For part 2 of this n-part series on marketing measurement techniques, we’re turning our attention to the methods employed by web analytics tools to capture the source of an inbound click and turn that into a report about whether marketing is working for you.
Bye bye, Referrer
Back in simpler, more innocent times (i.e. up until about five years ago), you could get a pretty good idea of where your traffic was coming from by looking at Referrer data in your web analytics tool (for some reason lost in the mists of time, web browsers always (or always used to) report the previous URL they were looking at whenever they request a new URL, the previous URL being known as the “Referrer”). This information can still be interesting to look at, but its quality has degraded terribly over the years, for a variety of reasons, the main ones being:
- Many browsers now block sending Referrer information, considering it an invasion of privacy
- Certain types of server redirect don’t pass on Referrer information
- Many marketing systems redirect traffic through gobbledygook URLs from which you can’t extract any useful information
Hello, landing pages
So Referrer data has really fallen by the wayside and has been largely replaced by a new technique, known as landing pages. The principle behind this method is that you create a unique page on your site for each marketing campaign you’re running – or even for each element of each campaign. The key thing to ensure is that only your marketing directs traffic to those pages – they’re not linked from anywhere else either inside or outside your site. So when you come to analyze your traffic data, you know that if you see page views for those pages, people must have come to your site via the marketing you’re doing.
It’s not as onerous as it sounds to create unique landing pages for each campaign you’re running, because it’s actually only the URL for the page that has to be unique, not the page itself. Almost all web servers are perfectly happy for you to append dummy parameters to the end of a URL (as long as you still have a valid URL from a syntax point of view) and will ignore the parameters they don’t recognize.
For example, the URL for the home page for mirrormirror (my wife’s e-commerce site) is
but it’s perfectly valid to include a dummy ‘src’ parameter, as below:
Clicking on either link above will take you to the home page. But when the data is analyzed, the “src=iansblog” part will identify the clicks that came from this blog.
It’s perfectly permissible to add more than one dummy URL parameter to a landing page to identify more than one attribute of a campaign, as in the following example:
Here, the src, pub, kg and kw parameters identify the source (Search), publisher (Google), keyword group (“Widgets”) and keyword (“blue widgets”) of the particular click in question. Which parameters you choose is up to you, though your web analytics tool may specify that it will only extract parameters with certain names.
If you have free rein over which parameters to add, though, how do you choose? That’s where you need a taxonomy.
Once you get the hang of ‘tagging’ your landing page URLs, you can apply the principle to all the marketing you’re doing – at least, all the marketing where you have control over the landing page URLs (some notable exceptions are organic search and affiliate marketing). If you apply the dummy parameters in a consistent hierarchy structure, or taxonomy, you can then compare the performance of different elements of your overall marketing mix side by side much more easily.
Let’s use an example to illustrate. Say you’re doing paid search marketing, e-mail marketing, and are running some banner ads. You want to create a categorization hierarchy (the taxonomy) that you can use to organize all the elements of these marketing channels. So you might use the following hierarchy:
“Channel” refers to the marketing channel – in this case, Search, E-mail, or Display Ads.
“Campaign” refers to a grouping of marketing activity, such as a collection of keywords on a particular search engine, or a particular e-mail run-out, or a banner campaign.
“Placement” is an online ad industry term that refers, broadly, to the location of the ad. But it can be applied to describe the “location” of any clickable marketing link, such as a link within an e-mail, or a particular search keyword.
So the trick is to pick values for these categories which make sense across the different kinds of marketing you’re doing. In our example, a Search taxonomy might be (the parts in square brackets are just to remind you which bit of the taxonomy is which):
Google general widgets [campaign]
Here you can see that the placement is really the keyword. For E-mail, the structure would look like:
Widget promotion mail 2-27-07 [campaign]
Blue widget picture link [placement]
Finally, the structure would look like this when applied to display ads:
Display Ads [channel]
Spring widget promotion – hobbyist sites [campaign]
youandyourwidget.com homepage 468×60 [placement]
Because the categorization is used consistently across the different types of marketing, you can now compare these channels, campaigns or individual placements against one another in a meaningful way.
Of course, you could add at least one or even two more layers to this hierarchy (four levels in total seems to be a generally useful number), but the more you have, the more onerous your instrumentation task is going to be.
Generating an overarching taxonomy for your marketing can be a little challenging, due to the diverse nature of the different marketing channels, but it is worth it. Some web analytics tools make it a bit easier by allowing you to define the taxonomy within the tool (usually not to more than one or two levels) and then generating the dummy landing page parameters for you (it’s still up to you to put them into your marketing click-through URLs, mind you). But many web analytics tools fight shy of enforcing an overarching taxonomy, introducing channel-specific categories (such as keyword, or ad creative size) at the lower levels. This makes those tools more usable (certainly not to be sniffed at), but it does make multi-channel comparative marketing analysis more difficult.
More to come…
That’s enough for this week. In the next installment, we’ll look at the methods
web analytics tools use to allocate marketing response to conversion, comparing and contrasting in-session conversion allocation with multi-session conversion allocation.