A lot of attention is being paid to lead scoring these days. Lead scoring is simply prioritizing and ranking incoming or existing leads according to some criteria. After helping a few dozen companies set up their own lead scoring systems, I've found there are a few key areas where people can falter:
1. Forming a lead scoring system out of concrete.
Really needs to be made of silly putty, something that you can endlessly reshape and refine.
2. Combining declared, inferred and implicit information into one score.
What people say about themselves if often different from what you can infer from them (via data augmentation services, for example), and all are quite different from activity scores (see Steve Wood's article at ...)
And for more accurate declared information, try progressive profiling. What progressive profiling does is let you ask only a few questions at a time. For example, the first time a visitor fills out a form, ask only for the first name, last name, company and email address. The next time, since you know their identity, you can ask more specific information such as title, size of company and BANT information. The urge to gather as much of the information as you can at one time will result in a polluted database over time (a TechTarget reports that users only correctly fill out BANT info about 23.4%% of the time).
Making sure that you keep these three types of information about an individual discrete from one other will help you in your data quality quest.
3. Not getting sales committed to the plan.
If sales doesn't agree that you are profiling and scoring on the right criteria, no system, no matter how slick, will generate the kind of results you are looking for.