September 28, 2010 ↘︎

Segmentation is the key to success.

It often strikes me as strange that people still look at numbers in the aggregate.  Knowing that you get a certain amount of page views, or a certain amount of visitors and so forth, doesn’t really tell you anything.

In order to get some insights of value, things that you can really act on, you need to segment your traffic and conversions.

But so few people really do it, and even fewer do it really well.  And doing it really well can help you in so many different ways, from optimising your site, to optimising your marketing spend.

People are different; they do things differently.  They interact with your site differently.  Broadly speaking, online, people fall into segments, and segmentation can help you to understand those differences and give you a way to identify those people in the future, so that you can hopefully optimise the experience for them – whether it be through more relevant content, landing page optimisation, or keyword reinforcement etc.

The segmentation challenge

Part of the challenge in segmenting is that you don’t really know what you’re looking for, until you’ve found it.  And “it” is something that’s interesting.  Interesting enough that you can do something with it.  Interesting enough that it stands out from the crowd.

The problem with report based segmentation is that you create a segment (might be something along the lines of traffic from organic search) and you then have to wade your way through the static reports to figure out if there’s anything interesting in them.  In most cases, you will fail to find anything interesting, or usable, because your segment (organic search) is still too broad.

Segmentation is an ad-hoc activity

In reality, finding interesting segments is really an ad-hoc thing.  Interesting things tend to be found through trial and error; and report-based segmentation doesn’t make it easy at all, as you typically can’t dive into data and look at things from different angles.

One of the key tools in the Adobe suite of products that we, at Murdoch Uni use, is Omniture Discover (or more correctly Adobe Discover – powered by Omniture).  From it’s name, you should get that you can “discover” things, through ad-hoc segmentation.  Ad-hoc being segmentation on-the-fly.  Don’t see anything interesting?  Try something else.  Then, when you’ve found something interesting, you can save the segment that meets that same criteria and apply it to your reports in SiteCatalyst, so that you can monitor it over time.

Basic Rules

There’s some basic rules of segmentation though:

  1. Segments need to be measurable and identifiable
  2. Segments need to be accessible and actionable
  3. Segments need to be large enough to be profitable

If your interesting segments don’t meet all three criteria, there’s a good chance that they’re not “interesting” segments.  In other words, a good chance you can’t really do anything with/to them.

Using Discover, we can take a look at traffic to the courses section of our site and do some ad-hoc segmentation to see if anything interesting comes out.  Taking a look at traffic in the aggregate, we see the following:


Which doesn’t really tell us much…  We got 1.67 million page views in a certain time period.

So let’s segment that by Visitor Loyalty – looking at First Time Visitors, versus Repeat Visitors, to see if there’s anything there.  Using Discover, you just drag over the segment First Time Visitor and Loyal Visitor onto the workspace and you instantly get the same metrics for those two additional “pre-configured” segments.

segment2 But that still doesn’t really tell us much, other than the obvious…first time visitors look at more pages than repeat visitors, probably because repeat visitors know where the content is they are looking for and can get to them quicker than browsing around.

Now lets segment that by traffic that comes to us from Google Organic search.  We just drill down on the section Courses, and look at our campaign traffic from Google Organic Search, which then segments everything.


Still doesn’t really tell us too much – they appear to be just a subset of the overall segment, behaving in a similar manner.  So we still can’t do too much with that information.

So we’ll segment by time spent, to see if there’s anything interesting in that.  Again, we just drill down on the Google segment showing Time Spent per Visit.


Ah, now that’s interesting…

There’s a high incidence of First Time Visitors from Google who spend between 1-5 minutes on our course pages, whereas, Repeat Visitors tend to spend 30-60 minutes (excluding those that spend 10-30 minutes).

Ok, so what…

Well, we can do something with that.

For instance, we can try to engage 1st Time Visitors from Google more…try to get fewer visitors to spend 1-5 minutes, maybe through cross-promoting content,  maybe through different calls to action.

We could use Test and Target to behaviourally target content at them.  Or we could test different page layouts, or different content altogether, or we could test different promo-modules.

There’s really no end to what we might want to try to do, armed with this information and the capability (and desire) to do something about it.

Segmenting your channel spend

Another way you should be using segmentation is to look at channel effectiveness.  Using ad-hoc segmentation you can see, for example, how many people convert from different segments.

Let’s assume that you have a Display Campaign running.  You’re running display ads across many network sites, which are all driving clicks to your site.

You’ll of course know you’re spend (if not, you should do!).

Let’s assume that you are looking for a conversion of some type – it might be a lead, or a sale.  Doesn’t really matter in this example.

And so you’ll ultimately be able to calculate cost per lead, or cost per sale.

However, your cost per lead or cost per sale will vary substantially by different segments.  Three key things that make up that variance are:

  1. Sub-segment rates
  2. Sub-segment bounce rates
  3. Sub-segment conversion rates

Let’s assume that in this example our sub-segments are 1st Time vs. Loyal Visitors and that we get a 60/40 split.

From a bounce standpoint, your bounce rates might be 40% vs. 20% (by sub-segment)

Now factor in that only some of those remaining visitors will decide to start to convert, but they won’t all convert.  Let’s assume that 30% of 1st Time Visitors start to convert, and 40% of loyal visitors start to convert (start to buy something, or start to sign up for something).

Then you factor in your conversion rates.  Lets assume that only 35% of 1st time visitors actually convert, versus 40% for loyal visitors.

Those segments and rates are purely an example, but are feasible given the audience segment.

The following depicts what the above calculations translate into, if you start with a million impressions, on a CPM based display campaign.


Notice the Cost Per Conversion by Segment…$66.37 versus $32.57.

That’s twice as much to get a new customer to convert than a loyal customer.

So, what can you do…?

How about trying to reduce the bounce rate for first time visitors.  If you can drop it to 20%, you reduce your acquisition cost to $49.67 (from $66.37).  And then you can try to increase your conversion start and completion rates for that specific sub-segment, which will all help towards reducing the cost of acquisition.

Discover Discover

Segmentation in Discover is incredibly easy.  But it doesn’t do only that – it allows you to visualize your traffic and conversions as well.  If you’re just using SiteCatalyst, you really should consider getting Discover – it will, I guarantee, blow your socks off!  It will give you access to view your data in ways that you have only imagined (and wished for).  It will save you time, and probably money in the long run.  It will help you vertically through campaigns and horizontally across campaigns.  (The graphic header to this blog is actually a visualisation of traffic across our website).

Segmentation is the key to success

Segmentation will help you out in so many different ways – and there’s virtually an unlimited number of ways to segment and get valuable insights on your user behaviour.

But the really important thing though, is what you do with it all.

Once you’ve got the insights, don’t be afraid to try to improve things.  It’s measurable, actionable, and profitable!

DB logo
DB logo
DB logo