# Self-Check Key

> Illustrative teaching case. Northgate Commons is not a real property, this is not a real transaction, and the term sheet is not a real financing quote. Every figure is fabricated for instruction.
>
> Open this after you have built the model, not while you are building it. The
> value of the near-miss table below is that it names the mistake, and it can
> only do that if you made one honestly.

Check your work in order. Each checkpoint depends on the one above it, so a miss at the top explains every miss below. The tolerance band is what a reasonable rounding path can produce; a value outside the band points to a method difference, not a rounding difference.

## Summary

| Checkpoint | Answer | Tolerance |
|---|---|---|
| Normalized year-one net operating income | $1,090,004 | to the dollar |
| Supportable loan amount and the binding test | $10,634,000 (Minimum debt yield of 10.25%) | to the nearest $1,000 |
| Unlevered IRR over the five-year hold | 9.68% | within 25 basis points |
| Levered IRR over the five-year hold | 13.42% | within 50 basis points |
| Equity multiple over the five-year hold | 1.84x | within 0.05x |

## Why tolerance bands at all

Two people can underwrite the same deal correctly and land a few basis points
apart. Annual versus monthly periods, rounding a loan to a thousand or to ten
thousand, a solver that stops at a slightly different place: none of those are
errors. A band that admits them keeps the check honest about what it is actually
testing, which is method rather than arithmetic hygiene.

The first checkpoint has no band, and that is deliberate. Every input to
normalized net operating income is either a stated figure or a stated percentage
of a stated figure. There is no rounding path that moves it. If your number is
not the number, something in the method is different, and the table under that
checkpoint will usually say what.

---

## Checkpoint 1. Normalized year-one net operating income

> ### $1,090,004

**Tolerance:** to the dollar. Accepted range: exact.

Every input is a stated figure or a stated percentage of a stated figure, so there is no rounding path that moves this number.

### If you got something else

| Your answer | What produced it |
|---|---|
| $1,118,412 | You took the 2026 trailing twelve months exactly as the seller reported it. None of the three normalizations were applied, and reserves were left out as well. |
| $1,304,912 | You pulled the roof out of repairs and maintenance and stopped. The management fee, the tax reassessment, and reserves are all still missing. |
| $1,248,962 | You removed the roof and added a management fee, but left property taxes at the seller assessed value and took no replacement reserve. |
| $1,181,954 | You removed the roof and reset taxes, but the property is self-managed so there is no management fee line to correct; you have to add one that was never there. |
| $1,126,004 | All three normalizations are right and reserves were left out. The term sheet escrows $300 per unit per year, so it is a real cash cost. |
| $903,504 | You added the fee, reset taxes, and took reserves, but left the August 2026 roof replacement sitting in repairs and maintenance. It is capital, not an operating expense. |
| $1,469,615 | You used the broker year-one pro forma. It applies rent growth, uses a vacancy factor below both the submarket and the property, and carries no management fee, no reassessment, and no reserves. |

If your answer is not on that list and not inside the band, go back to the
source documents rather than to the model. Re-read the statement notes and the
disclosures, then look at the month-by-month repairs and maintenance line and ask
which months break the pattern. Almost every miss at this checkpoint is a line
read from the statement without the note that governs it.

---

## Checkpoint 2. Supportable loan amount and the binding test

> ### $10,634,000

**Binding test: Minimum debt yield of 10.25%**

**Tolerance:** to the nearest $1,000. Accepted range: $10,633,000 to $10,635,000.

Lenders round loan proceeds down to a round figure. Anything inside a thousand dollars is the same answer.

**The three tests**

- Maximum loan under the loan-to-value test: $11,310,000
- Maximum loan under the coverage test: $11,678,361
- Maximum loan under the debt yield test: $10,634,185

### If you got something else

| Your answer | What produced it |
|---|---|
| $11,310,000 | You ran the LTV and DSCR tests and stopped. The debt yield test is the tightest of the three here, and it is the one most often skipped. |
| $11,678,000 | You sized on DSCR alone. DSCR is the loosest constraint on this deal, so it is not the one that governs. |
| $13,732,000 | You sized the DSCR test off the interest-only payment instead of the fully amortizing constant. The term sheet sizes on the greater of actual and fully amortizing debt service, so the 30-year constant governs even during the interest-only period. |
| $14,337,000 | You applied the debt yield test correctly but to the broker pro forma net operating income rather than your normalized figure. The mechanics were right and the input was not. |

If your answer is not on that list and not inside the band, confirm the
normalized net operating income from checkpoint 1 before touching the sizing.
The three tests all run off that one number, so an error upstream reproduces
itself in all three and can still leave a plausible-looking loan. Once the net
operating income ties, check which test binds rather than which loan is largest.

