Layered Rendering provides a means to limit combinatorial explosion for highly configurable products with many options.
Great, what is combinatorial explosion? When products are highly configurable, containing a number of unique combinations of configurable attributes upwards of several thousand, or even into the millions, it is not practical (or scaleable) to retain (and maintain) an equivalent number of renders.
For example, here we have a sofa.
This sofa is offered in 6 colors, the Accent Pillows are offered in 6 colors, there are 5 leg types, and an optional skirt. In sum, there are 360 unique combinations of the 4 attributes listed. Without layered rendering, it would be necessary to retain 360 unique renders to serve up on-demand, based on a given customer's configuration.
However, Layered Rendering provides a path to limiting the number of required renders. Separating the product on the basis of its configurable, component parts allows for defining said layers. In this case, the model is decomposed into the 4 layers below.
Now the options for each of these layers can be rendered independently, decreasing the number of required renders from 360 to just 23.
Some Assembly Required
Once individual renders exist, the Layers Service retrieves the requested images, per the configuration, and the final render is assembled and produced.
Re-Configure, Retain Performance
In the context of a live 2D configuration experience when the Threekit Player is initialized, the product retains a default configuration. When the shopper changes an attribute to a different option, all elements of the model not affected by that change will remain static. This allows for very performant configuration experiences.