What are some common mistakes leaders make in software R&D?

1. Not getting formal signoff on wants/needs/requirements
2. Underestimating development time
3. Not wire framing the UI
and on and on ....

CTOs don't tend to have a good understanding of what PhDs are trained in. They're trained in research, which is obvious, but they're also trained in how to answer questions. The analogy I use is from the Simpsons: Lisa Simpson holds up a rock and says, "This is a rock that keeps away tigers." And someone else says, "How do you know you?" She replies, "Well, there's no tigers around here." If you don't critically analyze the output of your R&D department, you can end up with that scenario. A PhD is trained to know how to test whether or not that rock keeps away tigers. To do that, you need to take it to a zoo and walk up to some tigers to see if they run away. Then you take a different rock to those tigers and see if they run away. That tells you if it’s just the rock you currently have, or if it’s rocks in general that are effective. Then you need to know if it’s only effective on the tigers at this particular zoo. At that point, you’d have to visit some tigers at a different zoo. 

It's a very common mistake in industry R&D as well as academic R&D — turns out you need a rock that keeps away tigers in the wild, but you've only tested it on tigers in a zoo. A lot of companies don't tend to publicize such mistakes, but you hear about them through the grapevine: that huge development that they did publicize got a lot of attention, but when they put it in the field, it only kept away tigers at the zoo.