Picture, as an illustration, an investment committee meeting on a Thursday in September. On the screen, the spreadsheet for a high-end development shows a 23% margin. Someone is pleased about the rate cut announced the day before. Nobody asks what happens to that margin if construction costs rise 15% and prices slip 5% in the same half-year. The spreadsheet knows only one future.
That September is not hypothetical. On 16 September 2026, Brazil's Monetary Policy Committee (Copom) cut the Selic target to 13.75% a year, according to the Central Bank's interest rate history. It was the fifth cut since March, when the rate began to come down from 15%. Nine days later, FGV IBRE published the September INCC-M, the national construction cost index: up 0.25% in the month and 6.61% over 12 months. The annual figure has eased from 7.07% a year earlier, but it remains high.
Falling rates and rising construction costs pull the result in opposite directions at the same time. A conventional feasibility sheet tells you whether the project works. An advanced model tells you under what conditions it stops working, how much cash runway exists until then, and who authorizes what when the numbers start to move. Rates offer comfort. Only costs will tell whether that comfort was earned.
A single future is a budget, not a decision tool
A single scenario carries a hidden assumption. It supposes that price, absorption pace, cost, schedule and interest rates will all behave together, exactly as planned. One variable off track is enough to turn the projected result into fiction.
InnovaSphere works with five absorption scenarios: optimistic, base, conservative, pessimistic and extreme stress. Each one has to move the whole model: construction schedule, equity required, month of peak cash exposure, break-even point and investor return. A pessimistic case that only trims revenue, while leaving schedule and financing cost untouched, is cosmetic.
The simulator that accompanies this article shows the logic on a small scale. Enter the base cost and the contingency, and it returns an optimistic case, a conservative case and the reserve. What matters is the gap between those numbers. If the reserve covers only part of the distance to the conservative case, the rest needs a source already identified. Hoping the base case holds is not a source.
Sensitivity shows where the project breaks before the market does
Scenarios combine variables. Sensitivity analysis isolates them: you move one assumption at a time, hold the others constant and measure the effect on margin and on cash exposure. Ranked from most to least influential, the variables form the familiar tornado chart.
The most valuable question runs in reverse: what value of each assumption wipes out the profit? In the worked example below, the base project only breaks even if construction cost rises 46%, or if prices fall 26%. In isolation, those buffers look enormous. With cost already 15% higher, however, a price drop of 17.6% is enough to erase the result. That break point, written next to each assumption, is what a committee needs most and sees least often in presentations.
The current moment calls for testing rates and costs separately. Since July 2023, the INCC has used a new weighting structure, built from budgets of residential projects completed between 2018 and 2022. The index is now reported by construction standard: economy, mid-range and high-end. For a luxury project, stressing cost against its own standard rather than the overall average is the difference between testing real cost and testing an abstraction.
Every cost line needs a dated, traceable source
A credible model starts with cost, and credible cost has a declared origin. SINAPI, maintained by Caixa together with IBGE, is described by Caixa as Brazil's main cost reference for engineering works, as required by Decree No. 7,983/2013. Since 2022 it has published parametric compositions. Caixa presents them as support for feasibility analysis, before a budget based on preliminary or detailed design exists.
The other benchmark is the CUB/m², which state construction unions publish every month under Law No. 4,591/1964. The rule most useful to modelers sits in paragraph 3 of Article 54: a CUB-based budget only counts as current if it uses the cost for that month or one of the two preceding months. A cost from six months ago is not an assumption. It is a memory.
The CUB page maintained by CBIC warns that the indicator is a partial cost: foundations, elevators, equipment and site works are left out. Multiplying CUB by floor area underestimates precisely the heaviest items in a high-end project.
For private projects, the vocabulary of the SINAPI decree is a sound discipline. It requires the contractor's overhead and profit markup, known in Brazil as BDI, to show at least central administration, taxes on price, risk, insurance and guarantees, and profit. With that breakdown, a builder's proposal can be compared line by line against the public reference instead of being argued as a single number.
