Tony Fadell

So it wasn't like something in a consumptive form where you could really do user tests with it.

And so we set out a set of tests

like, okay, how fast can I type this text?

How fast can I do this on a hardware keyboard?

And then how can we do this on the virtual one with multi-touch?

And it was a hardware software integration challenge of how we could get this to work.

So we were going back and forth and back and forth.

Oh, that doesn't quite work in the software.

Oh, we need to change this in the hardware.

And so this was a,

over a set of months would okay the hardware keywords here and you know and depends on how pro you know how much you've been using it but if there's this margin error and we could really understand it this over time was started way down here and it started to get it you know and it got a little faster and a little faster and a little faster and how many errors not just how fast but how many errors and how do you correct the errors and all of those things and at the end of the day i was able to convince myself it wasn't going to be a hardware issue

And I was convinced at some point that we were good enough.

Were we as good as a hardware keyboard?

No.

But were we good enough?

Yes.

And then other people came to that conclusion.

But at the same time, there were other people who were adamant that the hardware keyboard has to be there.

And they were unrelenting.

And so it came down to, so this was a classic, like I say in Bill, the data versus opinion-based decisions.