Parking Is the Unit Count

Nobody tours an apartment for the drive aisle. But the drive aisle may have decided how many apartments there are to tour.

On the site I'll walk through below, 200 homes at 1.5 stalls a home put down almost twice as much parking as building: 99,000 SF of stalls and aisles next to a 50,000 SF housing footprint. That's not a rounding error in the site plan. That is the site plan.

I know this one personally. On The Cameron in Tempe I did nearly 20 iterations to solve the deal. Years later I finally did the math: 502 stalls in the garage and maybe 100 on the surface, about 600 stalls per iteration. Twenty iterations, 12,000 stalls drawn, one real building.

Settle the parking at concept, because it sets the unit count before any unit plan does.

Stalls are homes times a ratio

The arithmetic is short, which is why it's so easy to ignore.

Required stalls are homes times the parking ratio, rounded up. Each stall also needs its share of aisle and drive. My study folds that into one allowance, 330 gross SF per stall including circulation. Stack the stalls on more than one level and the footprint divides:

parking footprint SF = homes × stalls per home × 330 ÷ parking levels

The ratio is local. Your jurisdiction and how people actually reach the project set it, and I won't pretend there's one national number. The 1.5 here is the study's assumption, not a recommendation, and 330 is an allowance, not a stall layout.

One site, worked out

The study uses a 200,000 SF site (about 4.6 acres), 200 homes at 1.5 stalls a home, and housing at 850 net SF a home, 85% efficient, on four floors.

• 200 × 1.5 = 300 stalls. 300 × 330 = 99,000 SF of parking on one level.
• 200 × 850 ÷ 0.85 ÷ 4 = 50,000 SF of housing footprint.
• 99,000 + 50,000 = 149,000 SF, leaving 51,000 SF of site.

Per home, that's 250 SF of building footprint and 495 SF of parking. Every home you add costs land twice, and the parking bill is the bigger one.

Now put the same 300 stalls on two levels. The parking footprint halves to 49,500 SF. The housing doesn't move. The land left nearly doubles, from 51,000 to 100,500 SF. Nothing about the housing changed, and the parking decision handed back a quarter of the site.

It hands back homes, too

Push the home count up until the leftover hits zero.

• One level of parking: the site tops out between 260 and 270 homes. 260 leaves 6,300 SF; 270 runs 1,150 SF past the property line.
• Two levels: 400 homes fit with 1,000 SF to spare.

Try it: drag the housing bar, the stalls-per-home bar and the bottom of the parking stack, and watch the leftover land against the red site line.

Same site, same ratio, same buildings. The parking levels moved the ceiling from about 260 homes to 400.

The ratio moves it too. Stay on one level and drop to 1.0 stall a home, and the ceiling rises to 340. Raise it to 2.0 and it falls to 210 (counted in steps of ten homes). On surface-parked garden deals, where every stall lands on grade, the parking field, not the buildings, usually decides how many units the site holds.

Those ceilings aren't yields. A zero remainder leaves nothing for setbacks, the drive to the street, open space or a stormwater basin, so the real count is lower. The point is which term sets it.

What stacking costs

Stacking trades land for structure. The study keeps cost per stall as its own assumption: $8k a stall on one level, $30k on two, $45k on four. Those are placeholders, not quotes. On them:

• 300 stalls cost $2.4M on one level and $9.0M on two. That's $6.6M more to recover 49,500 SF, about $133 per square foot of site.
• Four levels at $45k cost $13.5M and recover only 24,750 SF more than two, about $182 a square foot.

The geometry explains the slide. The footprint is 99,000 ÷ levels, so the second level gives back 49,500 SF, the third 16,500 and the fourth 8,250. The first level you stack does most of the work. After that you're paying more for less ground.

Push it to the extreme: 400 homes at 2.5 stalls on one level needs 330,000 SF of parking, and the site is 230,000 SF short. Four levels pull it back to 82,500 SF with room to spare, at an assumed $45M of parking. The land problem became a money problem.

Whether $133 a foot of site is a bargain depends on what the land costs and what you'd put on it. That's a local answer. The structural answer isn't local: the first deck is the cheap one.

A stall is a room for a car

It helps to stop treating parking as paint on asphalt. A stall is a space sized for one occupant, a car, with its own dimensions and a route in and out. The aisle is circulation. Ramps are stairs.

