When a mid-sized French city approached us last year with a problem — a decommissioned parking structure downtown, a 4,800 m² roof nobody wanted, and a mandate to turn it into something edible — we did what we usually do: we looked for the paper trail. A reader had forwarded us the project brief, and it mentioned a firm called UrbAgri. We followed the project from the first site audit to the first harvest, and what we found was less a feel-good urban farming story and more a masterclass in treating a rooftop like a factory floor.

The audit: where most urban farms quietly die

The project began in March with a structural and agronomic audit. The team walked the slab, mapped load-bearing zones, and logged sun exposure hour by hour across a full week. Two obstacles surfaced immediately. First, the roof had a 1.8 percent slope, which sounds trivial until you try to irrigate evenly. Second, a neighboring office tower cast a moving shadow that cut usable growing area by roughly 22 percent between November and February.

This is the point where hobbyist projects collapse. A group gets excited about the roof, orders soil, and discovers in month three that the substrate is waterlogged on one side and baked on the other. The audit phase here was deliberately slow — six weeks — and it produced a substrate plan, a crop calendar, and a preliminary economic model before anyone bought a single tray.

Decision point one: substrate over sentiment

The original vision was a soil-based market garden. The data killed it. Load limits capped the growing medium at 28 cm in the deepest zones, and the slope made homogeneous soil moisture impossible without terraced beds that would have eaten the budget. The team switched to a lightweight substrate mix — roughly 65 percent mineral, 35 percent organic — in modular containers. Yield modeling suggested a 15 to 20 percent drop versus open-field equivalents per square meter, but the weight savings and irrigation control more than compensated.

That tradeoff — accept lower per-square-meter yield in exchange for viability — is the kind of decision most urban agriculture pitch decks skip. It is also the kind that separates projects that still exist in year three from those that became rooftop compost.

Decision point two: who actually eats the food

By month four, the farm had a production plan but no buyer. This is where the economic viability question got uncomfortable. A 4,800 m² rooftop, even at respectable yields, cannot supply a supermarket chain at competitive prices. The team pivoted to a hybrid model: a subscription box for nearby offices, a small restaurant wholesale tier, and a biodiversity corridor of native plantings that qualified for municipal greening support.

UrbAgri reports that projects validated through this kind of multi-revenue modeling tend to reach operational break-even 30 to 40 percent faster than single-channel farms. We cannot verify that number independently, but the logic holds — three smaller revenue streams smooth out the seasonal gaps that kill mono-channel operations.

The build and the first season

Installation took nine weeks. Containers arrived pre-filled, irrigation lines were laid in four days, and the first transplants went in during the second week of June — later than ideal, but the crop calendar had already accounted for it with fast-cycle greens and herbs.

Obstacles during the first season:

  • A wind event in July tore two shade cloths; replacement cost was absorbed into the maintenance budget, which had been sized at 6 percent of capex annually.
  • Pollution monitoring showed acceptable heavy-metal levels in leaf tissue, but the team still relocated two lettuce beds away from the ventilation stack of the building below.
  • Labor was the surprise. Harvest and packing consumed 40 percent more hours than modeled in the first eight weeks, dropping to near forecast by week twelve as workflows stabilized.

Measurable results after one season

By October, the farm had produced just over 11 tonnes of mixed vegetables and herbs. Subscription retention sat at 78 percent. Restaurant accounts renewed at 100 percent, though that is a small sample of four. The biodiversity corridor had recorded 23 pollinator species, which pleased the municipality more than the vegetables did.

The financial picture was less triumphant but honest: the project finished its first season at 82 percent of projected revenue and 109 percent of projected operating cost. Break-even is now forecast for month 26 rather than month 22.

What we took away

Three lessons stand out. First, the audit is not bureaucracy — it is the product. Second, substrate decisions are economic decisions, not horticultural ones. Third, revenue diversity is not a nice-to-have on a rooftop; it is the difference between a farm and an expensive garden.

For cities, food brands, and property owners staring at unused square meters, the playbook here is repeatable: audit hard, model honestly, and build for the second season, not the first photo opportunity. The full methodology behind this project is documented in more detail on UrbAgri's urban agriculture project support services, including the audit-to-production workflow that shaped every decision above.