Ratio of Success to Dud projects?

What proportion of your DS projects do you consider successful? How many are actually operationalized and deliver on the promise they originally made? I'm currently running at about 70% prove useful and go on to enjoy a reasonable shelf life. That's \~30% are failures. Is this too high?

Edit: On a 12 month horizon...

Edit: I'm referring to your work product that goes on to become the basis for initiatives, designs, solutions, services, products, decision support. Versus how much of …

datascience projects success