Make It Pencil

You are holding a casebook with a voice, which is a strange animal, so let me explain the species.

Real estate books come in two kinds. The first kind tells you what to think: frameworks, definitions, the vocabulary of the business, delivered from altitude. The second kind tells you stories: the deal that almost died, the fortune made at the courthouse steps. The first kind is true and forgettable. The second kind is memorable and useless, because the story only teaches the person who lived it. What actually trains judgment, in this business and every other, is the thing between: a concrete situation, incomplete information, real numbers, a clock, and a decision that is yours to make and defend. Business schools discovered this a century ago and called it the case method. Real estate, an industry that runs entirely on judgment under uncertainty, has almost no case literature outside a few graduate programs. This book is my attempt to fix that.

The structure is simple. Each chapter opens with a concept: how to think about markets, land, buildings, capital, or time. Each concept is then handed to a case, a fictionalized composite built from real fact patterns, with names invented and numbers rounded but internally consistent, so that every exhibit can be recomputed by hand and every claim can be checked. The cases share one universe. You will follow a firm called Meridian Residential Partners through a market it cannot crack, land negotiations, contracts, capital raises, a lease-up that misses, a loan that matures at the wrong moment, and a fund that must choose between its investors' money and its own future. The people recur. The consequences carry forward. A concession granted in one case becomes a skewed comparable in another; a contract read line by line in one case becomes the checklist another case runs. That is how the business actually works, and it is how the book works.

A few instructions for use, learned from watching people read early drafts.

Do the cases. The chapters will make you feel like you understand; only the cases will tell you whether you do. When a case ends with "write the memo," write the memo, even three bullet points on a napkin. The gap between reading a decision and making one is the entire content of professional judgment, and no amount of reading closes it.

Foot the exhibits. Every table in this book reproduces from stated assumptions with arithmetic you learned in school. When an exhibit says a deal yields 6.6% against a 7.0% hurdle, multiply it out yourself. Partly this builds the muscle. Mostly it instills the deeper habit: in practice, the numbers handed to you are somebody's argument, and reproducing them independently is the only way to know what they actually say.

Argue with the assumptions. Each exhibit states its assumptions in the open and tells you they are attackable. That openness is the method, not modesty. The difference between a sophisticated investor and a spreadsheet operator is that the sophisticated investor knows which cell the whole answer lives in.

Who this book is for: the analyst in their first years, who will find here the things that rarely get written down because most senior people assume every junior already knows them; the career-switcher from finance, law, engineering, or construction, who knows their corner and wants the whole board; the reader still in school who suspects, correctly, that the textbook's clean diagrams omit the parts that decide outcomes; and the private investor who has bought a building or two and wants to understand the game the institutions are playing around them. You need no prior vocabulary. You do need a willingness to do arithmetic, because this business is arithmetic wearing a leather jacket.

One more thing, about why. It would be easy to read a book of yield-on-cost calculations and promote structures as a manual for extraction. I wrote it as the opposite, and the dedication says so plainly. This country is short millions of homes. Sentiment does not solve shortages. Getting land entitled, buildings financed, and projects delivered, at every price point, over and over, without losing capital or nerve, does. Skill in this craft is not in tension with the housing problem; it is the only thing that has ever solved it. The book exists to move skill from experienced hands to new ones, faster than apprenticeship alone can manage.

The firm you will follow spends much of this book trying to break into a coastal city where everything says build and nothing can be built. Its first case is smaller and closer to the ground: a second entrance, a private road, and a twenty-page easement nobody had read. Sharpen a pencil.

Look at any American skyline, or any American arterial strip, and you are looking at a building code rendered in three dimensions.

