| ||||||||||||
PRINT
COMMENT
SUBSCRIBE
RSSUnfortunately, personalization gets increasingly confusing when one considers all the capabilities and technologies available to help in the process. One expert tells marketers that personalization means incorporating a product recommendation engine; another calls for segmenting visitors into different like-minded groups. Marketers should avoid getting lost in definitions and recommendations, understand that many approaches to site-side content personalization exist, consider each tactic as just one spot on a personalization continuum, and progress through the continuum to continue boosting conversions and revenue.
On the "one to all" end of the continuum, marketers offer a single site with no variations. Each visitor receives the same content and promotions. On the opposite "one to one" end of the spectrum, marketers serve unique content and offers to each individual. While most marketers cannot realistically achieve this one-to-one ideal today, many seek to get as close as possible, knowing that conversion rates and revenue improve with each step taken along the continuum.
Three fundamental tiers of segmentation can help marketers get more personal:
1. Academic segmentation
2. Pre-meditated segmentation
3. Real time, automated persona segmentation
Academic segmentation, the simplest form, consists of random A/B or multivariate testing. Academic segmentation randomly serves visitors varying messages, button colors, page layouts or variances of any other element a marketer desires to test and tracks outcomes to determine the most effective elements. Testing can be applied to any part of a Web site, from the home page to a shopping cart or registration form and virtually any page in between. Academic segmentation helps marketers answer a specific question or determine an ideal design element or workflow. It helps them optimize their site to increase performance but only on a one-to-all basis, applying findings to improve the single experience they offer to all visitors. Marketers can get more personal by applying A/B and multivariate testing in tandem with pre-meditated segmentation.
Sometimes called segmented testing and targeting, pre-meditated segmentation gets much more personal by focusing on groups pre-defined by the marketer under any logical classification system. The types of pages users view, the time of day they visit the site or any host of other factors can define these segments. More advanced than a one-to-all approach, pre-meditated segmentation enables marketers to develop and serve messages and offers specifically for different groups, but marketers can take content personalization efforts even further.
The ultimate nirvana of content personalization, true one-to-one communication, is only achieved through real time, automated persona segmentation. Already generating impressive results for marketers large and small, automated persona segmentation uses next generation algorithm-based applications. The algorithms tap a vast array of available user data (IP addresses, time of visit, browser type, etc.) to make real time decisions, learn from those decisions and their associated outcomes, and get smarter with every piece of content it serves to visitors. Marketers already using pre-meditated segmentation or those considering their first personalization program can use these technologies to offer highly personalized content and promotions to visitors and improve conversions and revenues without spending much time or effort behind the scenes.
Where to focus?
Content personalization efforts can flourish on home pages, category and product pages, shopping carts, forms, subscription and registration pages, and just about any other area of a Web site. They can optimize site design and structure, serve up targeted promotions, and better engage visitors. Today's most sophisticated personalization applications help marketers segment customers across the entire personalization continuum.
It's never been easier for marketers to effectively personalize site-side content for visitors, but they should remain wary of the many shades of personalization and focus instead on the continuum. Ultimately the same goal remains; achieve an optimized customer experience that results in improved conversions and increased revenues.
2 people recommend this article.
RSS FEEDS
I think most people would consider "academic segmentation", when it's applied in a "one to all" format, to simply be testing. It's not really segmenting people, but rather determining which of several variations is the winner -- e.g., a champion v. challenger methodology with A/B testing -- keeping the given audience group constant. I'd caution against labeling this segmentation of any kind, as I think it distracts from the real meaning of segmentation, which your other two categories address.
Semantics aside...
In the category of pre-meditated segmentation, I'd suggest that one of the most successful techniques now being used is one-click (or two-click) behavioral segmentation in landing pages and post-click marketing. Rather than trying to infer segmentation across a large number of web pages -- where the respondent isn't necessarily aware that they're being segmented, and therefore have no option to correct segmentation assumptions made on their behalf -- it's often easier to come right out and ask respondents to make an explicit segmentation choice.
A classic example of one-click behavioral segmentation is a software company that sells data storage solutions asking respondents on the landing page: are you interested in solutions (a) for a small-medium business or (b) for a large-scale enterprise. Not a form, but a couple of one-click images -- essentially a kind of contextual navigation.
In that vein, I'd humbly suggest that such explicit segmentation may be more of a "nirvana" than the examples you give of automated persona segmentation. My concern with algorithms of the kind you mention -- emphasizing things like IP address, time of visit, browser type, etc. -- is that the actual correlation to segments that matter to respondents seems questionable. As with inferring segmentation from site-wide activity, there are lots of ways that respondents can be accidentally assigned to the wrong segment -- without them even knowing about it.
I'm not saying there isn't any value to that sort of machine learning approach. However, I do think that the combination of pre-meditated segmentation -- especially such segmentation that is achieved through the knowing participation of the customer -- with some additional machine learning enhancements can get to you further down the road of segmentation-driven improvement.
Great topic -- best of luck!