Greetings from Seoul! Perhaps it’s a bit ironic, particularly for someone who rarely travels overseas, that I find myself writing about public sector sites while abroad. There is nothing like a 12 hour flight to produce a flow of words, however, and I’ve just finished off a post for Clicktale’s May Madness. So as promised, my topic today is the application of Semphonic’s Two-Tiered Digital Segmentation to the Public Sector.
Why Public Sector? First, because Semphonic’s Public Sector practice has grown dramatically in the last few years. It’s become so important to us that we actually went through the process of GSA certification (about as painful from a corporate perspective as giving birth). I also think it’s a fascinating practice area.
Public Sector sites face challenges different and in some ways more severe than those (already challenging enough) endured by the private sector. Government sites have struggled with the basics of measurement. Fettered by extraordinarily restrictive visitor tracking rules (thankfully somewhat relaxed), the basic mechanics of segmentation have been largely missing from the public sector. That, however, is really just a small part of a larger problem. Public Sector sites have struggled mightily with the most basic problem in Web analytics; deciding what is success on their Website.
There are no transactions on Public Sector sites; there are no leads; there are no ad impressions on every page view. In such a situation, what is meant by success?
Most Public Sector sites have chosen to answer this question outside of behavioral analytics. They’ve come to rely on Voice of Customer measures of satisfaction and basic measures of Reach as captured in Visits to the Website.
It’s far from a terrible solution. Satisfaction measures embodied in opinion research provide an excellent measure of overall site success, a means of comparing sites and even a method of isolating success by task. It’s no wonder that Foresee has become the de facto standard for measuring Public Sites and that the adoption of VoC in public sector is at least on par with the private sector even while Web analytics practice has severely lagged.
There are, however, drawbacks to this extreme reliance on Opinion Research. Pure VoC isn’t very fine-grained. As a way of identifying site problems and opportunities it’s rather poor. As a means of site targeting it’s useless. It’s not that VoC isn’t a powerful and appropriate tool, it’s just that giving up on behavioral analysis on the Web is like giving up wine when you’re visiting Napa; wrong place, wrong time.
I believe that our Two-Tiered Segmentation provides a behavioral approach to measuring success on Public Sector sites.
The idea behind Two-Tiered Segmentation is simple: proper segmentation in the Digital world requires two-dimensions. The first is the classic dimension of Visitor Type, the second is Visit Type – what the Visitor is trying to accomplish. Almost every aspect of Web analytics can be framed within this simple Two-Tiered Segmentation of “Who” and “Why”.
In the past few posts, I’ve showed how this scheme can be used in Financial Services, Hospitality, and Media. It’s our belief that almost every type of Web site can and should be analyzed within this type of framework – including Public Sector sites.
The search for good measures of success on Public Sector sites has always been defeated by the very problem that Two-Tiered Segmentation is designed to address. Brute-force site-wide measures like “views per visitors” or “average visit time” can’t be applied consistently across different use-cases. Increasing views per visitor when deploying engaging content is great, increasing views per visitor for getting to a form is not so ideal. For a Website with both functions, all you get when you track the site-wide metric is noise.
A Two-Tiered Segmentation solves the problem. It parses up every visit to a Website into a specific type of Visit by a specific audience. In doing so, it creates a context for metrics that is both powerful and comparable.
Probably no vertical is more diverse than the Public Sector. I cannot hope to create a Two-Tiered Segmentation that is completely representative. Instead, I’m just going to assume a Public Service website for example purposes. The Visitor Type Segmentation might look something like this:
In this case, I’ve assumed the site has a particular target audience in addition to some common additional communities (Professionals, Educators, and Students). If a Website is designed to reach a particular target audience, it’s essential to try and identify whether you’re visits fall into that group or not and to make sure your Reporting and Analysis reflect that. Being successful with the wrong audience is no success.
Unlike my previous segmentations, a Public Sector site will not typically have a “value” dimension to sub-segment the audience. Some of the Visitor Type Segmentations I suggested for media will sometimes apply (Local/National/International, Access Channel, and Site Relationship). Site Relationship, in particular, is often a good Public Sector Visitor Segmentation:
Public Sector sites have often failed to incorporate Visitor Segmentation into their approach just as Private Sector sites have usually missed the boat on Visit Type. It’s a mistake either way. Using only Visit Type simply isn’t enough to achieve clarity in your metrics and KPIs.
Suppose, for example, that I wanted to measure Engagement on the site. No matter what measure I create or how complicated I make it, the standard of success for a Health Professional is going to be different than for a Student. The two types of visitors arrive at the site with different needs, different interests, and different content that will satisfy them. Capturing success in a single Engagement Metric will necessarily confuse the issue – no MATTER WHAT MEASURE OF ENGAGEMENT YOU USE. There is no one metric of Engagement that can possibly measure engagement accurately across two different populations.
What’s more, if I don’t segment, my view of site success will become dependent on changes in my customer mix. If Students register as more engaged than Health Professionals, then an increase in Student visitors relative to Health Professionals will indicate improved site-wide Engagement when, in fact, no such improvement exists.
Engagement, like any other KPI, simply cannot be applied site-wide without creating unacceptable levels of noise.
In common practice, Visit-Type is probably more congenial to Public Sector analysts. The concept of Use-Cases is widely understood and, what is more, ties in well to existing VoC practice.
Here’s a sample Visit Type Segmentation for our hypothetical Public Service Site:
One interesting aspect of this Visit Type segmentation is the split of “Finding a Form” into two Use-Cases. We find that almost all Form sessions fall into two basic patterns. In one pattern, the visitor arrives knowing the form needed and is searching for that form. In the other pattern, the visitor arrives at the site knowing they need a form but unaware of the appropriate Form Name / Identifier. In one sense, both visits have the same success (download a form). However, the level of success and the supporting metrics (time to find / satisfaction) are often quite different.
These are very common Use-Case on Public Sector sites (and not just Public Sector Sites). Remember, when a segmentation scheme would conflate two populations with significantly different performance, you’ll always want to at least consider a further sub-segmentation.
It’s within the matrix of Visitor Type and Visit Type that truly interesting KPIs, site analytics, comparative benchmarking, and testing opportunities all exist. Of these, I’ve talked a great deal about KPIs, testing and analysis. So I want to emphasize the idea of comparative benchmarking. Behavioral benchmarking of Public Sector sites would be a huge advantage to all concerned. With a consistent Two-Tiered Segmentation across multiple sites, that type of benchmarking would be possible. Given specific use-cases (Visit Types) and generalized audience types, it should be possible to create a real benchmark for success across multiple web properties. I think that’s a compelling prospect for Public Sector analytics and one I would love to tackle someday.
This is the fourth Two-Tiered Segmentation matrix I’ve demonstrated. I think that’s enough (maybe more than enough) to convey the general idea. In my next post, I’m going to tackle the art of actually building a segmentation in a Web analytics tool.
