet been fully identified.

Some quantities remain poorly defined.

Important interfaces still need to be addressed.

In this case, the uncertainty is not simply due to the natural variability of the execution.

It stems from the insufficient knowledge available.

This is known as epistemic uncertainty.

It can exist because:

• the project has low maturity;
• relevant information has not yet been collected;
• investigations are insufficient;
• technical decisions remain open;
• certain interfaces have not been resolved.

Its most important feature is reducibility.

Further geotechnical investigation can reduce uncertainty.

Engineering development can reduce uncertainty.

A survey of utilities and site interferences can reduce uncertainty.

This does not mean that all epistemic uncertainty can or should be eliminated. At some point, obtaining additional information may no longer be technically or economically justifiable.

But there is one key difference:

In random uncertainty, we know the phenomenon, but its outcome varies. In epistemic uncertainty, part of the problem lies precisely in the available knowledge about the phenomenon.

5. Does epistemic uncertainty fall outside BDI?

Not necessarily.

This is an important point.

Being random or epistemic explains the origin of uncertainty.

This does not alone determine who will bear its economic consequences.

Consider a geotechnical uncertainty.

It may result from insufficient information and, therefore, have a strong epistemic component.

But the economic treatment of this exposure will depend on additional questions:

• Who should get the information?
• Who has the capacity to reduce uncertainty?
• What information was available during procurement?
• What contracting regime applies?
• What does the risk matrix establish?
• Who contractually assumed the consequences of that event?

We then come to a fundamental distinction:

The nature of uncertainty explains its origin.

The risk matrix and the contract define its allocation.

The budget should economically reflect this allocation.

This logic is especially important under Law No. 14,133/2021.

6. What has Law 14.133 changed?

Law No. 14,133/2021 gave far greater relevance to the contractual allocation of risk.

Article 6, item XXVII, defines the risk matrix as a contractual clause that identifies risks and responsibilities between the parties and characterizes the contract's initial economic-financial balance in relation to subsequent events.

Article 22 allows the procurement notice to include a risk allocation matrix between the contracting authority and the contractor.

The law makes the matrix mandatory in specific situations, including large-scale works and services and integrated or semi-integrated contracting regimes.

One of the provisions most relevant to the BDI discussion is Article 103.

Paragraph 3 provides:

"The allocation of contractual risks will be quantified for the purpose of projecting the effects of their costs on the estimated value of the contract."

The importance of this determination is enormous.

The legislation isn't just saying:

identify the risk.

Nor is it only saying:

determine who takes the risk.

It adds:

quantify the economic effects.

This creates a direct connection between:

identification → allocation → quantification → budget.

Article 103 also states that the matrix must promote efficient risk allocation, considering, among other factors, the nature of the risk, the party benefiting from the contractual obligations, and each party's ability to manage the risk.

The risk matrix, therefore, should not be just a qualitative table attached to the contract.

It has economic consequences.

7. Risk matrix and BDI are not the same thing

It is important to avoid another confusion.

The existence of a risk matrix does not eliminate the risk portion of the BDI.

The instruments have different functions.

A risk matrix establishes the contractual allocation of exposures.

A risk analysis can assess and quantify those exposures.

The budget should incorporate their economic effects.

And the BDI risk portion may be one place where certain exposures assumed by the contractor are priced.

Other exposures may be:

• in direct costs;
• in insurance;
• in guarantees;
• in readjustment mechanisms;
• in contingencies of the contracting party;
• or specific contractual mechanisms.

This also highlights an important problem: double counting.

A risk should not be simultaneously priced in direct cost, covered by insurance, generically incorporated into the BDI and even considered in another contingency without these relationships being understood.

8. Where does TCU Decision 2,622/2013 fit?

One of Brazil's principal references for BDI in public works remains TCU Decision No. 2,622/2013—Plenary.

The decision established BDI reference values for different types of construction and for individual components, including the risk portion.

For the risk component, the following reference values are found:

These ranges have enormous importance for control and benchmarking.

But their meaning must be understood correctly.

They are statistical benchmarks.

They do not necessarily represent the specific risk exposure of each project.

9. The problem is not the TCU Decision, but the automatic use of its ranges

Imagine two highway projects.

