There is a fundamental and deeply frustrating divide between the knowledge and skills necessary to support personalization and database marketing and deep knowledge about the digital channel where customers increasingly live.

Traditional database marketing is powerful, proven, actionable, and sophisticated. It turns tiny bits of personal information into levers that drive marketing performance. Unfortunately, it has been focused almost exclusively on channels like direct mail and outbound calling that are, if not quite dead, most assuredly moribund.

Digital Analytics, on the other hand, lives in all the really interesting, vibrant and growing channels – the Web, Mobile and Social. Sadly, it hasn’t developed a comparably rich set of analytic techniques and it’s often delivered disappointing value even to organizations that have invested heavily in developing their Digital Analytics capabilities.

Wouldn’t it be nice if you could take those powerful database marketing techniques: data enrichment, segmentation, predictive modeling, and control groups and apply them to Digital? You can, but as you’d probably expect, it’s not trivial. Digital Channels present unique challenges to traditional database marketers. The data in Digital is fundamentally different and harder to use. Instead of demographics like Age and Income, digital data is mostly anonymous and describes a series of virtual events (usually page views on a Web site). Translating that series of events into meaningful data is extremely difficult – especially for traditional marketing analysts unfamiliar with Digital Analytics.

Bringing sophisticated marketing analytics to drive personalization in Digital is a three step process:

Each step takes some significant re-thinking in making the leap from traditional to Digital Analytics.

Any effort at Personalization should begin with Segmentation. Segmentation is about understanding the key fault-lines in the data you have that define groups of customers with important, shared concerns or needs. Digital Analytics has deployed only the most segmentation (things like Customer/Prospect or New vs. Repeat). That just isn’t enough to drive personalization or interesting visitor-level analysis.

The appropriate technique for Digital channels is a two-tiered segmentation that includes both Visitor and Visit Type. The Visitor-Type is the “Who” and the Visit-Type is the “What”. Here’s a sample of a Two-Tiered Segmentation for an Online Brokerage site:


The unique aspect of the two-tiered segmentation is the Visit-Type. It provides the key intelligence about what the user is trying to accomplish and how successful they were. By combining this with a traditional Visitor-Type segmentation, you create a matrix that is the foundation for good Digital Website Reporting, Personalization, and even for modeling data in your customer warehouse.

Each element of the matrix deserves a unique testing/personalization strategy.

The real trick is extracting those Visit Types from online behavior. They don’t show up as raw data elements in your Web analytics solution. Instead, they have to be inferred. The best technique is to build these visit types in a hierarchical segmentation that uses behavioral signatures based on the type of content viewed, first click, method and detail of search, and navigational methods.

In particular, it’s vitally important to classify your online content as richly as possible so that you can infer meaning from behavior. These “meta-data” classifications make it much easier for the analyst to use web data. A wide-range of meta-data classifications are desirable including common ones like these:

  1. Functional Purpose
  2. Site Hierarchy
  3. Product Family
  4. Topic Classification
  5. Intended Audience
  6. Intended Sales-Stage
  7. Page Components
  8. Component Classification
  9. Amount of Content
  10. Page Length
  11. Content Source
  12. Publish Date & Days since Changed

Remember, this is meta-data ABOUT your content, not the visitor, but it’s used to make the connection between what a visitor views (the page data) and who the visitor is.

These meta-data classifications are the key bridge to creating useful segmentations from Digital data.

Once you’ve created a segmentation scheme and figured out how to classify visitors and visits, the last step is building an offer/creative testing strategy for each group. If you’re doing the process right, this step should be data-driven as well. To make that happen, you’ll need to integrate your VoC online survey data into your behavioral data. With that integration, you can create a demographic and attitudinal profile of every group you’ve identified. It’s that profile that lets you understand the key drivers of choice for each target group and create an appropriate testing strategy. If you don’t do this, your creative folks are just flying blind.

None of the three steps discussed here – two-tiered segmentation, classification based on meta-data, and integration of survey data – are beyond the current generation of Web analytics tools. On the other hand, none are directly supported, commonly practiced or widely implemented and all will take some real work. The complex, hierarchical segmentation capability necessary to support a full Two-Tiered segmentation is particularly demanding both of tool and analyst.

It’s worth it.

These three steps are a bridge between online data and traditional database marketing techniques and create the necessary framework for personalization programs, A/B and multivariate testing programs and even good management reporting.

They are the key to doing Digital Analytics well.