Or looked at a slightly different way. We should only build things that customers want, and they find it very difficult to explain what they want unless we give them an application to try. When we do give them an application we will receive conflicting requests, and we have to decide which fit the product, or whether we can meld them into a more encompassing or strategic form. Once the customers start giving feedback my thinking is that the "vision" you talk about should just be our interpretation of what the customers ask for rather than being independent of them. Or less absolute, perhaps the proportion of development effort expended on non-customer-requested changes should be small compared to things which address their currently stated requirements. I'll stop rambling now ...