The shapes this industry builds are intersections of construction cost curves with legal ceilings, not primarily aesthetic choices. Wood frame is cheap and the code caps how high it can go. Concrete and steel remove the cap and roughly double the cost per foot. At the boundaries between construction types sit cliffs that generate whole product categories. The garden apartment exists because three wood stories over surface parking is the cheapest legal way to stack homes. The five-over-one podium exists because the code lets five wood stories ride on a concrete base, and the moment land gets expensive enough to justify the concrete, most parcels in the district sprout the same silhouette. The high-rise exists only where land prices clear the full concrete premium. Height is land price divided by construction cost, and the skyline is that division performed in public.

The construction type is therefore the underwriting decision that pretends to be an architectural one. It fixes the height envelope, bounds the cost per square foot, and dictates the fire separations, the sprinkler standard, the parking solution, the insurer, the lender, and the exit cap rate. By the time an architect issues design development drawings, the type has already foreclosed two thirds of the design space. Choosing it is the developer's job, and it should happen before the architect is engaged for design development. The architect's job is to optimize within the type, not to pick it.

The Ladder and Its Cliffs

The International Building Code, adopted in some amended form by nearly every U.S. jurisdiction, sorts all buildings into five construction types defined by the combustibility of the structural elements and the fire-resistance rating of those elements. The A and B subtypes distinguish protected from unprotected assemblies. Layered on top is a height and area chart (IBC Tables 504.3 and 504.4) that tells you how tall the building can be for a given occupancy. For R-2 residential, which covers nearly all multifamily, that chart is the first document to open in early feasibility.

IBC construction types for R-2 residential, with 2025 to 2026 hard cost ranges

Type	Structure	Stories	Height	Hard cost per GSF	Typical role
V-B	Wood frame, unrated	3	60 ft	$180 to 220	Suburban garden
V-A	Wood frame, 1-hour	4	70 ft	$200 to 250	Taller garden
III-A	Wood over non-combustible shell	5	85 ft	$260 to 340	Urban infill workhorse
IV-C	Mass timber, exposed	8	85 ft	$310 to 400	Mid-rise timber
IV-B	Mass timber, mostly encapsulated	12	180 ft	$310 to 400	Mid to high-rise timber
IV-A	Mass timber, fully encapsulated	18	270 ft	$310 to 400	High-rise timber
I-B	Concrete or steel, 2-hour	12	180 ft	$360 to 480	Mid to high-rise concrete
I-A	Concrete or steel, 3-hour	Unlimited	Unlimited	$450 to 650+	Towers

Stories and heights are the R-2 maximums under the 2021 IBC with a full NFPA 13 sprinkler system; an NFPA 13R system caps any type at four stories and 60 feet, and jurisdictions amend the tables, so confirm the local row before relying on it. The Type III-A range is for the wood portion over a podium. Those cost ranges are rough national averages for 2025 and 2026 and will miss by 20% in either direction depending on market, labor environment, prevailing wage exposure, and specification. The shape of the curve is durable across cycles even when the levels are not.

Read the table as a staircase rather than a ramp. Each step up in fire rating buys height and density at the price of materials and labor, and the risers are not evenly spaced. The largest riser in the entire code sits at 75 feet, measured to the floor of the highest occupied story above the lowest level of fire department vehicle access. Cross it and you trigger Chapter 4 of the IBC: standpipes, fire pumps with backup power, smoke control, elevator recall, areas of refuge, emergency responder radio coverage. The delta from a 74-foot building to an 80-foot building runs $20 to $40 per gross square foot on the entire building, not just the marginal floor. On a 100,000 square foot project:

Low end: 100,000 × $20 = $2 million
High end: 100,000 × $40 = $4 million

That is $2 million to $4 million for six feet of height. So podium projects hold the top occupied floor at 74 feet 11 inches. The parapet above it can go where the roof needs it, because the trigger is measured to the floor people live on and not to the roofline.

