in Uncategorized

Thinking About Audience Segmentation

While this is not an explicit Explicit Design volley, it’s definitely related.

A holy grail in Web site information architecture is the ability to cleanly segment content by audience type. Much of the content on a website is not applicable to every single person, but without good segmenting methods, we have to expose all that content to everyone.

Some sites have it easy. An Adaptive Path favorite is Hay Net.


However, few folks have such easily separated audiences. This is particularly true in high-tech marketing, an area I’ve been deeply involved in ever since working with PeopleSoft in 2001. Every high-tech marketer wants to target different messages at different levels within the potential customer’s organization — executives get a Business Value message, directors and managers get an Ease of Integration message; developers get technical specifications, etc.

One company actually went pretty hard to market with these distinctions: Siebel. Thanks to the Wayback machine, I was able to dredge up their “custom views” over the last few years.


Well, I went to today, and noticed that they’ve dropped Custom Views. If you ever clicked around their Custom Views, you know that they never really did the concept service — there’d be a single page for each audience, but beyond that, the distinction was lost.

But I don’t think the real problem was execution. I think it was more fundamental. It’s pretty much impossible making good clean distinctions that visitors can select themselves by. The changes in Siebel’s options show this. In June 2002 they add “Small and Medium Business”, in part reflecting the importance of the market, but also because an “executive” of a small business has a lot more in common with a “business manager” of a large corporation.

I find the combination, in April 2003, of “executive” and “business manager” telling. It’s basically them throwing in the towel for this kind of audience segmentation. What’s happening here is that Siebel is realizing that meaningful segmentation doesn’t have as much to do with job titles and self-identification, as it does with the types of tasks the people are engaged in. When it comes to enterprise software, Executives and Business Managers start to blur — you pretty much just want the best value for the money.

And this is why Siebel was wise to ditch the Custom Views altogether. It was an attempt on their part to be Explicit — what could be more explicit than getting people to *exactly* the content best suited to their needs — but it ran against the reality that, when it comes to marketing products, “audience type” is the wrong type of explicit. All that matters to the visitor is their task at hand, not what some company thinks of their job title, and they’ll click immediately to whatever they believe will support that task — which is likely one of the clearly labeled main navigation areas.

All this said, I don’t advocate never segmenting by audiences. But you can only successfully do so when the audience accepts the labels as meaningfully applying to themselves. One realm where this works is higher education. UCLA’s home page clearly distinguishes its audiences.


The reason this works is that the *tasks* — what people want to do at UCLA’s site — cleanly break down by audience types. Future students have interests distinct from current students. Current students distinct from faculty. Etc. Not to say they’re mutually exclusive — but a visitor can look at those selections and click with assurance. You can’t do that with a selection like “business manager.”

Anyway, I feel this really resonates with the notions of Explicit Design. If you can be explicit and meaningful — such as Hay Net, or UCLA — then by all means you should. But if you can’t be explicit, as in segmenting potential customer audience types for high tech, then attempts at being so will backfire, because this faux explicitness confuses the visitor when they don’t see an option that inspires confident selection.


  1. We’ve tried this several times down at the hospital I work at, on various projects. It’s worked to varying degrees of success. On our Intranet it was fairly easy to segment some of the Clinical information and tasks from the Administrative, for example.

    In most cases it doesn’t work as well. I’ve found that overcoming our internal “org chart” structure to be as much as a barrier to audience segmentation as anything else.

    Some times even if you can be explicit and meaningful as far as the user is concerned internal politics and the org chart IA and labeling schemes move in to block your way.

  2. I too work on an intranet (mine is for a global electronics company), and we’ve succumbed to a hybrid “org chart/job function” design due to the same internal politics Keith mentions.

    The governing architecture is a complex set of “microsites” based on the organization chart because that is the most recognizable structure to the people who built it in the first place.

    To address issues of content “findability” (a big issue when I came on board), we started organizing links and content inside the microsites by task or job function. For smaller department sites we use task because it is simpler, but for the larger departments with multiple functions we found it easier for the employee to use if we grouped them by job function.

    I should clarify that these are not portals, but there is segmentation on two levels (org chart and job function) that is proving successful. Even with our limited resources for testing, we can already see better task success and employee satisfaction with the intranet growing.

  3. Funny…I interpreted the segments COMPLETELY backwards: Need Hay to me said, “If you need hay, follow this link so you can obtain hay.” Have Hay says to me, “If you have hay and want to find folks that need it, follow this link.”


  4. Good post. IBM does segmentation on their homepage as well, and it seems to work. So does Dell.

    As a sidenote: I’d give up on the whole “Explicit Design” thing. I don’t think it’s working – not for me at least. “Findability” might work for Peter Morville, but “Explicit Design”? Nah.

  5. Heh. I don’t know if “Explicit Design” as a term is working for me. But there’s still something about the concept that I think that resonates, so I’m going to keep at it for a bit.

    Re: Hay Net. Yeah. After having such delightfully clear segmentation on the home page, it does fall apart once you click on a choice — what’s behind the links isn’t managed well, and so the results are confused.

