The work, decision by decision

The work, gate by gate

1. Run it live, on the real project. What was known: the brief was hypothetical. The question I asked was whether a hypothetical could show how the model holds under real constraints, and the answer was no. So the whole response was built on the real rental community, told as one engagement from the first lead email to a signed feasibility agreement. What it changed: every recommendation had to survive real plans, including their flaws. The concept set disagreed with itself, and the response had to say which figures it carried and why.
2. Map the front end as a stage-gated pipeline. What was known: a lead arrives thin, usually a plan set and a sentence. The question was where the plant should spend estimating hours and where it should stop. I laid the front end out as a stage-gated pipeline, the figure on the cover: intake and a first call, a free rough estimate, a paid feasibility package, then a preconstruction handoff, each with an entry test, a named output and a gate decision. What it changed: the question stopped being "how do we win this lead" and became "what does this lead have to show us before it gets the next hour of our time." The discovery questions that sort a project from a shopping trip are in Attachment A.
3. Reframe the comparison before any number exists. What was known: developers shopping modular carry two or three budgets built on different scopes, and an all-in per-foot target with nothing defined inside it. The question was how to compare a factory-built unit to a site-built one honestly. The answer was to split every budget into common costs, the work that exists no matter who builds the units, held at parity across every bidder, and vertical delivery, the finished unit above the slab, where stick framing, panel systems and volumetric modular actually compete. My first email to the developer planted that frame before the estimate existed, and offered to map the competitor's quote onto our scope sheet at no charge. What it changed: the conversation moved off the sticker and onto the one scope where the delivery methods differ, and onto schedule.
4. Build the estimate, then check how it was made, not what it said. What was known: the number had to be fast and had to be defensible. I ran the quantity takeoff twice, blind, and reconciled the passes on area, unit count and module count; both converged exactly. I stated the band, plus or minus 15 percent, and the exclusions on the face; carried the factory line as a labeled placeholder until the plant confirmed its own pricing; and put contingency only on the site-built scopes, where the local trade market creates the risk. The check that mattered most was on me. On the first pass, a comparison figure I recalled from an earlier project drove a rate that had been tuned so the total would land in a competitive-looking range. Once I found the original proposal document, the recalled figure was high by double digits. The total had not looked wrong; that was the danger. A number built to match a belief is not caught by looking at the output. It is caught by checking how it was made. What it changed: source before memory became a written rule, and never tune an input to hit a target became the rule beside it. Both are in Attachment A.
5. Lead with the schedule, and gate the detail behind a paid feasibility phase. What was known: at this scale, carry, interest and rent not yet collected outweigh a small difference in construction cost. The question was how to turn a free estimate into a commitment without giving the plant's work away. The schedule led every page, because units are built on the line while the site is graded and the two paths meet at set. Everything deeper than the rough estimate sat behind a fixed-fee feasibility agreement, nonrefundable once work starts and credited in full against a build contract signed inside a stated window. The package adjusts the plans to factory tolerances, reworks slabs and utilities for set, maps the approvals pathway and ends in a written go or no-go. What it changed: the estimate became the doorway, and the fee became the test of whether the developer is real.
6. Write down how the front end connects to the line. What was known: the company had defined what its front end does and how it is measured, and that definition said little about production. The question was what whoever runs the front end has to know about the factory to avoid selling what the line cannot build. I took the company's definition as written and added operating expectations, the questions to answer in the first weeks, and working assumptions about who originates work and who owns the pipeline. What it changed: the front end was defined as the protector of the line, not only as a sales function.
7. Test the business behind the plant. What was known: a front end can run perfectly and the plant can still fail, because the front end only feeds the line. The question was what else had to be true for the business to work in its first year. I tested the model from a plant operator's seat, a housing-finance seat, a tax-and-incentives seat, a labor-cost seat and a technology seat, each asking its own question of the plan. Those questions are the closing box of this report. What it changed: the front-end work stopped being a sales plan on its own and became one part of a business that had to answer every one of them before the line ran full.

Schedule first, then the honest split

The schedule is the figure a developer can check without trusting me. Price is where a factory looks most like everybody else; schedule is where it is different. The mechanism is parallel work. Design freezes and a production slot is reserved, the line builds units indoors while the site is graded and foundations go in, and the two paths meet at module set. Weather and deliveries stop driving the calendar, and progress is documented at each factory inspection where a third party can see it. Every month saved is carry, interest and rent collected sooner, which at the scale of the worked example outweighs a small difference in construction cost. That is why the comparison in the estimate was built to be read schedule first and price second, and why the first email to the developer led with time and transparency rather than a number.

The split chart counts scope lines rather than dollars, on purpose, because the point is the structure. The common bucket is the same for everyone. A bid that looks dramatically cheaper there is usually missing scope the developer will buy later at change-order prices, so the estimate holds it at parity and moves the argument to the vertical, where the modular path adds factory, transport and set and takes the field framing out. Where the plant has real advantages inside the common bucket the estimate names them rather than burying them: factory inspections replace field framing inspections, the line can run extended hours without site labor premiums, and there is no travel or mobilization inside the factory scope. The competing system is described by its structure and never by name; the critique is structural, not personal. Two delivery-model choices were flagged as design items rather than decided quietly. Delivering the drawn attached units as detached two-story units kept the factory product standard and the state approval clean, and put the cheapest square footage on the second floor. A box width at the edge of a legal haul was flagged too, because if the plant caps its box narrower, the module count doubles and mate-lines are added.