Smaller cliffs sit inside each step. Type III requires non-combustible exterior walls, which in practice means fire-retardant-treated studs in the exterior assemblies at a 30 to 50% premium over untreated dimensional lumber, with tighter availability. Moving from NFPA 13R, the residential sprinkler standard with reduced coverage in attics and closets, to full NFPA 13 adds roughly $0.50 to $1.50 per gross square foot. Builder's risk premiums on wood frame have risen roughly two to four times since 2018 on the strength of a bad large-loss fire history, and the premium delta between Type V and Type I on a $50 million project can run $300,000 to $700,000. Operating insurance carries the same asymmetry in perpetuity: roughly $200 to $350 per unit per year for concrete against $400 to $800 for comparable wood frame. Capitalized at a 5% cap, a $300 annual delta is $300 ÷ 0.05 = $6,000 per unit of value that never appears on the construction cost line.

The offsetting cliff runs the other way, and it is schedule. A wood-frame five-over-one builds in 12 to 16 months. An equivalent concrete building takes 22 to 30. On a $40 million construction loan at 7%, six to fifteen extra months of carry is roughly $1.4 to $3.5 million of additional capital tied up:

Annual interest: $40 million × 7% = $2.8 million
Six extra months: 0.5 × $2.8 million = $1.4 million
Fifteen extra months: 1.25 × $2.8 million = $3.5 million

That is the largest hidden cost of escalating to Type I and the reason developers fight for the tallest podium and the most wood the code will permit.

The rule that falls out of the staircase is simple to state and constantly violated: build the cheapest type the site, the zoning, and the rent will support. Skipping a step upward is a common form of value destruction in multifamily, and it almost always traces to an architect or a lender pushing for performance the rent cannot underwrite. Skipping a step downward shows up as forgone density, which is less visible and just as expensive.

Four Products, One Acre

Any multifamily site reduces to one identity. Gross lot times one minus the setback factor gives the buildable lot. Buildable lot times coverage gives the footprint. Footprint times rentable stories gives gross residential area. Gross residential times efficiency gives rentable area. Rentable area divided by average unit size gives units. The construction type does not appear explicitly because it is hiding inside coverage and rentable stories, which is where it does its work.

Run that identity across the four products on the same one-acre parcel and the industry's entire product taxonomy falls out of the arithmetic.

Four products on one acre at 1,000 SF average unit size

Input	Garden (V-A, 4 stories)	Wrap (III-A, 5 stories)	Podium (5-over-1)	Tower (I-A, 25 stories)
Buildable lot	37,026 SF	38,333 SF	37,026 SF	39,204 SF
Coverage	35%	65%	65%	30%
Footprint	12,959 SF	24,917 SF	24,067 SF	11,761 SF
Rentable stories	4	5	5	25
Gross residential SF	51,836	80,980	120,335	294,025
Efficiency	85%	80%	80%	76%
Units at 1,000 SF	44	65	96	223
Parking at 1.3 per unit	57	85	125	290
Hard cost per GSF	$200	$260	$310	$450 to $650
Approximate cost per unit	$236K	$324K	$389K	$593K to $857K

The wrap's gross residential area nets out the internal wood-frame garage, which takes about 35% of each floor; the podium's five stories are the wood stories above the concrete base; the tower's coverage is the tower plate. Parking is surface for the garden, internal for the wrap, podium for the five-over-one, and podium or below grade for the tower. Cost per unit is hard cost only.

Four products, four densities, four cost bases. The garden's coverage is only 35% because surface parking is taking the rest of the lot, and its efficiency is the highest of the four because a true walk-up has no elevator core and short corridors. The wrap solves a narrower problem: a site too expensive for surface parking but not expensive enough to justify concrete, answered by putting a wood-frame garage in the middle of the floor plate and wrapping units around it. The garage eats 30 to 50% of each floor's interior, so 65% coverage still only produces 65 units. The podium moves the parking into a concrete base that the code treats as a separate building, frees the entire floor plate above for units, and nearly doubles the garden's density. The tower's coverage collapses to a small plate on a podium and wins on density only by stacking floors.

