Landing page optimization with audience segmentation
Scott Brinker on Wednesday, October 22, 2008 at 11:33AM Search Engine Land recently ran a post of ours, A Completely Different Kind Of Landing Page Optimization, which talks about segment optimization as a better way of improving the performance of your landing pages over one-size-fits-all content optimization.
Segment optimization is about determining how many different landing pages are optimal for a given campaign, and determining how each should be different from the other. Instead of stretching one page to try to please everyone — which is quite hard to do — segment optimization breaks out several specialized landing pages that each focus on pleasing a particular segment of your audience.
The article walks through an example of a language learning vendor, illustrating ways in which keywords can be used to identify different audience segments and match landing pages to them accordingly — showing why traditional A/B testing or MVT landing page optimization would produce suboptimal results. It then provides a step-by-step approach you can take to start leveraging segment optimization in your search marketing.
Along these lines, you might be interested in our previous post on landing page planning and strategy with the use of a “message match message map” chart.
Landing Pages, SEM, Segmentation in lead generation 
Reader Comments (6)
Scott, great post on Search Engine Land. Although I agree entirely with your idea of segmenting targeted traffic to highly relevant landing pages, qualifying that traffic and directing them to the correct page would be critical. That becomes something of a challenge in paid search campaigns where keywords can easily overlap audiences, especially when you're narrowly segmenting that audience. Perhaps you can address in greater detail the issue of qualifying PPC traffic in another post?
Charles -- thanks for the kind comment! Yes, you're absolutely right that segmentation gets challenging when you can't determine a respondent's segment from their keyword with a high degree of confidence. The argument could be made that **most** search advertising has the risk of overlapping segments.
This is where post-click marketing needs to pick up the ball. As you may have seen from some of the other posts on this blog (such as Why 2 clicks are better than 1), we talk a lot about how marketers can use one or two click behavioral choices on their landing pages to effectively segment respondents after the ad.
I plan on a follow-up article for Search Engine Land that addresses this technique as a "part 2" to the article from yesterday.
Hi Scott, great article on the value of segment optimization. One way to get an idea of which segments are significant to optimize separately is to first run a single test across all of your segments and examine which ones behave differently. This progression is the most common path our customers take to ramp up from testing into segmented optimization and targeting.
Here's a post I wrote recently about the value of bringing testing and targeting together:
http://blogs.omniture.com/2008/10/09/optimization-is-greater-than-the-sum-of-its-parts/
Hi, Lily -- glad to hear you enjoyed the article!
At the highest level, we agree that the best way to learn how to segment your audience -- and how granular your segmentation should be -- is to look at the results from previous tests to find interesting patterns and create new hypotheses.
However, I disagree that running a single, generic test across all of your respondents is the best way to reveal segmentation insight.
The problem with a single test is that by design it's trying to please everyone. Very few marketers are willing to run tests that they know will lose customers, where 1/5th of the audience may love it, but the other 4/5ths are completely turned away from the brand. So in a one-size-fits-all test, almost all of the variations being tested revolve around the common denominators of the entire audience. For businesses that have diverse customers, that can be a pretty low common denominator. It's really hard to learn anything meaningful about segments that way.
We recommend a different approach: start with a hypothesis about a segment. If that's a segment that you can identify by their search terms, that's great (at least as a starting point). Build a separate landing page specifically for that segment, direct those explicit searchers there, and see how that compares against your generic control.
The advantage of this approach is that the marketer can craft that segment-specific landing page ideally for that niche, with full passion and authenticity, and with minimal risk of negative side effects with other segments. You can really learn about what works with that segment versus your generic messaging and how much additional value can be gained -- even from just a single test. If it's a promising outcome, you can then expand your testing with that segment even further. If it's not, move on and try another segment.
Not to toot our horn too much, but this segment-centric strategy is what gives our customers 100%, 200%, 300% lift -- whereas content tests inside a generic one-size-fits-all page are lucky to achieve a fraction of that. It's not the technology, it's the marketing strategy.
Sometimes with software tools, it's easy to fall into a trap of letting the tool dictate the strategy. It needs to be the other way around.
Hi Scott, I don't think I said that a generic test was the *best* way to discover all significant segments, but rather that it can be appropriate for those that are less clearly defined as in the case of Charles's comment.
It is an approach that balances the potential of a hypothesis against the level of effort required to create multiple landing pages.
Also, running a generic test is only the first step toward segment optimization. I think we are in agreement that once you discover those important segments in your audience, you then need to continue to test within them, and that's where you will find significant opportunities for ROI.
Lastly, I agree that a testing platform is just an enabling technology for marketers, and that the ideas and strategy are what will ultimately dictate success. However, it's also important to be realistic about what's possible technically and politically within a company's organization, and consider those factors when designing an optimization strategy.
Hi, Lily.
I'm afraid I still disagree that a generic landing page is a good first step towards segment optimization. The marketer has a major disincentive to try segment-specific hypotheses in engaging with a general audience, and therefore gets little segment-specific insight back. Actually, I'd argue that looking for segment insight in the results of a generic landing page test is just as likely to lead you astray. It goes to the basis of the scientific method: hypothesis, experiment, conclusion. If you're running generic hypotheses and experiments, jumping to segment-specific conclusions is a risky move.
In the case that Charles brought up -- generic keywords that don't unequivocally reveal the respondent's segment pre-click -- the segmentation trigger needs to move into the post-click funnel. In many cases, that can be accomplished by putting a segmentation choice on the landing page and then branching accordingly. But that also requires intentional segmentation content and testing by the marketer.
However, I absolutely agree with you that the level of effort required to create multiple landing pages is often the real culprit, creating unfortunate trade-offs between what is strategically "best" and "what's possible technically and politically". If technical issues are preventing a company from executing the ideas and strategy that will ultimately dictate success, there's a big problem.
For that reason, we focus on solving that exact technical challenge -- how to enable marketers to quickly create, deploy, test, and manage a large portfolio of Long Tail landing pages -- to shift the fulcrum of what's realistic. It's a different underlying thesis than the origin of Offermatica/Test & Target, Optimost, and the MVT crowd, which focused on multivariate testing as a specific experimental methodology instead of holistically addressing the creative production and management of post-click experiences.
When a marketer can copy an existing page -- or work with pre-fab designs, encapsulating their brand standards, to rapidly produce a new page -- and easily plug in content with that specific segment in mind, in a matter of minutes, without having to get IT in the loop, the technical barriers fall away. When front-line marketers acquire the same control over their landing pages that they have over, say, their AdWords campaigns, and can strategically and creatively tie the two together, that's when the magic starts to happen.
Politics, while always a challenge in any form of innovation, are easier to navigate when the marketer can take action quickly and cost-effectively, with minimal dependencies on other people's time and budgets, and deliver measurable results. Pilot programs are great for this, low risk for high return. The greater risk is that if you don't increase the fundamental agility of your post-click marketing, your competitors will eat your lunch. According to the latest research by Compete, competitive differences in post-click marketing can be as high as 500% or more in overall ROI.
Sorry, but if it is "unrealistic" in an organization for a marketer to effortlessly try a segment-specific landing page, they're facing a bigger problem than segmentation discovery. Landing pages can't be mired in the overhead of traditional web site change orders. The cycle speed of competitive online marketing is too fast for that.