Cancellations and delays run on a statutory calendar
Many models treat buyer cancellations and delays as exceptions. Brazilian law gives them precise rules, with dates and amounts. Article 67-A of Law No. 4,591/1964, added by the 2018 cancellation law, lets the developer retain the brokerage commission and a penalty of up to 25% of what the buyer paid. Under a segregated-asset structure, the penalty can reach 50%.
What changes for cash is the calendar. With segregation, the refund is due within 30 days after the occupancy permit, when construction has already stopped consuming funds. Without it, the balance is paid in a single installment 180 days after cancellation, and it may land mid-construction. If the unit finds a new buyer sooner, the amount falls due within 30 days. A serious stress test simulates a cancellation rate and places each refund in the month the law sets.
Schedule risk follows the same logic. The law allows a grace period of up to 180 calendar days after the contractual delivery date, with no penalty or termination, provided it is agreed clearly and prominently. Beyond that limit, a buyer who did not cause the delay may terminate and recover everything paid, plus the penalty, within 60 days. Our stress case uses up the entire grace period and tests the month after. That is where an operational delay becomes a liquidity event.
Asset segregation turns cash prudence into an obligation
Under the segregated-asset regime, the land, buildings and rights of a development are kept apart from the developer's own estate and answer only for that project's debts. Article 31-D of Law No. 4,591/1964 spells out what this means day to day: a dedicated bank account, quarterly trial balances and, at least every three months, a report on construction progress against schedule and funds received. The developer, the law says, must preserve the resources needed to finish the work.
It is the best statutory definition of cash policy available in the sector. A model that already projects cash flow per project, with its own account and a quarterly report, complies with the law as a by-product of management. Segregation also opens the special tax regime, a unified monthly payment of 4% of revenue actually received. Because the base is cash collected, the tax follows each scenario's pace and must be calculated month by month.
Electing the regime is irrevocable while credits or obligations toward buyers remain, and the current text already cross-references the 2025 tax reform. That is why the choice enters the model as an explicit assumption, dated with the day the statute was read.
A hard-cash policy turns the model into a rule
A model filed away after approval protects nothing. A hard-cash policy ties the model to daily practice: reserve levels set in advance determine which expenses may be authorized. We use a color scale. In green, the approved budget runs normally. In yellow, discretionary spending needs extra sign-off. In red, only signed contracts and critical work fronts proceed. In black, a formal contingency plan kicks in, with immediate reporting to investors.
Three rules complete the policy. Each construction phase starts only on revenue actually received, not merely contracted. Management fees and supplier payments are tied to measurable deliverables and real cash inflows. And there is a minimum contingency of 5% of construction cost, revised whenever sensitivity shows more risk than the reserve covers. The hard part is writing the triggers before you need them, because under pressure every exception looks reasonable.
The model that protects decisions is the one that defines, before the crisis, who may spend what at each cash level.
Traceability completes the design: every assumption with a source, a date and an owner, every version with what changed and why. Investors can then debate assumptions rather than the credibility of whoever wrote them.
Three group companies work at the edges of this model, not at its core. C&S Engenharia turns quantities into referenced budgets and comparable progress measurements. InnConta validates the tax assumption before it becomes irreversible. Valida Auditoria checks projections against the segregated account's balances and site measurements.
