I ask people to give me problems instead of a solution.
Example: Can you build a website where I can enter a list of numbers and then create various formulas and way to format the data? Of course I can, but you're going to be paying me a lot of money to build a spreadsheet. Actually, they're not going to pay you to build a full-functioning spreadsheet web app because they think this is a "simple website" just like all the rest and should cost the same. What is the problem? You need a way to enter a bunch of numbers and add them up, get a spreadsheet app.
Rarely will you be given a problem that won't have restrictions (time, budget, current technology, etc.), so why compound the matter by adding more? If you tell me how to build it, you're giving me restrictions.
Once someone starts making assumptions (it should be a web app), they leave out important details about the problem. Without knowing the full-extent of the problem, I'm going to be making design decisions that may contradict what you really want. Even when you tell me how to build it, there won't be enough detail to code it fully. Example: Can you add a report with X, Y, and Z data? Only after hours of creating a beautifully formatted report ready for color printing, they ask you if it can be in Excel without the formatting. Now you tell me.