Seen that way, a parking structure is a building with its own program. My study puts two 60 ft deep plates side by side: residential modules 24 ft wide on a 6 ft corridor, and parking with its own stall width, stall depth and aisle. Whole stall rows have to fit, just like whole units do. Stall geometry pushes column placement and the shape of the plate, and those choices then fight with whatever housing or retail sits above or beside it.

That's a design lens, not a claim that cars and people need the same things. Its value is making the area commitment visible next to the homes it serves.

Then the water takes its cut

Every square foot of roof, parking and drive you draw makes water. Rain that used to soak in now runs off, more of it and faster. In a regulated jurisdiction the rules may require the site to detain, infiltrate or treat that runoff before it leaves, usually by holding it and letting it out slowly. So the basin isn't a civil detail added later. Stormwater is the first tenant on the site, and it doesn't negotiate.

My stormwater study uses the same 200,000 SF site. Runoff is rain depth times area times a volumetric runoff coefficient: 0.9 for roofs and paving, 0.2 for lawn. A detention pond on its own land has a 4 ft working depth, and its outlet passes 2,160 ft³ while the rain falls (an assumed 0.1 cfs for 6 hours). Whatever the pond can't hold and the outlet can't pass goes over the rim.

• 1.5 inch storm, 80,000 SF of hardscape, a 6,000 SF pond: 11,850 ft³ of runoff. The pond is 40% full.
• 3 inches: 23,700 ft³. The pond is 90% full with 2,460 ft³ of room left.
• 4 inches: 31,600 ft³. 5,440 ft³, 17% of the runoff, goes over the rim and off the site.

Now hold the storm at 3 inches and change the site instead. Go from 80,000 to 120,000 SF of roofs and paving and runoff rises to 30,700 ft³. The pond that was 90% full overflows by 4,540 ft³. The storm didn't change. The site plan did.

Try it: drag the rain gauge, the edge of the paving bar, the pond wall and the pond floor, and watch the water rise and spill off the site.

A bigger pond fixes it and takes land to do it. A 12,000 SF pond at the same depth is 59% full in that storm, and it costs 6% of the site. Then assume an 8 inch 100-year storm (the real depth depends on the place) and even the bigger pond floods: 30,907 ft³, 38% of the runoff, over the rim. A pond is sized for one storm. Where the overflow goes is a site-plan decision too.

Parking and water are the same decision

Here's why these belong in one essay. Stacking parking doesn't just hand back land for buildings. It takes hardscape off the site, and hardscape is what fills the pond. In the stormwater study, the difference between 80,000 and 120,000 SF of paving is the difference between a pond with room left and one that overflows.

Halving the parking footprint on our 200-home site takes 49,500 SF of paving out of the equation. That shrinks the pond too.

Your options for the water are the same shape as your options for the cars. A surface basin takes land. An underground vault under the parking stores water beneath a surface you were paving anyway, and you pay for it in excavation, access and maintenance. Green infrastructure, bioswales, permeable paving, green roofs, attacks the coefficient instead of the depth, and local standards decide how much credit each earns. Water quality is a separate requirement with its own space. Either the basin takes land or it takes budget. It doesn't take nothing.

And the permit sequence makes all of it early. A project that discovers its stormwater approach is wrong halfway through review pays for the same site plan twice. Draw the basin before you draw the building.

The site has to work for people, too

One more constraint that belongs at concept, even if it's mostly an industrial one. A warehouse or plant needs workers as well as land and trucks. The labor shed, the area a facility can draw employees from, can explain why a site works for one operation and not another.

My labor-shed study is deliberately crude: a circular commute radius on a uniformly populated plane, compared against 300 positions. Real labor markets have barriers, congestion, competing employers and skill gaps that a circle ignores. The useful point is narrower. A building can be easy to build and hard to staff, and parcel price alone won't tell you which. Test the shed against the real shifts and skills, not population totals.

The test fit is the due diligence

Put it together and you get the real job of a test fit: checking whether a site can support the program that justified its price. Parcel, zoning envelope, structure, mix, parking and water, in one early drawing.

Parking is often the binding constraint even when floor area looks available. If the assumed count doesn't fit, the land decision needs another look before anyone commits to it. A test fit isn't entitlement and it isn't final yield. It tells you what deserves a deeper look before you buy.

The line this argument started as, "the site plan is a parking decision," overstates it. A site plan is also a stormwater, access and code decision. But in this example, parking is the decision that eats the most ground, and water eats what's left.

Park it, drain it, then build it.

Drag the ratio, the parking levels and the pond on your own parcel before you price the land. Book a TestFit demo and run this on your own site.