Worked example: one project, three scenarios
The table uses a hypothetical development. Every figure is a teaching assumption by InnovaSphere, not market data or a real case. Base construction cost matches the simulator's default. Tax is applied to the full potential value of the units (VGV), as if everything were collected with no cancellations. Figures in millions of reais.
| Item | Base | Conservative | Stress |
|---|---|---|---|
| Assumption: VGV of 100 units | 60.00 | 57.00 (price -5%) | 54.00 (price -10%) |
| Assumption: construction cost | 30.00 | 34.50 (+15%) | 34.50 (+15%) |
| Assumption: land, design and permits | 9.00 | 9.00 | 9.00 |
| Assumption: commercial and administrative expenses (8% of VGV) | 4.80 | 4.56 | 4.32 |
| Assumption: RET tax (4% of revenue received) | 2.40 | 2.28 | 2.16 |
| Assumption: overhead during delay (6 months at 0.15 per month) | 0.00 | 0.00 | 0.90 |
| Result: total cost | 46.20 | 50.34 | 50.88 |
| Result: project profit | 13.80 | 6.66 | 3.12 |
| Result: margin on VGV | 23.0% | 11.7% | 5.8% |
| Result: cost overrun beyond the 5% contingency (1.50) | 0.00 | 3.00 | 3.00 |
Three readings follow. The conservative case halves the margin, and the stress case cuts it to about a quarter of the base, without any single assumption looking absurd. The 5% contingency covers a third of the conservative cost increase. The remaining 3 million needs a defined source before construction starts.
The third reading is the most uncomfortable. The stress case still ends in profit, but only the month-by-month model shows whether peak cash exposure fits the available capital. The risk grows if cancellations happen outside segregation and refunds coincide with the construction peak.
When it does not pay off to sophisticate the model
Advanced modeling costs money to build, maintain and govern. Sometimes that cost exceeds the value of the information.
- Small, short project with locked cost. Under a fixed-price construction contract, which the development statute defines as non-adjustable, and with all-equity funding, five scenarios add little to a good price and schedule sensitivity.
- Very early stage, with no design or site defined. Before preliminary design, a meticulous model produces false precision. Ranges based on parametric compositions inform better than dozens of tabs.
- No governance to enforce the triggers. If nobody has a mandate to halt spending when cash changes color, a sophisticated model only breeds false comfort. A simple model that is followed protects more.
Checklist for a model that protects decisions
- Define the five absorption curves and confirm each one changes schedule, capital and return, together with the commercial curation team.
- Calculate the break point for price, cost, pace, schedule and rates, and write it next to each assumption, with the investment committee.
- Reference every cost line to SINAPI, to the CUB of the month or the two before, or to a dated quote, adding the items outside the CUB.
- Break down the builder's BDI into the items listed in the SINAPI decree and compare it with the public reference.
- Simulate cancellations and delay under the statutory rules, placing each refund in its legal month, together with legal counsel.
- Validate segregation and the special tax regime election against the current text, with tax advisers.
- Formalize the color triggers and who authorizes each level, and submit the model to independent review.
Frequently asked questions
How many scenarios does a model need?
There is no legal number. We use five because they span optimistic to extreme stress, with intermediate steps that are useful for triggers. What matters more is that each one moves schedule, capital and return.
What is the difference between sensitivity and stress testing?
Sensitivity moves one variable at a time and shows which one weighs most. Stress testing combines lower prices, higher costs, cancellations and delay to check whether the project survives a severe but plausible picture.
SINAPI or CUB: which reference should I use?
They serve different purposes. The CUB/m² is a partial cost for a standard design, useful for orders of magnitude and required when registering a development. SINAPI offers detailed compositions for line-by-line budgeting. Use the CUB as a check and the referenced budget as the base.
Is a 5% contingency enough?
It depends on the risk range sensitivity reveals. In the example, it covers a third of the conservative increase. It is a discipline floor, not a guarantee.
How often should the model be updated?
Monthly, tracking the CUB, cost indices and the actual cash position. In segregated developments, the quarterly rhythm of trial balances and progress reports is the natural milestone for full reviews.
Legal basis
- Law 4,591/1964: arts. 31-A to 31-F, 32 (h), 43-A, 54, 55 and 67-A.
- Law 13,786/2018: cancellations and delivery grace period.
- Law 10,931/2004: arts. 1 and 4 (RET).
- Decree 7,983/2013: art. 9 (BDI).