  6. This is a good post. I can imagine using a simple tool–a matrix of audience types and tasks–with the client to determine if audience segmentation would work for them. If there’s lots of “bleed,” segmentation is probably not a good choice.

    As for Explicit Design, I say keep working it. I had a hard time with Findability at first (not the concept as much as the label), but it’s stuck since.

  7. Peter, I come at segmentation as a marketer. (Hey, guys, hold the rotten tomatoes. Contrary to popular opinion, not all marketers are assholes.)

    Segmentation is nothing but divvying up people you find it worthwhile to treat differently from one another. Segmentation is part of the pre-work for every ad you see on television. It has an impact on the design of the ad — but without the self-categorization of the old Siebel site.

    It seems to me that applying the 80/20 rule to potential customers is the key here.

  8. Great subject…we’ve been working on similar lines on the Microsoft UK site and getting some psotive feedback through research since launch of our new homepage. We’ve a number of initiatives focussed on gearing our content/servcices/offering to segments (audience-centric, rather than product centric) where possible. Lots more to do in this area.

    Feedback most welcome!

  9. The problem with “Hay Net” is that – even with only two distinctions – the distinctions can be very misleading.

    One could reasonably assume if they need hay they should click through to “Need Hay”. Instead that section has listings of others who also need hay; the listing in “Have Hay” are… those who _have_ hay.

    So, if you need hay… click on “Have Hay” to find it. And if you have hay you wish to sell, click on “Need Hay” to find buyers.

    So even in this basic case it’s important to understand whether the designations are for ‘me’ (the visitor) or for ‘them’ (the ads and those that placed them).

  10. One of the concerns of segmentation like UCLA’s is that you have to make sure you identify every segment you care about, and ensure that the organization is on board with ignoring or minimizing the others. For example, I would think that the media would be an important audience for UCLA, but does “visitors” really speak to them?

    I’ve been working on usability for a large health system’s public website, and have arrived at a hybrid segmentation approach: patients, visitors, medical personnel and everyone. Information about general “health & wellness,” for example, applies to all, so it resides in the last “segment.” Based on our user research and testing, its looking like the most effective way of giving users quick access to the most needed/requested information, while providing a workable navigation scheme for a very complex (broad and deep) content structure.

  11. Need to learn what the user does and infer his/her interests from that and then slowly start to feature elements that match that or similar profiles. Also flap arms and fly to moon.

  12. We seem to have had some success segmenting folks on the homepage – not as explicit as some others, but seems to be doing the trick.


  13. I agree with you, Peter, that “All that matters to the visitor is their task at hand, not what some [entity] thinks of [their title], and they’ll click immediately to whatever they believe will support that task.”

    But in my experience as a prospective grad student, an enrolled grad student, and a university staff member (i.e., someone who’s looked at a lot of university web sites in the past 5 years), I’ve rarely found audience segmentation of the sort used by UCLA useful or confidence inspiring. IMHO, this is because the audience lines are just as fuzzy in academia as in industry.

    For example, both future and current students (not to mention staff) legitimately have an interest in curriculum-related content (degree requirements, course descriptions, etc.), logistics (campus maps, parking permits), student life (organizations, activities, etc.) and lots of other info. So even if multiple paths ultimately take everybody to the same content, there’s still no guarantee that they’ll be able to, as you say, “click with assurance.” I can’t tell you how many times I’ve had to pause and think, “OK, what audience do they think I’m in *this* time?” That it’s possible to be in more than one audience at the same time makes this even worse!

    I’m not arguing against segmentation, but I don’t think it’s easily done, and I don’t think UCLA’s attempt–one that’s similar to many university sites–is any better than Siebel’s.


  14. I agree with Michael Fry on this and will go a bit further. I have been frustrated by poor attempts at such segmentation for years, both in trying to serve clients (or bosses) who insist that we try and by sites that try and segment me. I think there are a couple of important things to note.

    First of all, I think the University example is telling – it is easy to assume that in a university context, the segments are going to be quite distinct. But that assumption is meaningless until thoroughly tested with the segments themselves.

    Second, segments work really well when they exist in the product, information, or other subject of the website itself. For cell phone companies, for instance, that actually have different pricing for businesses and consumers. The danger is in assuming too much. Just because business users expense their phones and so can afford more $$ doesn’t mean a “personal” customer doesn’t want your more expensive offering. Making it less accessible to them based on incorrect segmentation can equal shooting yourself in the foot.

    Third, and stemming from #2, in my attempts to DO this kind of segmentation in large sites, I have found that in fact you want to allow EVERYONE to access ALL of the content through each view (and NOT by “pretending” to be in a different group – the key to the segmentation is that it is just one particular entry point. I learned this the hard way – by segmenting stuff OUT of different views then having to put it right back in later on. Views provide emphasis and prioritize certain information over other information, they don’t become the sole container in which any information resides.

  15. As a sidenote: I’d give up on the whole “Explicit Design” thing. I don’t think it’s working – not for me at least. “Findability” might work for Peter Morville, but “Explicit Design”? Nah.

Comments are closed.