- Search traffic
- Level of competition
- The ease in which a revenue stream can be made from the traffic (for example, having dominance in an insurance niche is more valuable than dominance in an animal niche given equal traffic).
- PPC cost for niche
These are not mutually exclusive.
In terms of automating it you would have to use some sort of statistical distribution for the value contibuted by each of the factors. Those at the top of each factor are worth exponentially more than those only a few steps away. For example in length, a 2 letter .com may be worth 100 times more than a 3 letter, and a 4 letter is worth squat. Another example is Edwin's niche dominance, you can be 1 letter away from the dominant term and be worth 100 times less, and 2 letters to be worth nowt, perhaps petinsurance vs. petsinsurance vs. petsinsurances
It would be a very difficult problem, but you may be able to leverage Google's adwords auctions for a very blunt approximation. It wouldn't work for the 2 letter type anomalies, but for keyword domains how about something along the lines of
[factor 1 x exact searches x ppc cost] + [factor 2 x phrase searches x ppc cost] + [factor 3 x broad searches x ppc cost] + [factor 4 x no of advertisers] + [factor 5 x no. of exact search results] + [factor 6 x no. of broad search results] + [factor 7 x no. of domains containing keywords]
Then have a 1000 or so domains across many niches manually valued, enter all the Google data and try some sort of regression analysis for the various unknown factors?
Just throwing it out there...