Predictive analytics needs a foundation of clean data. Here are top tips from our most recent lead gen implementation on Salesforce.com. You can use these immediately in any environment:

1. Address standardization and change of address. Typically up to 30% of records lack complete address information. This affects both deliverability and duplicate search. If you have not done so in 3 years, run a Change of Address (NCOA) process to get your customer’s new addresses. Stay up to date on your customers because 10% of businesses move each year. When you receive a street address change, but it’s a PO Box in the same city, retain both — one for mailing and one for secondary validation.
2. Dupes are the silent killer. Because a single comprehensive definition does not fit all, de-dupe the files using a variety of match logic. You can try a loose match logic (a few criteria, gives more duplicates) or tight match logic (more criteria, resulting in fewer duplicates). Take address elements into account, but use transactional information to determine which record to keep or drop. For example, between duplicates you might want to merge a record with multiple contacts into a record with the largest sales amount or the record with the longest time on file. Run dupes within your accounts, as well as against other sources. If you see the same record in two tables, delete or mark clearly why and set a shelf life to expire one.
3. Fix bad data entry practices. With key fields you use the most, particularly text fields, consolidate the misspellings and mixed-cases that make reporting difficult. As it can feel like a huge task, you do not have to do everything at once – fix just five fields this week, more next week. Here is an opportunity to update correction rules and fix legacy errors. As a bonus, segmentation and reporting become much easier.
4. Match to a B2B database or Tier 1 compiled list. While the obvious next step can be to append the firmographic elements for analytics, you can use information from the match rate to help further clean your data. Segment the unmatched customers and evaluate if they: are duplicated somewhere else, have failed address standardization, have any useful fields, or are possibly orphaned from a prior merge exercise. It’s springtime, prune the dead weight!
5. Append firmographic data. While technically it is not cleaning, append and enrichment of your data can provide big dividends from a hygiene perspective. Compare all contact information against the external data and fix format errors, missing extensions, suite number etc., while adding new contact info you did not have.