Get close to your marketing data. As the old saying goes, you cannot expect what you do not inspect. Assign specific staff to browse your customer data before and after each update, especially when importing large numbers of records. They will catch cosmetic mistakes, but more importantly, begin to see patterns with your customers (buying trends by date, by customer type, etc.) that would not be apparent with standard database queries.
First things first. Before you buy software for your marketing database, you must study the marketing data itself - it likely has old records, duplicates, and other problems created by inconsistent data entry procedures. If you don't set high standards for the quality of your customer data at the outset, it will degrade rapidly over time.
Be careful of the company(s) you keep. Non-profit organizations, associations, and government offices should not be mislabeled as "companies" in your database. Use "organization" or a similar reference.
Use prospect data skeptically when updating your mailing database from prospect lists. Check quality very carefully. Look for internal dupes within the list, blank name or address fields, and other address problems.
Cleanliness is the best policy. To improve your merge/purge processing by at least 10-20%, get your customer file processed with "Address Standardization" to clean up the variations in addressing. For example, 1600 Penn. Avenue Northwest will get standardized by USPS to 1600 PENNSYLVANIA AVE NW. This process will improve the match rate with other prospect files to reduce your duplicate mailings.