The choice among them is a question about two numbers: land basis and achievable rent. As a rough rule in high-cost metros, below $25 per buildable square foot of land with rent around $1.80 per square foot per month, garden pencils. Between $25 and $60 of land with rent at $2.20 to $2.80, wrap pencils. Above $60 with rent above $2.80, podium pencils. Type I concrete needs roughly $4.50 per gross square foot per month to underwrite at current costs and a 5% exit cap, which is achievable in maybe twenty submarkets in the country.

Parking is the mechanism that drives that escalation, and its cost curve is steeper than the structure's. A surface stall costs roughly $4,000. An above-grade podium stall runs $25,000 to $35,000, and a second podium level with its structural transfer to the wood frame above runs $30,000 to $50,000. Below grade runs $50,000 to $90,000 once excavation, waterproofing, ventilation, and perimeter reinforcement are priced. Structured parking consumes roughly 350 gross square feet per stall above grade and 30 to 70 more below. At one stall per unit, a 96-unit five-over-one needs 96 × 350 = 33,600 square feet of parking against a 24,067 square foot footprint, so single-podium projects need either below-grade supplement or a lower ratio. Each of those escalations has to be absorbed by the residential rent above. That is the reason infill rents must clear suburban rents by so much to produce the same yield.

For triage, collapse the identity into one coefficient per product: units per acre per rentable floor at 1,000 square foot units. Garden is 11, wrap is 13, podium is 19, tower is 9. Multiply by rentable floors, by acres, and by 1,000 ÷ your actual average unit size. A 2.5-acre student housing site as a five-over-one at 800 square foot units: 19 × 5 × 2.5 × (1,000 ÷ 800) = 297 units. The long-form identity on the same inputs returns 300. Three units of slop across a 300-unit answer you can produce standing on the sidewalk. Note that the tower's coefficient is the lowest of the four. Per floor, a podium beats a tower. The tower only wins because you cannot legally build a 25-story podium.

Why the Code Writes the Skyline

The five-over-one is the single most consequential construction innovation in multifamily in twenty-five years, and it exists because of one provision. IBC 510.2, horizontal building separation, allows the podium and the wood-frame structure above it to be regulated as two separate buildings stacked vertically, provided a three-hour horizontal assembly separates them. Without that section, the whole stack would be classified by its most restrictive element and the wood-frame portion would be impossible. One paragraph of code, and the arterial corridors of every dense American metro between New York and Los Angeles look the way they look. California and Washington amended further to allow six wood stories over a podium, so the densest podium stock sits on the West Coast.

The corollary is the error that ruins more first models than any other. People see a seven-story building, multiply the footprint by seven floors and by efficiency, and land 40% too high. The bottom two floors are parking. They produce no rentable square footage. Adding a podium level adds 70 to 90 cars and no units, because the unit count is governed by the wood-frame stories above, which the code caps at five or six. Adding podium height moves those wood floors closer to the 75-foot cliff without adding to their number. Podium height is a parking decision, not a density decision.

The same logic explains why regulatory changes move skylines faster than markets do. In the cities that have reduced or eliminated parking minimums, freeing a podium level or a below-grade level for residential or amenity can shift density on a one-acre site by 20 to 30% overnight. State density bonus regimes have gone further. California's Density Bonus Law allows up to 50% additional density and up to three additional stories above base zoning for a 5 to 24% affordable set-aside. Florida's Live Local Act preempts local zoning for multifamily with an affordable component. Washington's HB 1110 requires cities to allow up to four units on any single-family lot. In those states the base zoning envelope is now the floor, not the ceiling. A one-acre site zoned for 65 units per acre that supports 100 under a 50% bonus has just moved from a four-story garden to a five-over-one, which is a different construction type, a different cost curve, and a different lender.

Which cap binds is the central judgment in feasibility, and it varies by geography in a way that tells you something about local politics. Most urban infill sites have height or FAR binding first with coverage close behind. Most suburban sites have the parking ratio binding first with a units-per-acre density cap close behind. Run the bottom-up envelope against the zoning cap, then against the IBC cap, then against the parking minimum, and take the lowest of the three. A city where coverage is the binding cap on every podium has been planning for less density than its zoning advertises.

