We've been advocating respondent segmentation in post-click marketing for a while now, in particular early segmentation, before a respondent converts. The motivation is to fill the gap in the funnel: somewhere between winning that first click from 100% of your respondents and actually converting a tiny fraction of them (~3% industry average), there's a mysterious pre-conversion void, where the far majority of respondents simply vanish without a word. The question is always: why?

Directed behavioral segmentation dispels this void by inducing respondents to click through simple choices on the first page or two of a landing path. To the respondent, these choices are presented as shortcuts, to quickly direct them to the content that's most relevant to their particular interests. (Example: I click on a Caribbean resort ad, and the first page of the landing path asks me to choose romantic rendezvous or bring-the-kids family vacation.) To the marketing manager running the campaign, aside from targeting the right content to the right respondents (which naturally improves the conversion rate), the tracking of these segmentation choices reveals who's responding to which ads -- even if they don't convert. This helps explain why certain people are or are not converting. Then, instead of trying to globally optimize a marketing campaign for all audiences -- which inevitably devolves into generic, middle-of-the-road tactics -- you can maximize for each audience segment independently, vividly, passionately.

Although this technique has proven quite effective, a downside has been that it implied marketing managers had to define a hierarchy of segments and subsegments first and then build paths within that structure. It is certainly a very good idea to think about segmentation at the start of a campaign, but the long tail of web marketing is fluid, and it's often advantageous to let your ideas of segments and subsegments evolve through quick and easy experiments "in the wild". Think about segments, but don't overthink them -- the post-click marketing ecosystem of create, test, analyze, adjust is better served by cycle speed than lengthy deliberation.

Inspired by the tagging paradigm popularized by Web 2.0 sites such as flickr, del.icio.us, and other social media -- where users "tag" content with their own keywords, and collectively everyone's tags are aggregated into an emergent "folksonomy" instead of a rigid, top-down taxonomy -- we've adopted that idea for respondent tagging. It reverses the flow: instead of users tagging content, content tags users (or, more accurately, users' content choices cause them to be tagged).

Rather than laboring too hard over a segmentation hierarchy, a marketing manager simply tags a respondent with a keyword at any point in their post-click marketing experience. Click on a choice on the landing page? Tag it -- it's analogous to assigning a segment. Submit a form with a particular answer to a question? Tag it too, a way to mark respondents with different qualifiers.

It's the same underlying principle we've been championing -- segmenting and qualifying respondents is essential to good post-click marketing -- but tagging is just plain easier than getting caught up in hierarchical segmentation. You can change tags or add new ones in a matter of seconds. The order in which respondents are tagged -- a segment vs. a subsegment -- can be switched around effortlessly, without losing the fidelity of the information. And you can layer your own meta-data on top as you see fit, facilitating better hand-offs to subsequent systems and later stages in your sales and marketing funnel.

The beauty of a tagging approach is that you learn about the emergent structure of your respondent universe, more rapidly, with less allegiance to pre-conceived notions of your audience -- notions that just as often hold you back from discovering the real drivers in your customer base.

Tagging does have its critics, primarily in a social media context. The concern there is that the lack of coordination among participants inventing tags can generate meta noise and idiosyncrasies. However, respondent tagging in the other direction has three countervailing forces to impose just the right amount of order:

1. As a manager, you can enforce as few or as many rules of tagging -- how tags are created and used -- as you want. Different organizations can choose to make this process as tightly coordinated or as loosely distributed as they see fit, whichever best serves the needs of their market. Even within a firm, this might differ from campaign to campaign, some more structured, some more experimental.

2. Unlike social media, where a tag is often merely a classifier, the context of when and how a respondent is assigned a tag during post-click marketing is rich with added meaning. A respondent's walk through a path has direction and order -- first choice on a page, second choice on a page, choices before conversion, choices after conversion, etc. Tags at these different points can have different significance, which can help identify the pivotal tags in a campaign's success.

3. The correlation of tags with conversion rate and cost-per-acquisition (CPA) metrics provides another dimension by which to rank the relative significance of tags, which may vary across different paths, traffic sources, or even overall campaigns.

The great thing about this additional meaning in #2 and #3 above is that the marketing manager doesn't have to consciously think about any of this in advance. Simply tag where you want, and then, post-hoc, you can see which tags emerge as influential. This is the basis for what we call context-valued tagging (CVT), a class of algorithms that enable you to mine rich information from very simple tagging.

We believe this gives you the benefits of a tagging methodology, while exercising enough managerial control -- as much as you decide valuable in the balance of flexibility and consistency -- to run a dynamic, vibrant, and stable platform for post-click marketing analysis.

Test, test, test. And now tag, tag, tag.

-- Scott Brinker