Three Tasks to Perform When Building a Marketing Database

When it is time to build a marketing database, a company's internal IT group almost always lobbies tenaciously for the assignment. For IT professionals, this only seems natural. After all, the word "database" carries all sorts of connotations of nuts-and-bolts computer science.

So, why are internal marketing database builds so often unsuccessful? The reason is that much of what determines success has nothing to do with database technology. Consider just seven of the scores of issues that must be confronted when constructing a database:

  • Address standardization and correction; parsing and matching technologies; and programming logic must be carefully integrated to match (i.e., "group") accounts, leveraging a combination of factors such as name, address, company name, "hard links" such as phone number, and sold-to/bill-to/ship-to relationships.
  • For B-to-C account records, matches must be determined at the individual and household levels. Also, the matches must be unified into a non-circular cross-reference that assigns each account to a single individual, and each individual to a single household. Likewise for B-to-B account records, the same must be done at the individual, site, and super-site ("company") levels.
  • Incremental periodic matches of new accounts to the existing base must be executed, to add activity to existing individuals, households, sites and super-sites, and create new ones.
  • The entire database must be periodically re-matched whenever the number of address changes reaches critical mass, or key USPS tables have been updated.
  • All programming logic must be maintained such that it can be easily adjusted and enhanced, and data re-consolidated, based on the results of ongoing quality assurance of the matches.
  • Logic must be dynamically imposed to make the historical data consistent and usable, and correspond with "real world" behavior. For example, "split-shipment" order records must be aggregated into "true" orders.
  • Rapid classification of customers as of any point-in-time must be done, at the individual, household, site, and super-site levels, employing atomic-level histories up to that point-in-time only. This multiple-level aggregation of data, performed dynamically and at-will, is essential for: 1) accurate response attribution, analysis, scoring, and selection for promotions, and 2) the appropriate allocation of marketing-spend to each customer.