Last year a distributed solar client asked for a rooftop power station 3D interactive showcase. The first proposal and the live version turned out to be very different products, and the review left a few lessons worth writing down. The proposal stuffed everything the client wanted into the scene: per module shadow analysis, live yield simulation, weather data feeds. On an ordinary office laptop the frame rate dropped below usable, and the build schedule kept stretching. The live version was rebuilt around one rule: keep what the eye can judge inside the scene, move anything that needs computing next to it.
Where the first proposal stalled: the cost of simulation style interactions
Shadow analysis means calculating occlusion hours module by module. Yield estimation pulls local weather data. Inside browser based 3D, either the data hookup fails or the frame rate collapses into a slide show. One easily overlooked premise matters here: the actual users are sales staff carrying a tablet to present in front of the building owner, not engineers sitting at a workstation. Who watches, and on what device, decides which interactions are even feasible.

Decision one: yield calculation moves out of the scene onto a chart page
The owner cares most about the return on investment, yet that math does not need to run inside a three dimensional scene. It moved to a standalone tool page: enter roof area and local sunlight hours, get an annual yield and savings table. The 3D scene keeps a clear job, showing where the arrays sit, how many panels fit and how far they stand from obstacles. Each side does one thing, pages link to each other, and the development effort dropped by an order of magnitude.
Decision two: tilt and orientation stay in the scene
Roof pitch and panel orientation are things a human eye judges directly, so these interactions stay. Drag a slider, the tilt changes in real time, and the owner immediately sees how orientation shapes the layout. Interactions with visible results are the strongest suit of 3D, and if cuts have to happen, they happen elsewhere.
Decision three: module selection reveals three parameters only
The proposal wanted a dozen parameters behind every module. The live version carries three: model, wattage, warranty years. An owner tapping a module on a tablet wants confirmation of the panel grade, not a datasheet. The full parameter table sits on its own page for anyone who wants to read deeper.
The build schedule compressed from the three plus months quoted in the proposal to seven weeks, and the change the sales team reported was more practical: owners now drag the tilt slider themselves during meetings, and the doubts that used to stall signing shrank noticeably. This pattern is not solar specific. The medical catheter 3D interactive case cut interactions by visitor scenario, and the filling machine case study explains why the live version shipped seven fewer interactions than the proposal. Solar companies also have another 3D route, the configurator, covered in the solar configurator build guide with option logic and quote linkage.
Related Q&A
Why not compute yield simulation directly inside the 3D scene?
The computation is heavy and depends on weather data feeds. In a browser it either wrecks the frame rate or returns unreliable numbers. Splitting it onto a separate tool page keeps both sides simpler, with the scene showing only a result summary.
Do rooftop scenes need real buildings as reference?
No. Three typical roof types, flat industrial, pitched residential and carport, cover the vast majority of demonstration cases, and real client projects get their own models added later.
Does this kind of showcase actually help close deals?
Reading a static rendering and dragging a tilt slider to watch the layout change are two different decision experiences. During the sales conversation the showcase lets owners reach conclusions themselves while the sales person only guides.