The first consists of a relatively conventional intervention, with mature engineering, known geotechnical conditions and few interfaces.

The second has:

• Tunnels
• major bridges and other special engineering structures;
• complex geotechnics;
• numerous interferences;
• environmental interfaces;
• high construction complexity.

Both belong to the category:

Construction of roads and railways

But do they necessarily have the same risk exposure?

And more:

0.56% or 0.97% effectively represent the economic risk assumed by the contractor in each of them?

Not necessarily.

This is where the weakness of mechanically applying reference values becomes clear.

TCU Decision No. 2,622/2013 should not be treated as though it established a universal risk rate for every type of construction.

The quartiles function as benchmarks.

In fact, values outside the reference ranges do not automatically mean irregularity. Specific situations may justify different values, provided that they are technically demonstrated.

10. TCU itself recognizes the need for quantification

There is one particularly interesting aspect.

TCU Decision No. 2,622/2013 itself recognizes that risks must be identified and measured.

When discussing contingency estimation methodologies, TCU addresses different techniques and recognizes the need for consistent, systematic, transparent and reliable measurement.

Among the approaches discussed are techniques based on:

• specialized judgment;
• historical references;
• expected value;
• parametric modeling;
• risk simulation;
• hybrid methods.

Therefore, there is an important difference between:

use historical data as a reference

and

assume that a historical reference already quantifies the specific risk of the project.

These are two different things.

11. Benchmark is not risk analysis

This is perhaps the simplest distinction for understanding the problem.

A historical range roughly answers the question:

“What percentages were observed across a given population of construction projects?”

A specific analysis seeks to answer:

“What is this project's economic exposure to the risks assumed by this party?”

The first question produces a benchmark.

The second seeks to measure an exposure.

Both are useful.

But they are not equivalent.

A benchmark can serve to:

• verify reasonableness;
• compare results;
• identify outliers;
• support simple or repetitive projects;
• assist in situations where specific information is limited.

The problem arises when the benchmark completely replaces the analysis.

12. A 2013 decision in the context of a 2021 law

There is also a relevant temporal issue.

TCU Decision No. 2,622 dates from 2013.

Law No. 14,133 dates from 2021.

This does not mean that the TCU Decision is outdated.

Its reference values remain highly relevant and continue to be used by Brazilian public oversight bodies.

But the regulatory environment has evolved.

The decision was developed mainly to establish references that support the analysis and oversight of public-works budgets.

Law 14,133 reinforced a more structured contractual logic:

identify → allocate → quantify → price.

Especially after Article 103, paragraph 3, it is difficult to treat a generic historical percentage as a universal substitute for quantifying the economic effects of allocated risks.

Thus, perhaps the issue is not to abandon the TCU benchmarks.

The necessary evolution is another:

use them increasingly as benchmarks and reasonableness tests, and less as an automatic answer to the question “what is the risk of this project worth?”

13. Integrated, semi-integrated contracting and contracts with project responsibility

This discussion becomes even more relevant when the contractor assumes responsibilities for engineering development.

Under the integrated and semi-integrated regimes in Law 14,133, responsibilities are distributed differently than in conventional contracting based on a design previously developed by the contracting authority.

In the private setting, similar reasoning may arise in certain EPC, Design & Build, and Turnkey contracts.

The greater the contractor's autonomy to:

• develop the engineering design;
• choose solutions;
• define methods;
• manage interfaces;
• reduce certain uncertainties;

the greater its ability may be to assume and price certain exposures.

But this does not mean that any deficiency of information can simply be transferred.

Contractually transferring an uncertainty without providing reasonable conditions for its assessment and management does not necessarily mean producing an efficient allocation of risks.

14. What about extraordinary risks?

Not every exposure should simply be added to the BDI.

Extraordinary events or risks subject to specific contractual treatment may involve, as the case may be:

• certain environmental responsibilities;
• expropriations;
• legislative or tax changes;
• acts of the Public Administration;
• exceptional events;
• facts capable of affecting the economic-financial balance.

It is not appropriate to establish a universal rule saying that these events always or never belong to the BDI.

It is thus necessary to verify:

• the applicable legislation;
• the risk matrix;
• the contracting regime;
• the responsibilities established;
• and the contractual mechanisms provided for.

