Can you actually predict shoppers' behaviors by having them pretend to shop while sitting in front of their computer?
Apparently you can. Decision Insight from Kansas City claims a 90% correlation of their virtual research results with actual store sales data from IRI.
What have they tested? Shelf sets, packaging, SKU mix, promotions & displays, and even pricing. Grocery Headquarters has a Heinz ketchup case study about shelf sets; reducing the SKU count by 25% grew category sales--good for both Heinz and the retailer. Another example about Kellogg's Smart Start cereal packaging is at AFN.
These artificial, desk-bound, not-in-shopping-mindset virtual studies seem to be highly predictive of real-world results. Virtual store research might not just be good enough: it might be good.