DB Design :: Normalizing Personal Contact Information

I have a large data set with 10s of millions of rows of contact information.  The data is in CSV format and contains 48 columns of information (First name, MI, last name, 4 part address, 10+ demographic points, etc.) and I'm struggling with how I should design the database and normalize this data, or if I should normalize this data.

My 2 thoughts for design were either:

Break the columns into logical categorical tables (i.e. BasicContactInfo, Demographics, Financials, Interests, etc.) Keep the entire row in one table, and pull out the "Objects" into another table (i.e. ContactInformation, States, ZIPCodes, EmployementStatus, EthnicityCodes, etc.)

The data will be immutable for the most part, and when I get new data, I'll just create a new database and replace the old one.

The reason I like option 1 is because it makes importing easier, since I can just insert the appropriate columns from each row into the appropriate tables.  Option number 2 feels like it would be faster to get metrics on the data, like how many contacts live in which states, or what is the total number of unique occupations in the data set.  Plus I'll be able to make relationships between the tables, like which state is tied to which zipcode, which city is tied with which county, etc.  Importing that data might be more tricky, since I don't think SQL Bulk Copy will allow for inserting into normalized tables like that.

The primary use for this data is to allow our sales force to create custom lists of contact information based on a faceted search page.  The sales person would create the filter, and then I will provide them with the resulting data so they can start making business contacts.  Search performance needs to be good.  Insert, update, and deletes won't happen once the data has been imported.

What should I look for in designing this database?  Any good articles on designing tables around wide data sets like my contact information?