Yield on Cost Versus Development Spread

For the practitioner all of this compresses into the prototype decision: what to build on a parcel that could legally hold several answers. The core arithmetic is a two-number comparison. Untrended yield on cost, stabilized income divided by total cost, measures the project's efficiency. Development spread, yield on cost minus the exit cap rate for that product, measures its profit. The two can rank the same prototypes differently, and they diverge when product types trade at different exit caps, which is when the decision matters most. Profit itself has two expressions and they are not the same ranking. The spread is a difference; value over cost, yield on cost divided by the exit cap, is a ratio. They agree most of the time and part company when two prototypes sit within a few basis points of each other on spread, at which point the ratio is the one to follow, because it is the one that scales with the dollars deployed.

Work it. Suppose the garden prototype on a parcel unlevers to a 6.4% yield on cost and the podium to 6.2%. On yield, garden wins. Now attach exit caps. Institutional buyers pay 25 to 50 basis points tighter for concrete than for wood frame, and urban podium product generally trades tighter than suburban walk-up, so call the garden's exit 5.75% and the podium's 5.25%. Value ÷ cost = yield on cost ÷ exit cap. Apply it to each prototype:

Garden: 6.4 ÷ 5.75 = 1.11, an 11% profit on cost
Podium: 6.2 ÷ 5.25 = 1.18, an 18% profit on cost
Podium over garden: 18 ÷ 11 = 1.64

The lower-yielding project creates roughly 60% more value per dollar deployed. A developer who chooses on yield alone is answering "which project is efficient" while the capital is asking "which project makes money."

The reverse case is just as real and less discussed. A prototype that wins on spread only because you underwrote a tight exit cap for a product you have never sold in that submarket is a position in an assumption, not a better project. Spread is a difference between a number you can compute and a number you must forecast. Treat the forecast with the suspicion it deserves, and stress it before you stress the rents.

The less obvious input is the boundary of your own data. A firm's cost estimates are knowledge inside the product types it has actually built and hopes everywhere else, and hopes require pricing: fatter contingency, a higher required yield, or a structure (a guaranteed-maximum-price contract, a fee-build partner) that pays someone else to hold the uncertainty. Most first-time-product disasters were priced as if the firm's confidence traveled with it.

Mass Timber and the Buildings We Already Have

Two developments are moving the ladder itself, and both live in the four-to-twelve-story band where podium runs out and concrete gets expensive.

The 2021 IBC split Type IV into subtypes and expanded the height table dramatically: for R-2, eight stories and 85 feet for IV-C with exposed timber and a fire-rated charring layer, twelve stories and 180 feet for the mostly encapsulated IV-B, eighteen stories and 270 feet for the fully encapsulated IV-A. Two things drove that. Fire research, including ASTM E119 testing on cross-laminated timber panels by the U.S. Forest Service and the tall wood tests at NIST, established that mass timber chars at a predictable 1.5 to 2 inches per hour and self-insulates, so residual structural capacity after an hour of fire exposure is calculable. And the carbon math turned. A mass timber building sequesters roughly 1.5 to 2 tons of CO2 equivalent per cubic meter of structural wood; for a typical eight-story residential building the embodied carbon delta against concrete is on the order of −2,000 to −4,000 tons versus +4,000 to +6,000.

The economics are still in transition. Direct hard cost sits roughly 5 to 12% above equivalent concrete or podium construction depending on market and supplier. Three things compress the gap. Speed: panels arrive cut to size and assemble like a kit, framing 25 to 40% faster than concrete, which shortens interest carry and pulls lease-up forward by three to six months. Weight: a CLT structure weighs roughly a fifth of an equivalent concrete frame, which buys cheaper foundations and matters enormously on poor soils or vertical additions. And a small documented rent premium of roughly 2 to 5% for exposed wood in the few markets with comp sets deep enough to measure it, Portland, Seattle, and Vancouver. Panel supply is no longer the bottleneck; design experience is. A mass timber project needs a structural engineer fluent in CLT provisions, an architect comfortable with panel tolerances, and a contractor who has built one before. Once the hard-cost gap to concrete closes inside 5%, speed and carbon should make timber the default in that band. That is a forecast, and it has been made before.

