Like others have mentioned, you've got to focus on the energy savings to justify the cost.
For some reason Nest has hidden this away in their whitepaper. But - if you take the time to read that - you find, nested (ha ha - get it) towards the end, the conclusions of their simulations.
As a quick aside, let me just say, that for a company essentially selling excellent product design - they did a shit job at presenting quantifiable, easy to understand energy savings data.
Anyway, they devided their simulated case studies into two main groups, people who are gone from the house 8-hours a day 5-days a week (e.g. - your average working professional), and people who aren't (e.g. people with kids at home, or who work from home).
They found that the people gone from the house 9-5 will save an average of $227/yr, and people who need to keep their homes comfortable during the day will save an average of $116/yr.
These savings of course go up if you take a long vacation (they estimate 4 weeks of vacation will add an additional $58-$68 of savings), or if you follow the suggestions of the thermostat to lower or raise your comfort-temp by 1° (an additional $54-$60).
At my work, we do quite a bit of marketing for energy-efficiency companies. The most effective thing they can do is to go into a business or home and say, "We're going to install this energy-saving equipment. We'll pay for the install, and you can pay us back out of the energy savings. After the installation is paid off (usually 1-3 years with the companies we do work for) the savings go straight into your pocket."
Now - this might not be possible with this specific device, but they could at least make the argument that the device will pay for itself in ~3 years. I may or may not want to spend $375 for a fancy thermostat, but that purchase barrier would be much easier to overcome if I could tell myself that it'd pay for itself in a few years.
I'm not sure why this isn't front and center on Nest's site. There might be some liability issues since they've only run simulations, and don't have actual user data. But it seems like a miss on their part.