You’ve successfully completed two of the five steps to optimize your data: Data Aggregation (getting the data required to address your information needs together in one place) and Data Augmentation (filling in the gaps in your data). The data that are identified and defined in the first two stages, whether internal or maintained by third-party entities, are often disparately sourced – with you acting as a sleuth to locate and acquire.
Next up – stage three: Data Processing, or making your data useful. In this stage, your data are effectively transferred and standardized. Not surprisingly, this stage often involves the most time and effort.
Think of stage three as the “trans” stage – translate, transform, and transfer:
• Data translations: Translating semistructured or unstructured data into measurable terms;
• Data transformations: Standardizing your data so that inputs can be linked to outputs or responses, or data from different sources can be useful in analyses; and
• Data transfers: Transferring data to a central accessible location.
Due to various limitations, we still only use a small percentage of the data available to us. Yet to accurately determine the extent to which our communications influence customer responses, we must know more about the practical relationships between contacts and responses. This requires that we quantify – translate – the contact data, defined by our marketing tactics, to facilitate the analysis and interpretation of its influence on responses.
Most marketing communications consist of unstructured or semistructured data including variable content designed for multiple purposes in managing the customer relationship. Unstructured data typically take the form of text, but also includes graphics, images, XML, Flash animation, and so on. The underlying structure of marketing communications adheres to an information-rich standard that can be leveraged for a broad array of analytical purposes – but only if the data can be made useful.
But this unstructured data essentially exist outside of relational databases, and is therefore not neatly organized into rows and columns that can, in turn, be easily accessed and utilized. The data must be captured and translated to be rendered useful in analysis. Yet when successfully performed, it will help you understand how your marketing programs and tactics are influencing consumer behavior.
Typically many forgo these efforts due to the inordinate amount of manual data processing required. Is the time-cost really worth it? There are tools on the market that process and classify unstructured data. Most are not customized for a specific application, are too broad-based, or lack context. Yet in the right hands, these tools and existing technologies can be applied to new areas and answer specific questions, resulting in timely and actionable customer intelligence. The solution requires that we go beyond simple rules-based Bayesian approaches to creating context-specific concepts and relationships. This approach can result in enhanced classifications or taxonomies to provide a rich context for analyzing consumer behavior.
Data tranformation is another sub-step step in which your data are standardized to a common form and definition. As most required data are maintained by multiple entities, these data tend to have different logical forms and formats.
Data must be gathered – transferred – into one location to be analyzed and executed upon. Data nirvana is having all information in one data flow – for processing, analysis and execution! But, alas, many marketers are typically far from achieving this state.
Next month, we’ll cover data analysis, which is about making your data meaningful. We’ll also take a look at Exploratory data analysis (EDA), in which data is validated and described to ensure that all the necessary data are in the correct and valid form for analysis.
Katie Cole, Ph.D., vice president of research and analytics for Denver-based Merkle|Quris, the e-mail marketing agency of Merkle.