The recommendation back to leadership, filed with the estimate, was yes on fit (one envelope, two fit-outs, highly repetitive boxes, a thin local labor market that favors factory work, and a known approval pathway) and yes on modular, on schedule grounds, with the feasibility agreement as the next step. That does not make it a dead lead. It means the door stays open and the work stays paid for. One question went to the team: whether the plant's trade bench could reliably cover a site that far from the metro.

What stayed as the company wrote it, and what I added

Kept: the company's own definition of the front end and the measures it had chosen for it, early-estimate accuracy against final budgets, win rate, client satisfaction in the early phases, fewer scope gaps at handoff, and the speed and clarity of project definition. I took that as written and built under it. Kept: lead generation through the chief executive and the existing network, with the front end's job starting the moment a lead exists.

Added, and delivered as analysis for leadership to adopt rather than installed by me: the stage-gated pipeline with its entry tests and gate decisions; the paid feasibility phase as the qualification gate, with a hard design-freeze gate before a production slot commits; the common-cost and vertical-delivery split on every budget; the four checks every early estimate passes before it leaves the plant; and a set of operating expectations for whoever runs the front end. Know the factory the way production knows it. Build the trade-by-trade library of where modular wins and where it does not. Know where modular breaks down, for the client and on the line. Never overpromise deliveries as the pipeline grows. Treat early estimates as the contingency-heavy documents they are, tighter where the line controls the outcome and wider where the site and the local trade market do. Start design-build conversations earlier than the intake flow requires, so production hears about a project before it is sold. The full operating document is Attachment A and the risk register is Attachment B.

What was faulty in the original logic, and why it had to change now: the front end had been defined as a sales function, measured on wins and estimate accuracy, with nothing in it that tied the promise to the line's capacity. That definition is fine for a contractor whose capacity is a subcontractor phone call away. It is fatal for a factory, where capacity is fixed and the cost of an unfillable promise lands on production. It had to change before the first sale, because the first overpromise is the one that sets the plant's reputation with exactly the developers it most wants.

What it costs to hold the line, and what I would watch

Saying no to a full price on demand costs a few leads that were only ever shopping, and a few that were real but impatient. Printing the band and the exclusions on the face of every estimate costs the false comfort of a clean total. Source before memory costs an hour under a deadline, every time, and the deadline is when the shortcut is most tempting. Charging for feasibility in a market where competitors give it away costs some conversations. I would pay all four again, because the plant's estimating hours are inventory and the fee is the cheapest test of whether a developer is real.

The part nobody asked for was a question I put to leadership at the end. I built the core of that package in one afternoon, with tools that did not exist in this form a year earlier. When every shop can fold feasibility-depth analysis into a free estimate, and they will, the paid feasibility phase stops being a revenue line. It becomes a filter. Charge the fee because a developer's willingness to pay it is the test; do not build the plant's economics on it, and do not give it up either.

What I would watch, from the register: the developer's capital stack, so no engineering leaves the plant outside the paid phase; the destination state's acceptance of the plant's home-state factory inspection, which is discretionary, so it is confirmed with the state program in the first week of feasibility and the per-module insignia is priced either way; the trade bench for the site-built scope, covered by a visible contingency on those scopes only and early subcontractor commitments; the design freeze against factory release, held by catalog-first unit selection and an owner sign-off milestone before the slot commits; and the county's conditional-use and site-plan deadlines, owned in the master schedule with every milestone mapped. Two items stayed open on the estimate rather than in the register: the water utility's intent-to-serve letter, which gates the utility scope, and the developer's required completion date, which decides whether the schedule advantage is decisive.

At the plant level the watch list is the physics of the line. Throughput per line per week against the pipeline being sold, because a full pipeline the line cannot deliver is worse than a thin one. Labor per module once wage rules are settled, because that is where the factory's cost advantage lives or dies. The transport radius, because the margin on a box dies quietly somewhere on the highway, and the trade bench at the far end of that radius, because the site-built scope still has to be built by someone who will show up. Those are the questions the analysis ended on, and they are left as questions.

What it produced

Delivered in one week. Built on a real project, a two-story rental community of about 200 units that had been bid before, run end to end from a thin first lead to a drafted feasibility agreement. On the worked example the factory path ran about 14 months against 18 to 19 months site-built; a comparison figure recalled from memory drove a rate tuned toward a competitive-looking total, caught high before the number shipped. The recommendation to leadership was yes on fit and yes on modular, on schedule grounds. The stage-gated pipeline, the paid feasibility gate, and the common-cost and vertical-delivery split went to leadership as analysis to adopt, not as something installed.

A slice of the project list

A few related projects.

• A PE-backed medical group: operating audit (2026)
• A property management launch (Phoenix, 2023 to present)
• A 204-unit multifamily: value engineering (Nevada, 2026)
• Canyon Corporate: estimating and preconstruction