I find citing research helps a lot. Keep on top of the latest reports and case studies and use them to draw analogies as to why your suggestion is the best given the situation.

Also, know your limits. Many times I make a recommendation and suggest that we split test it properly quickly. The "to test" list will grow and it will serve two purposes:

1. A reminder of important tests you really should do

2. A place where bad/low priority tests go to die. You do this by regularly pruning the list with the help of the client.