Two Excel deliverables are named on Titleman's own site: an underwriting model for acquisition
teams, and a feasibility model for developers. Until this workbook was built, excelUrl — the
product's own field for a downloadable Excel export — was empty on every real showcase entry, and
three of ten warm accounts had independently asked for exactly this kind of file. This piece tears
down the feasibility half: what it computes, the error caught while building it, and why the
example was deliberately chosen to fail.
The worked example: 84 homes, and the model says no
The example is a for-sale residential scheme: 84 homes, roughly 1,900 sq ft average, asking $7.98M for the land, homes sold at an average of $615,000.
| Measure | Result |
|---|---|
| Total development cost | $44,851,470 |
| Net revenue | $49,795,074 |
| Profit | $4,943,604 |
| Margin on cost | 11.0% — fails an 18% target by 7.0 points |
| Profit on revenue | 9.9% — fails a 15% target by 5.1 points |
| Peak equity | $19,829,439 |
| Return on equity / annualised | 24.9% / 6.6% |
| Residual land value at target margin | $5,327,745 |
The last line is the number a developer actually takes into a negotiation. The model does not just say the deal misses its targets — it says the site supports $5.33M and is being offered at $7.98M, short by $2,652,255. That is the question a feasibility model is bought to answer, and it is why this is a spreadsheet rather than a PDF: change the build cost per square foot or the sales price, and the number a developer can actually pay for the land moves while they watch.
A worked example that cleared every target would have been an easier story and told a reader less. This one shows the tool doing the job it exists for.
The basis error, and how it was caught
The first draft of this workbook charged soft costs and contingency on a subtotal that included the land. Soft costs and contingency are percentages of construction — land attracts neither. Because the error was expressed as a percentage, it compounded quietly through every downstream number: it added roughly 13% to total development cost before it was caught.
This is the second basis error found in two related workbooks built this week. The first — in a separate underwriting model — defaulted a purchase price to a county tax-assessed value instead of an income-derived price, which briefly inflated a levered IRR to 53%. Both were caught the same way: recomputing the model a second time, independently, and asking why the two disagreed. Neither was caught by inspection; both were caught by disagreement between two independent implementations of the same arithmetic.
In the shipped version, land sits below the construction cost stack, and the Costs tab carries a note at that exact line explaining why — so the next person to edit the file does not quietly undo the fix.
Verification
Recalculated directly from the file with an independent formula engine, then compared against a separately written implementation of the same model. Twelve outputs agreed to the cent: cost before finance $41,703,386, total cost $44,851,470, net revenue $49,795,074, profit $4,943,604, margin 11.0222%, profit on revenue 9.9279%, profit per home $58,852.42, peak equity $19,829,439, return on equity 24.9306%, annualised return 6.5663%, residual land value $5,327,745, land as a share of total cost 17.7921%.
What the model is honest about, on the workbook itself — not only in a cover note
- No property is sourced. Every input is a worked example, stated at the top of the Sources tab rather than in a footnote, because a template that reads like an underwritten site would be a false statement.
- Interest is modelled on an average drawn balance, not a full monthly draw schedule — accurate to within a few tenths of a point of margin on a normal build curve, and explicitly flagged as wrong if the draw is heavily front- or back-loaded.
- Residual land value is a first-order figure. Lowering the land price also lowers the facility size and its interest cost, so the true residual is slightly higher than shown — the model states plainly that this makes the number conservative, not optimistic.
- Annualised return is not an IRR. It assumes equity goes in at the start and comes out at the end; for staged equity, the workbook says to build a cash flow instead.
- Four automated reality checks run against the assumptions themselves, not the result — including one that flags when revenue growth is set to outrun cost inflation, because, in the source report's own words, "a feasibility that is wrong in the flattering direction does not look wrong — it looks like a site worth buying."
Why this matters to someone with money on the line
A feasibility model that always says yes is not underwriting, it is marketing. This one was deliberately built around an example that fails its own return targets, states exactly by how much, and hands over the one number — residual land value — that turns a failing deal into a negotiating position instead of a dead end. The basis error it caught in its own construction is not a flaw in the story; it is the reason the story is worth telling: the same independent-recompute check that caught a 53% IRR the first time caught a 13% cost understatement the second time, in a different workbook, built by a different pass.