The signal belongs to the neighbourhood, not the company

The people who can lease you farmland for a solar park are mostly invisible. Many have no website. Several have no published email address. They are in the commercial register, they farm several hundred hectares, and no sales tool built for software companies can see them at all.

The moment worth catching is not “this landowner wants solar”. It is the year after a large project is commissioned nearby. The neighbour has watched the lease payments arrive for a season. The question has stopped being whether this is real and become whether they missed it.

Land leasing is a twenty to thirty year decision that a family business makes roughly once. Nothing in the owner's own behaviour announces that they are ready, because there is no search, no job posting and no software purchase. The readiness is created from outside, by a project they can see from their own fields.

So the timing evidence is not about the company at all. It is about the parish. A zoning plan in the same municipality, or a park that went live down the road, predicts receptiveness better than anything you can learn about the business itself.

Municipal zoning plans. A project-specific development plan (in Germany, a vorhabenbezogener Bebauungsplan) is published by the municipality during consultation, with the area in hectares and the intended use. It appears while the project is still being decided, which is the earliest useful point.

Trade press for commissioned parks. Capacity, hectares and commissioning dates for large installations are reported with enough precision to cite.

The commercial register. Agricultural businesses appear with their legal form, seat and registration number, which is often the only structured record of them that exists.

State-level capacity policy. When a federal state changes how much land it will allow for solar, every landowner inside it is affected at once. Mecklenburg-Vorpommern doubled its solar allocation to 10,000 hectares in April 2026.

Match the farm's Gemarkungen, the cadastral districts it actually works, against the area a nearby project covers. That is the join. It is geographic, not firmographic, and it is why the usual tooling cannot do this: the relationship you need is between a parcel and a plan, not between a company and an industry code.

Two projects make the point. The Klimapark Steinhöfel covers roughly 500 hectares with a 753 MWp installation and began construction in Q4 2025. A separate project-specific zoning plan, “Agri-Photovoltaikanlage Plath I”, covers 61 hectares and went to draft in July 2025. In both cases the useful question is not who owns the project. It is who farms the fields beside it.

First, we let the general-purpose prospecting tool try. Asked for landowners in eastern Germany it returned three things. Energy buyers. A farm we could not confirm exists. And a large agricultural trading company that is a direct competitor of the developer we were researching for. Not a tuning problem. The tool indexes companies that behave like software buyers, and a 400-hectare farm with no website is not in the index at all.

Then we got the direction of the signal backwards. The first three candidates were cooperatives with impeccable evidence. One had leased the land for a 49 MWp park. One had started a zoning plan of roughly 100 MWp. One had part of a 150 MW park on its ground. Every row cited a real, checkable source. Every row was useless. They had all already signed with a developer. The letter's argument was that they should consider leasing land, and the citation supporting it proved they had considered it and chosen someone else.

That is the trap in evidence-led prospecting, and it is worth stating plainly: a source that proves the need is real often proves it has already been met. The strongest evidence and the best prospect pull in opposite directions. What you want is mid-decision, not already-decided, and mid-decision is by definition less well documented.

This does not scale into a list you can buy. Every row is a manual join between a cadastral district and a planning document, and the reward for the work is a handful of names.

Contact data barely exists. None of the three businesses in the final set has a website. Rather than guess at a domain, the record carries none.invalid. An obviously invalid value is more honest than a plausible invented one, and a system downstream should fail loudly rather than mail into the dark.

The neighbourhood signal is also an inference about a person, not a fact about a company. It says the idea is live nearby. It does not say this owner wants it, and a letter written on it has to offer rather than assume.