If your NLP is that good and you could figure out how to sell it for more money, you could potentially take on companies like Meltwater News.
Neither of the problems mentioned are fatal for an early company, you can get better at customer acquisition and raise prices (or better introduce some new tiers and grandfather your existing clients in).
"Despite the fact that the team thought our product was "clearly" better than Google Alerts, they thought our product had too high of customer acquisition costs and too low of a price point to leave us with big enough margins."
I'm always thinking of new ways to apply what we built over the years. I certainly believe the technology has other applications but Fready is the first one we've landed on.
I actually have an NLP server that converts any data sent to it to JSON format with names/companies/locations, but no business behind it. I'd be curious to hear a unique idea on applying this data if you have one.
However, OpenCalais is too low level. Which is fine, but you'd want to have a broader spectrum of services, including some applications that showcase the power of the platform. See more here: http://www.directededge.com/products.html
Also, language bind is a must. Having a REST API is fine, but having libraries that make it easy to access on any language is much better.
In the end it is all about execution. Make it really simple to use, show how to use, and let the users show you where the business model should be after a while they are using (Airbnb's presentation on the last startup school really shows how this can happen).
Do you think being based in Wisconsin hurt your chances of success?
I recently moved from Madison to San Francisco, and the difference in the startup scene is tremendous.
(I think they did this by converting email -> domain -> domain to ticker symbol -> ticker symbol to news feed rather than any sophisticated NLP thought)
Personally I'd be more worried about legal issues than profit margin for a service like this.
When you're using the content of someone who doesn't want to be indexed, there's no real legal defence for creating a derivative work.
(IANAL but I took a couple of courses in IP law at uni)