A plant plan, a pipeline promise, and a front end nobody had designed
The company was spinning off into modular housing: volumetric units designed, built and set from its own factory line in Arizona, aimed at workforce and low-income housing, with a new plant coming online. They engaged me, through Common Ground, for an initial analysis of the business model, drawing on the years I spent running a design-build general contractor and the preconstruction work I do now. The brief pointed at the front end of the plant, meaning how it reads a lead, talks to a developer, frames an early budget and walks a client through the first phase of a project. It was framed around a hypothetical affordable multifamily project, and the obvious deliverable was a tidy memo that walked each step in turn.
The people in it were the company's chief executive, who originated most of the leads through his own network and wanted the front end designed before he chose who would run it; the production side of the plant, which would have to build whatever the front end sold; and, in the worked example, a developer new at that scale who wanted a full price fast against an all-in per-foot target.
What they actually needed was different from what the brief asked for. A factory is a fixed cost with a line that has to stay fed. Its margin does not come from winning a lot of leads. It comes from three conditions holding at once: the line runs full, the work on it does not change after it is released to production, and the plant never promises a delivery it cannot make. Every one of those conditions is set before a single box is built, in the first calls, the first number, the first comparison a developer sees. Unqualified developers eat estimating hours. Budgets compared on mismatched scopes push a factory to chase a number it cannot build to. Design changes after release destroy modular economics at the worst point in the schedule. A sales team that oversells the line burns the plant's credibility faster than any competitor could. All four are decided at the front end, and the company had a plant plan and a pipeline promise but no design for the front end at all. My read of the modular builders that have failed in public is that most of them failed there first, on the promise, long before the line had a chance to prove itself.
A hypothetical cannot break, so I ran a real project through the plant
I had two ways to answer. Write the memo around the hypothetical project, which would have been faster and would have said all the right things. Or take a real project and run it through the plant as if it were live, so that every recommendation had to survive real plans, a real site and a real local trade market. I chose the second, because a model for a front end is only worth something if it holds under constraints, and a hypothetical has no constraints to fail against.
The project was a two-story rental community of about 200 units that I had bid before, with concept plans that disagreed with themselves on unit area and parking count, a small market about an hour from the nearest metro, and a county that had let an approval on a comparable parcel expire on a missed submittal. I told the whole thing as one engagement: the thin lead from the owner, my first email to the developer, the developer's reply asking for a full price fast against a per-foot target, the estimate, the schedule, my recommendation to leadership, and a fully drafted feasibility agreement. The core of that package was built the same day the brief arrived.
Three decisions followed from the choice to use a real project, and they are the reasoning in this report. Split every budget into what is identical no matter who builds the units and what actually differs, so the comparison happens on the only scope where the methods compete. Lead with schedule, because schedule is the one thing a factory does that a site-built contractor cannot, and at this scale carry outweighs a small difference in construction cost. And put a paid step between the free estimate and the plant's real work, because a plant's estimating hours are inventory, and inventory given away is priced at zero by the people who take it.
How I came at this one
The question I asked first was what has to be true for a factory to make money. That question fit because a plant is a fixed cost with a line that has to stay fed, and every condition that keeps it fed is set before the first box is built. The second question was who pays for the plant's thinking: a free step attracts browsers, a paid step aligns the incentives. The third was whether the number in front of me was a source or a memory: write the check down and run it every time.