Demystifying data quality: the top five questions for marketers
"Know thy customer” is the marketing wisdom that underscores every effective marketing campaign. Relevant insights and accurate information help marketers up-sell and cross-sell as well as ensure that their customized offers reach the right household or individual via mail, email or phone. Savvy marketers who procure third-party data to enrich their customer databases and expand prospecting opportunities want to be confident in the coverage, accuracy and quality of their data investment. However, many marketers are perplexed by the contradictory findings published by data providers and unable to effectively discern data quality and coverage. We asked Andy Johnson, Experian Marketing Services’ Vice President of MIS Data Sourcing, to identify some basic questions marketers can ask to demystify data quality.
1. Who conducted your provider’s data quality audit?
Johnson: Objective auditing is a critical component of evaluating data quality and coverage across a crowded competitive landscape. An independent third-party evaluation versus an in-house analysis creates an additional level of assurance that the results are credible. As a consumer, what is more compelling to you: the manufacturer telling you the product is the best or an independent consumer report rating the product as the best? The more objective and transparent the study methodology, the more confident you can be in the credibility of the results.
2. What are the data application and data attributes for the test specified in the research methodology?
Johnson: The data application and data attributes identified in the methodology discussion will set the stage for how the data will be matched and validated for quality and coverage across all providers. Typically, there are two types of data applications that are audited: enrichment or prospecting data. Enrichment data is most commonly used to append individual or household profiles so the depth and the breadth of attributes available (e.g., lifestyle, demographic, etc.) are benchmarked and compared. Mailing tests validate prospect contact information (e.g., name and home address) and the deliverability of those records. In order to ensure audit fairness, the competitor sample request should specify the data application and data attributes for the sample geography to ensure a fair comparison across competitors included in the audit.
3. How was the data pulled?
Johnson: The source of the data also should be identified in the methodology. So fairness is maintained, each competitor should be blindly solicited, receive the same sample request documentation and be responsible for producing the sample. If the competitive sample is pulled from a reseller or another party, the sample is potentially compromised. The age, accuracy, comprehensiveness and mix of data sources (e.g., known, modeled, inferred, etc.) will vary depending on the provider. A sample procured directly from the data compiler or aggregator lowers the risk that results will be skewed and unreliable.
4. What was the geographic sample for the test?
Johnson: The geographic sample for the test should be random and large enough so that it is statistically representative of the United States. As a potential data purchaser, you want to know the quality and reliability across the entire national database, not for a handful of ZIPTM codes. A statistically significant sample size is projectable and a reliable indicator of the data performance across the nation.
5. How was data matched, and how was the accuracy validated?
Johnson: Ultimately, the audit will provide a scorecard that helps the discriminating marketer assess coverage and quality. As a marketer, you want to know what percentage of records have valid, nonmissing values by provider. The overall match rate should be explained and put into context in order to understand the breadth, depth and accuracy under review. Census or syndicated research data at the geographic level is a reliable benchmark for validating the data’s accuracy. Higher counts or greater attribute coverage is of little value to the marketer if the basic IDENT information is wrong.
Unfortunately, there is no perfect data source, and there will be imperfection in any data file. The key is to understand what factors affect that variance, such as the ratio of known versus inferred data, how households are defined by provider, and how each provider sources and compiles data. Often, these factors are obscured in data quality audits, but in the real world they can have an enormous impact on marketing intelligence and campaign effectiveness.
To learn more about Experian Marketing Services’ marketing data quality, we invite you to read the 2011 ConsumerView℠ marketing data quality report.