The Building Is Priced at the Interfaces

A duct 6 inches bigger costs 20 inches of height. That's not a typo. In my floor-to-floor example, swapping a 10 in duct for a 16 in one means it no longer fits through the beam, so it drops below it, and the whole section grows by 20 inches.

That one line explains a lot of what I've learned about construction cost. Buildings don't get expensive in the middle of a system. They get expensive where two systems meet: the duct and the beam, the column and the wall, the podium and the wood above it, the steel delivery and the framing crew, the owner and the contractor.

This essay walks through those interfaces, with the arithmetic.

The section, inch by inch

Floor-to-floor is a stack of layers: finish floor, slab, plenum, ceiling, clear height. The plenum is the structure under the slab plus the services under that. Change any layer and the others move, along with the building's total height and how many floors fit under the limit.

The worked example below uses stated assumptions, not a design: a 2 in ceiling assembly, a 1.5 in finish floor, 1 in of insulation each side of the duct, 2 in between structure and duct, 3 in between the lowest service and the ceiling grid.

• Tight PT slab. A 7 in post-tensioned slab has no beam, so the 10 in duct hangs right under it. Plenum 17 in. With a 108 in clear height, the floor-to-floor is 135.5 in, or 11.29 ft. That's the reference.
• Steel beam, duct below. A 24 in beam under 6 in of deck and topping sends the duct underneath. Plenum 41 in. Floor-to-floor 158.5 in, 23 in more than the PT slab.
• Duct through the beam. Cut a web opening and the same duct passes through. Plenum 27 in, floor-to-floor 144.5 in. The example only allows the opening when the insulated duct is no more than half the beam depth. That's my assumption. The structural engineer sets the real limit.
• Higher ceiling. Raise the clear height from 108 to 120 in and the floor-to-floor goes up the same 12 in, to 156.5 in. Clear height adds inch for inch.
• Bigger duct. The 16 in duct's 18 in insulated height won't fit through a 24 in beam. It drops below. Floor-to-floor 164.5 in.
• Tall retail podium floor. A 144 in clear height, 12 in slab, 30 in concrete beam and 20 in duct give 216.5 in, or 18.04 ft. The plenum alone is more than three times the PT slab's.

What an inch costs

Now price it. The example assumes a 120 × 80 ft footprint, so 400 ft of perimeter, 12 floors and $110 per SF of facade.

One inch of floor-to-floor is 33.3 SF of facade a floor: $3,667 a floor, $44,000 across all 12 floors.

• Steel with the duct below adds 23 in against the PT slab: about $1,012,000 of facade.
• Threading the duct through the beam cuts that to 9 in, or $396,000. One web opening is worth about $616,000 of skin.
• The podium floor adds 81 in, but to one floor only: 81 × $3,667, about $297,000 of facade.

Those numbers count facade only. Structure, shafts, risers, elevators and everything else that grows with height come on top.

The original slug for this thesis said the GC sets the floor-to-floor. The body never supported that, and it isn't true. Contractors and trade partners can tell you whether a section is buildable and what it may cost. The architect, the structural and mechanical engineers, the owner and the code all shape the decision. The section is a shared decision, and the drawings should make its price visible to everyone at the table.

Construction type is a financial decision

Zoom out from the section to the whole structural system and the interface gets bigger. Construction type ties structure and fire-resistance strategy to the height, area and uses the code allows. It also drives cost, schedule and which materials and trades are practical.

A wood frame versus concrete or steel isn't a material-price comparison. Height, site, fire provisions, spans, labor availability and local contractor experience all vote. A podium puts two structural systems in one building and adds coordination right at the transition, which is the expensive interface you just saw in the section.

Early comparisons are useful only when every option carries the same program, schedule and scope. Generic cost ranges and story limits don't replace project-specific code review and pricing.

Columns should land on walls

The column grid and the room layout are two systems that have to share one plate. A column inside a room interrupts it. A grid that matches the rooms lets planning and structure work together.

My rhythm study uses a 168 × 60 ft plate with two rows of 30 ft deep rooms. A 28 ft bay holds two 14 ft rooms with the columns meeting their shared wall. Change the room width, the bay width or the grid offset and the relationship breaks. Matching dimensions isn't enough. The bay and the room module have to align in position, too. The study marks every column that lands inside a room and every foot of frontage left unassigned.

Try it: drag the bay width, the room width and the grid shift until every column lands on a wall.

