Start by defining a smart failure. Everyone in your organization knows what success is. It's the things you put on a resume: increased revenues, decreased costs, delivered a product etc. Far fewer know what a smart failure is — i.e. the type of failures that should be congratulated. These are the thoughtful and well planned projects that for some reason didn't work. Define them so people know the acceptable boundaries within which to fail. If you don't define them, all failure looks risky and it will kill creativity and innovation.
Questions to consider in defining smart failures: What makes a failure smart in our organization? What makes a failure dumb? Specifically, what guidelines, approaches, or processes characterize smart risk taking? What clear examples can we point to, to demonstrate smart failures? You want people to clearly understand the right and wrong way to fail.
Next, reward smart failures in addition to successes. Once you've defined smart failures, you want to reward them just as you do smart successes. It sends a powerful message about what sort of behavior is encouraged in your organization.
Chapter 4 in the Mistake Bank book is called "Smart Mistakes," and adds some perspective to this argument: on defining a smart failure, establishing "affordable losses" and ordering activities to fail more cheaply.
[There's a Kickstarter campaign underway to raise funds to get the book in print.]