Adaptive reuse is a regulatory pathway and a financing structure rather than a construction type, and its constraint is geometry rather than code. Residential units need windows: the IBC requires roughly 8% of floor area in glazed openings for R-2, half of it operable, which limits habitable depth to 30 to 40 feet from window line to corridor. Office buildings from the 1960s through the 1990s were designed to maximize lease span, running 45 to 55 feet on each side of a large core. Much of the 250 to 350 million square feet of functionally obsolete Class B office in American central business districts will never convert for that reason, and the successful conversions cluster in pre-war buildings with narrow plates and small cores. The work itself is MEP and envelope rather than structure: plumbing risers at every unit where an office had them only near restrooms, HVAC re-zoned from 1,500 square foot office zones to per-unit zones, new electrical risers and meter banks, operable windows, insulation to a residential energy code the building was never built for. A well-shaped, well-located building with historic tax credits converts at $200 to $300 per square foot all in, competitive with new construction. A badly shaped one runs $400 to $550 once the corridor reconfigurations and structural penetrations are priced, at which point building new wins. The sorting happens on the floor plate, and it happens before the pro forma.

The yield-versus-spread comparison now gets a parcel. The case gives Meridian ten acres and three defensible products: the garden it knows, the townhome build-to-rent product that is the market vernacular but which the firm has never built, and a four-story building that wins on paper and exists only if a rezoning arrives. The Prototype Decision is built so that no single row of the exhibit settles it, because in practice none does.

A Case in Development Strategy: Building Type, Unit Mix, and the Economics of Product Choice

Synopsis

Land is indifferent to what you build on it; capital is not. A year after the entry campaign chronicled in Breaking into Baymouth, the standing monthly call with Harbor Line finally produced a parcel that cleared the screen, and Meridian now has 10.0 acres under contract in the Baymouth region at $500,000 an acre. It faces three honest answers to the next question: what is the product? Garden multifamily is the firm's native prototype, the one its cost data, its lenders, and its LP all know. Townhome build-to-rent is the region's vernacular, what renters visibly choose and what exit buyers pay tighter cap rates to own, but it carries what Brooks called in an earlier case "more walls, more material, more everything," and Meridian has never put one in the ground; the Carverton hillside plan is still in entitlement and its budget is a lot budget, not a vertical one. A four-story elevator building beats both on paper and exists only if a rezoning that planning staff have not blessed comes through. Victor's three-column pro forma (Exhibit A) refuses to give one answer: the garden wins on yield, the four-story wins on development spread by two basis points and on total dollars, the townhome wins on profit per dollar deployed, and the townhome wins again on every qualitative dimension the columns cannot hold. The architect needs a decision before schematic design; every week of delay is carry. Choose the product, then defend the choice against the two columns you rejected.

Cast of Characters

Party	Role in the case
Meridian Residential Partners	Developer holding a 10.0-acre Baymouth-area parcel under contract at $5.0M ($500,000/acre), sourced through the Harbor Line relationship built in the entry campaign, zoned 16 du/ac; it sits inside the Urban Overlay within half a mile of the interstate, so 24 du/ac is a straight rezoning rather than a PUD, plausible and unproven.
Daniel Brooks	Chief Investment Officer; must approve one prototype before design dollars are spent.
Victor Amani	Development Associate; author of the three-column pro forma (Exhibit A). The first building in a market he chose will be judged on the column he recommends.
Martin Boyd	Project architect; advocates the four-story building if density can be won.
The market	A region whose vernacular is the townhome; whose renters pay up for attached product with garages; and whose institutional buyers pay tighter cap rates for it at exit.

