Sometimes, you don't have to tell anyone their idea sucks. You can make them reach that conclusion themselves by asking the right questions - or you'll realise it's not so sucky after all.

"Hmm. How is your idea different from X?" (Instead of "but X does that already, and has a 12 month head start.")

"But what happens when Y?" (Instead of "As soon as Y happens you're screwed.")

"Do you really need W & Z?" (Instead of "You're overreaching. You don't need W and Z in your first iterations.")

Questions let you find out if you're missing the bigger picture, and hopefully let them figure out for themselves what the problems are. If they blow you off they're probably beyond telling anyway.