Formally, if you want to be quite academic about it, the most appropriate way to examine this problem is to have created an experiment where sometimes respondents saw K products on the screen and sometimes respondents saw K+1 products on the screen. Then, you build utility models that predict those two scenarios so that you can have confidence in the appropriate cannibalization (cross) effects when the +1 product is added to the scenario...also so you can be confident that you've captured the response error (scale factor) due to seeing K products at a time on the screen vs. K+1 products on the screen. That said, most every researcher in practice doesn't have the time, budget, or inclination to do this.
Rather, researchers mostly rely on the benefits of individual-level modeling plus other tricks such as RFC in simulations to hopefully reduce IIA problems to something approaching near nothing.
I generally recommend using individual-level models under HB plus the RFC market simulation model if comparing K products in the simulation to K+1 products in the simulation.