If a risk was expressly allocated to the contractor and properly priced, its mere occurrence does not automatically create a right to economic-financial rebalancing.

Likewise, it does not seem reasonable to generally incorporate into BDI exposures whose economic responsibility remains with the contracting party.

15. How to quantify risk?

If risk has an economic consequence, and Law 14,133 requires its allocation to be quantified so the effects can be projected into the estimated contract value, a practical question follows:

How should this quantification be performed?

There is no single method applicable to all situations.

Depending on the complexity, materiality and maturity of the information, the following can be used:

• historical data;
• Benchmarking;
• parametric models;
• scenarios;
• expected value;
• statistical analysis;
• decision trees;
• Monte Carlo simulation;
• other probabilistic methods.

A small, repetitive construction project with extensive historical data may not justify a sophisticated quantitative model.

A unique, billion-dollar project with material risks transferred to the contractor will probably require different treatment.

The sophistication of the method should be proportional to the importance of the decision.

Quantifying does not necessarily mean building a complex model. It means technically demonstrating where the value assigned to risk comes from.

16. Random or epistemic: where does contingency come in?

The initial distinction is again important.

Random variability can be represented statistically relatively naturally when adequate data are available.

Certain epistemic uncertainties, on the other hand, require greater care.

An unknown quantity does not automatically become a probability distribution simply because we assign it three numbers—minimum, most likely, and maximum.

When knowledge is limited, the distribution itself used can carry strong subjective judgment.

Therefore, quantitative analysis does not eliminate the need to understand the quality of the information used.

A sophisticated simulation based on fragile assumptions continues to produce fragile results.

Engineering maturity and data quality remain critical.

17. A rule of thumb

Before defining the risk portion of a construction project, five questions can help:

1. Even with sufficiently mature information, would this variability continue to exist?

If so, there is a strong random component.

2. Could additional information materially reduce uncertainty?

If so, there is a strong epistemic component.

3. Who has the best conditions to control, reduce or manage its consequences?

This question helps to assess the efficiency of the allocation.

4. Who actually took the risk in the contract and risk matrix?

This defines who bears its economic consequences within the contracted conditions.

5. How has this risk been quantified and where is its economic consequence being considered?

Only then does it make sense to decide whether a given amount belongs in direct cost, insurance, contingency, the BDI risk portion, or another contractual mechanism.

18. From the tabulated percentage to the quantified risk

Perhaps that is the main point of the discussion.

For many years, the discussion on BDI has focused on questions such as:

What percentage to use?

Is it within the TCU range?

First quartile, mean, or third quartile?

These questions remain relevant.

But Law No. 14,133/2021 adds a prior question:

What risks are being allocated and what is the economic reflection of this allocation?

This change is important.

TCU Decision No. 2,622/2013 remains a valuable reference.

Its percentages provide important benchmarks and help oversight bodies identify values that deserve closer examination.

But a benchmark should not automatically replace a specific analysis when materiality and complexity warrant further treatment.

The greater the uniqueness, value, and exposure of a venture, the less satisfactory a response based solely on:

“For this type of construction, we usually use X%.”

19. Conclusion

The BDI debate should not start with the percentage.

It should start with risk.

And before risk comes uncertainty.

The distinction between random and epistemic uncertainty helps to understand where exposure comes from.

The risk matrix and contract determine how its consequences are distributed.

Risk analysis seeks to quantify them.

And the budget should economically reflect this allocation.

In a nutshell:

The nature of uncertainty explains its origin.

The risk matrix defines its allocation.

The analysis quantifies the exposure.

And the budget translates that exposure into price.

In this context, the reference ranges in TCU Decision No. 2,622/2013—Plenary remain highly useful as benchmarks and oversight tools.

But the legal development introduced by Law No. 14,133/2021—especially the requirement to quantify allocated risks so their effects can be projected into the estimated contract value—reinforces the need to move beyond the automatic application of historical percentages.

The challenge is not to abandon the benchmarks.

It is to use them for what they do best: compare results and test their reasonableness.

Because risk should not enter the BDI simply because there is a range available.

It should be included because it has been understood, allocated, quantified, and needs to be priced.