---

## Checkpoint 3. Unlevered IRR over the five-year hold

> ### 9.68%

**Tolerance:** within 25 basis points. Accepted range: 9.43% to 9.93%.

Annual-period cash flows, whole-dollar rounding in the projection, and different solver settings can move the result a few basis points either way.

### If you got something else

| Your answer | What produced it |
|---|---|
| 10.79% | You left the renovation program out of the cash flows. The 34 units a year for three years at $9,500 each is a real cash outflow above the replacement reserve already inside net operating income. |
| 9.13% | You capitalized year-five net operating income at the exit cap instead of year six. A buyer at the end of year five prices the forward year, so the year-six figure is the one to capitalize. |

If your answer is not on that list and not inside the band, work through the
cash flows before the return. Confirm the year-six net operating income being
capitalized at exit, then the net sale proceeds after costs of sale, then the
year-zero outflow. An unlevered IRR that is off usually traces to one of those
three rather than to the solver.

---

## Checkpoint 4. Levered IRR over the five-year hold

> ### 13.42%

**Tolerance:** within 50 basis points. Accepted range: 12.92% to 13.92%.

Leverage magnifies small differences. Monthly rather than annual debt service, or a loan rounded differently, moves the result more than it moves the unlevered figure.

### If you got something else

| Your answer | What produced it |
|---|---|
| 12.77% | You amortized from month one. The term sheet gives 24 months of interest only, which raises early cash flow and leaves a larger balance at payoff. |
| 12.72% | You repaid the original loan amount at sale. Three years of amortization run after the interest-only period, so the payoff balance is lower than the original principal. |
| 14.52% | Your equity at closing was price less loan. Acquisition costs and the origination fee are also funded with equity, so the initial outflow is larger. |

If your answer is not on that list and not inside the band, check the debt
before the equity. Confirm the loan amount and binding test from checkpoint 2,
then the interest-only period and the amortizing payment that follows it, then
the payoff balance at sale. The levered return is sensitive to all of them.

---

## Checkpoint 5. Equity multiple over the five-year hold

> ### 1.84x

**Tolerance:** within 0.05x. Accepted range: 1.79x to 1.89x.

The multiple inherits every rounding difference in the cash flows, and a loan rounded to a different thousand moves both the numerator and the denominator.

### If you got something else

| Your answer | What produced it |
|---|---|
| 0.84x | You divided profit by equity rather than total distributions by equity. Equity multiple counts the return of capital as well as the return on it. |
| 1.93x | Your equity base was price less loan and excluded acquisition costs and the origination fee. |
| 1.53x | That is the unlevered multiple on total capital. The equity multiple runs on the levered cash flows and the equity check. |

If your answer is not on that list and not inside the band, compare it
against your own levered cash flows rather than starting over. The multiple and
the levered IRR in checkpoint 4 read the same column, so a checkpoint 4 that ties
alongside a multiple that does not tends to point at the summation rather than at
the model.


---

## If you cleared all five

Two things worth doing before you call it finished.

**Sensitivity.** Move the exit cap 25 basis points in each direction and see what
happens to the levered IRR. Then move year-one net operating income by 5%. One of
those will move the answer noticeably more than the other, and knowing which is
the difference between having a model and understanding a deal.

**The memo.** The five checkpoint answers are the easy part to check and the smaller part of
the work. A committee reads the recommendation and the bridge to net operating
income. If your memo cannot be used by someone who has not seen your spreadsheet,
it is not finished.

---

*Illustrative teaching case. Northgate Commons, the market, the broker, the lender, the comparables, and every figure in this packet are fabricated for instruction. Nothing here is a real property, a real transaction, or a real financing quote.*