Part I: Three Products, Three Businesses

The three prototypes are different businesses that happen to share a parcel. Garden multifamily is a manufacturing business Meridian already runs: surface parking, walk-up stairs, $150,000 per unit of hard cost, a leasing machine tuned to 950-square-foot units, and a national buyer pool at exit. Townhome BTR is a hybrid: single-family construction methods (more foundation, more roof, more exterior wall per unit), lower operating cost because tenants behave like homeowners, deeper renter demand at $2,395 for an attached three-bedroom with a garage, and the region's most liquid exit. Against that, it carries a cost structure the firm has never estimated from its own completed projects and a fundraising story Brooks has already flagged: "it's not a great environment to go raise money for something you've never done before." The four-story building is a density arbitrage: it exists only above 16 du/ac, spreads the land over 240 units, and imports elevator, corridor, and podium-adjacent costs the other two avoid. In The Hillside Closing, one further fact makes the choice concrete: contract language follows product, and the seller's draft on Meridian's Carverton deal defined the map around the wrong product entirely.

Part II: What the Columns Say (Exhibit A)

Read the exhibit twice, because it reverses itself. On untrended yield on cost, the garden wins: 6.79% against the townhome's 6.76%, with the four-story's 7.03% inadmissible until a rezoning exists. This is the "more walls" intuition made numerical:

Townhome NOI advantage per unit: $21,122 − $15,058 = $6,064
Townhome cost disadvantage per unit: $312,505 − $221,750 = $90,755

The NOI advantage is fully consumed by the cost disadvantage. But yield on cost is only half of development profit; the other half is the exit cap rate, and the market pays 25 basis points tighter for BTR. On development spread, the ranking flips: the townhome's 151 basis points beats the garden's 129, and the four-story's 153 beats both. Two basis points is inside the noise, and Chapter 8 says what to do about that: rank on value over cost, where the order is townhome (1.288), four-story (1.278), garden (1.235). A third ranking sits in the dollar row, where the four-story's $15.5M beats the townhome's $9.9M only by deploying $55.8M against $34.4M. Stop at yield and you build the garden; price the exit and you build something else; price entitlement probability and you may build the garden anyway.

Part III: The Rows the Table Cannot Hold

Four considerations live outside the arithmetic. Execution risk: Meridian's garden cost number is drawn from six of its own completed projects; the townhome number is drawn from two market data points and a contractor's verbal; a 5% miss on townhome hard cost erases its entire spread advantage. Lease-up shape: 110 large units absorb differently from 160 small ones, and differently again from 240; The First Twelve Months, later in this series, shows what absorption misses cost. Capital: the LP knows the garden pro forma cold and would underwrite the townhome as a first-time product, with pricing to match. And the entitlement coupling: the four-story column is not a product choice but a bet on the rezoning analyzed throughout this series; choosing it means the prototype decision and the entitlement strategy become one decision, with one calendar and one kill criterion. Unit mix, finally, hides inside each column: the garden's 6.79% assumes a 15/55/25/5 mix of studios through three-bedrooms at the stated average, and shifting ten points from one-bedrooms to two-bedrooms moves the blended rent, the average size, and the parking count simultaneously. The prototype comes first; the mix is fought inside whichever column wins.

The Decision

Schematic design starts in three weeks; the architect prices each prototype's design fee differently and will not hold three teams on standby. Brooks wants the recommendation as a one-page memo: the chosen prototype; the two assumptions in its column most likely to be wrong and the diligence that would tighten them before design dollars are spent; the conditions under which Meridian would switch columns mid-entitlement; and, if the four-story is chosen, the entitlement kill date after which the fallback prototype proceeds automatically. Each column has a fact that would retire it. The garden is wrong if a regional BTR builder's hard bid lands within 5% of the townhome column's cost, because the cost disadvantage was the whole case against the townhome. The townhome is wrong if the LP prices a first-time product at more than the 22 basis points of spread it adds. The four-story is wrong if the pre-application meeting returns the answer staff gave 1249 Kempton alone. The memo is due a week from Friday, two weeks before schematic design starts. Write the memo.

