SolarResult
Panels are the easy part.

The unglamorous half
Solar is sold as a product decision and it is almost never a product decision. Two identical arrays — same modules, same inverter, same orientation — will diverge by a quarter of their output inside a decade, because one was surveyed by somebody who went onto the roof and one was surveyed from a satellite photograph.
The industry has arranged itself around the easy half. There is enormous appetite for financing an installation and very little for owning what happens to it afterwards. Warranties are written against the panel, which almost never fails, and not against the mounting, the sealing, the string design or the monitoring, which is where the money actually leaks.
So the interesting position is the unglamorous one: be the party that is still liable in year eight. It is harder to raise against and slower to scale. It also happens to be the only version of this business that gets better with age rather than worse.

Four things, in order
Survey
On the roof, not on a satellite image. Structure, penetrations, shading through a full year, and the route the cable will actually take.
Install
Own crews where possible. The failure modes that matter are all workmanship, and workmanship is not something you can procure by the square metre.
Commission
Prove the string design against the model before anyone signs. A commissioning report nobody reads is a warranty nobody can claim.
Stay
Monitoring that raises a person, not a dashboard. The whole argument is here: somebody has to still be liable when it is boring.
A solar company that installs nothing is a brochure with a balance sheet.
I have read a lot of those brochures. Several of them raised more money than we did.