Shade analysis that survives the site visit

Desktop shade models and what the crew finds on the roof disagree more often than anyone admits. Here is where the gap comes from, and the three checks that close most of it.

ARAditya RaoHead of Deployment Research

4 August 20262 min read

Shade analysis that survives the site visit

The gap between the model and the roof

A rooftop shade study built from satellite imagery is a good first pass and a poor commitment. Imagery is months old, parapets are flattened by the nadir view, and the water tank that went up last monsoon simply is not there. We compared 380 desktop studies against the as-built string layouts that followed them.

Sample video for checking the embed. Starts at 0:45.

380

rooftop studies compared with as-built

11%

median gap in modelled annual yield

1 in 6

sites needed a string redesign on site

Where the error concentrates

Source of the gapShare of affected sites

Obstruction added after the imagery date31%

Parapet height estimated, not measured27%

Neighbouring structure outside the plot boundary19%

Tilt or azimuth assumed from roof outline14%

Everything else9%

None of these are modelling failures. They are input failures — the study was honest about what it was given. The fix is not a better solver, it is a survey that captures the four inputs the solver is most sensitive to.

Sample Short for checking the vertical frame.

Three checks that close most of the gap

1Measure every parapet, do not estimate itOne tape measurement per roof edge. Parapet height drives the first-row setback, and a 200 mm error at the south edge moves the whole array.
2Photograph the horizon from each array cornerFour wide shots, timestamped. This is what catches the neighbour's stairwell that no imagery shows.
3Log obstruction heights against a known referenceA person of known height standing beside the tank turns a photo into a measurement the designer can use without a second visit.

We stopped treating the desktop study as the design. It is the question we take to the roof.
Design lead, 40 MW commercial rooftop portfolio

The short version

Desktop shade studies are inputs, not commitments — the median yield gap is around a tenth.
Most of the error is stale or estimated geometry, not the solver.
Four photos and one tape measure per roof close most of it in a single visit.

Found this useful?All field notes

Aditya Rao

Head of Deployment Research

Aditya leads deployment research at ReDesk, working with EPCs and installers to find where real projects lose time.

The gap between the model and the roof

Where the error concentrates

Three checks that close most of the gap

One email a week, no product spam