Different uses want different grids. My transfer study puts two grids with the same spacing on a 108 × 60 ft plate, offset from each other, and draws the transfer distance from each upper support to the nearest lower line. That's the residential-over-parking problem in miniature. Stalls and apartments want different modules, and the transfer between them has structural and cost consequences that need real analysis. There's no universal bay or slab that suits every project. The job is getting the architect and the engineer to test the same room and loading assumptions together.

Every trade is an interface

Each trade has defined work and an interface with the trades before and after it. Framing builds the structure. MEP routes through it. Finishes depend on everything beneath them. The schedule follows those dependencies, so a delay in one scope rolls into the next.

My trade study is deliberately tiny: one hypothetical section with a beam and a duct sharing a vertical zone and a stated 2 in separation. That's a clash check on one interface. Overlaying trade models and reviewing clashes is integration testing for buildings, and it's a lot cheaper to fix a conflict in drawings than after the adjacent work is in place.

I've used the microservice analogy for trades before. It's good for handoffs. It breaks down because trades share space, sequence and responsibility. They can't be treated as independent packages. That's exactly why the interfaces are where the money goes.

The longest lead sets the schedule

Some interfaces are in time instead of space. A building's completion depends on components sourced, fabricated, delivered and installed in sequence, and the longest-lead item can govern the schedule even when everything else is ready.

My procurement study is an explicit illustrative network: steel and glazing fabrication start together, framing waits for steel, envelope waits for framing and glazing, interiors follow envelope, closeout follows interiors. Durations are assumptions, not a forecast. Elevators are the classic long-lead item. Prefabrication moves assembly to a factory and adds transport and lifting limits. Both choices move dependencies around the supply chain. Neither removes them.

So procurement is part of design, not something that starts after drawings are done. And the schedule should state its lead-time assumptions instead of treating last year's durations as a promise.

The contract decides who keeps the difference

The last interface is the one between the owner and the contractor. A bid reflects a proposed scope, the subcontractor market, the schedule and the contractor's view of risk, at one moment. A bid is a market signal, but only when you understand its scope and its terms.

My contract study prices one illustrative project three ways: lump sum, cost-plus-fee and a guaranteed maximum price. Trades are $20.00M, general conditions $1.20M, with a 3% contingency and a 5% fee on top, so all three start from the same signed price. None of it is market data.

• On budget. The owner pays $22.90M lump sum, $22.26M cost-plus, $22.58M GMP. The contractor's margin is $1.70M, $1.06M and $1.38M. In the lump sum the margin is buried in unit prices. In cost-plus it's the fee, shown openly.
• 5% underrun. The lump sum becomes the contractor's best contract: the owner still pays $22.90M and the margin rises to $2.70M. Cost-plus passes the saving through, and the owner pays $21.26M. The GMP splits it: $22.08M and a $1.88M margin.
• 10% overrun. Under the lump sum the contractor's margin goes to −$0.30M, a loss. Under cost-plus the owner pays the whole overrun, $24.26M. The GMP cap behaves like the lump sum.
• 8% contingency instead of 3%. The signed price rises to $23.96M for the same work. Under a lump sum the unspent contingency becomes a $2.76M margin that never appears as a line.
• 9% fee instead of 5%. Easiest to see under cost-plus: $23.11M, with a $1.91M fee shown plainly. The lump-sum owner pays $23.74M and can't see the fee at all.
• 8% underrun, contractor takes 25% of savings. $2.24M is saved under the GMP cap. The contractor keeps $0.56M and the owner pays $21.22M. A lump sum would give the contractor all of it, a $3.30M margin.

Try it: drag the dashed cost line, the fee, the contingency and the savings split, and watch the owner's bill and the contractor's margin move under all three contracts.

No contract is good in the abstract. The same building at the same cost produces very different checks depending on which interface you sign.

When bids blow the budget

If bids come in over budget, redesign might be the answer. First diagnose the gap. The scope may have grown, the estimate may have been optimistic, the market may have shifted, or one trade may be driving the number. Different causes call for different fixes.

Normalize scope before you treat the lowest number as the best one. Alternates, exclusions, contingencies and schedule assumptions matter as much as the total. And date every number, so a changed market doesn't get blamed on a changed design.

Draw the joints first

Every system in a building is designed by somebody competent. The money leaks out where those systems touch, and nobody is assigned to the touching.

Price the joints, not just the parts.

Test your section, grid and structure against your unit plan before the bid shows up. Book a TestFit demo and run this on your own site.