Exhibit A: Three Prototypes on One Parcel

Three prototypes on the same 10.0 acres and $5.0M of land

	Garden multifamily (by right)	Townhome BTR (by right)	Four-story elevator (needs rezoning)
Density (du/ac)	16	11	24
Units	160	110	240
Average unit size (SF)	950	1,380	870
Monthly rent	$1,895	$2,395	$2,050
Rent per SF	$1.99	$1.74	$2.36
Other income (% of rent)	6%	3%	7%
Vacancy	6%	5%	6%
Operating expenses per unit	($7,600)	($7,000)	($8,400)
NOI per unit	$15,058	$21,122	$16,343
Non-land cost per unit	$190,500	$267,050	$211,560
Land per unit	$31,250	$45,455	$20,833
Total cost per unit	$221,750	$312,505	$232,393
Total cost ($M)	35.48	34.38	55.77
Untrended yield on cost	6.79%	6.76%	7.03%
Assumed exit cap rate	5.50%	5.25%	5.50%
Development spread (bps)	129	151	153
Value over cost (yield ÷ exit cap)	1.235	1.288	1.278
Profit on cost	23.5%	28.8%	27.8%
Profit in dollars ($M)	8.3	9.9	15.5

Assumptions: NOI per unit is rent × 12 × (1 + other income) × (1 − vacancy) − operating expenses. Non-land cost per unit is hard cost + 12% soft + 5% contingency + carry. For the garden column:

NOI per unit: $1,895 × 12 × 1.06 × 0.94 − $7,600 = $15,058
Non-land cost per unit: $150,000 × 1.17 + $15,000 = $190,500 (carried as a round $190,000 elsewhere in the series)

Carry per unit is $15,000, $15,500, and $15,000 respectively. Development spread is yield on cost less the exit cap rate. The four-story column is achievable only at 24 du/ac; by right it ranks nowhere. The townhome's lower vacancy and expenses reflect single-family tenant behavior. The four-story's rent per square foot premium reflects elevator product in a garden market and is the column's most attackable number.

Workpoints: every $5,000 of townhome hard-cost error carries $5,000 × 1.17 = $5,850 of total cost and moves its yield on cost roughly 12 basis points. Every 25 basis points of exit cap is worth roughly 4.5% of exit value (5.25 ÷ 5.50 = 0.955). Re-solving the four-story at 20 du/ac (200 units):

Land per unit: $5.0M ÷ 200 = $25,000
Total cost per unit: $236,560
Yield on cost: $16,343 ÷ $236,560 ≈ 6.9%

That keeps the four-story in the argument even under the compromise density from the Kostas case. Convert the townhome's 25 basis point exit advantage into dollars on its total NOI and set that against a 5% miss on a thinly sourced hard-cost number. A fourth column (three-story surface-parked, or a garden and townhome blend on the same parcel) is worth building before the memo is written.

Discussion

Brooks and Boyd are reading different rows of the same table. Brooks reads the yield row and the cost data behind it. The garden is the column the firm can underwrite from memory, with a lender and an LP who need no meeting to do the same. A first project in a new market is a proof point before it is a profit, and a proof point built on a cost number from two data points and a contractor's verbal is not proof of anything. Boyd reads the spread row. The exit buyer pays for the townhome, the garden's 129 basis points is the worst of the three, and the rezoning the four-story needs goes to the same planning staff Meridian is already in front of three parcels away, so choosing the garden means paying for an entitlement fight and then building as if it had been lost. On his reading the garden is not the safe column; it is the column that forfeits the density the firm is spending two years to win. The reader must decide which row governs a first deal in a new market, and whether an LP who knows the garden pro forma cold could be brought to underwrite anything else.