Profile cover photo
Profile photo
Matt Clarke
Ecommerce director, analytics implementation geek and former editor-in-chief. Currently on the Econsultancy MSc in Internet Retailing.
Ecommerce director, analytics implementation geek and former editor-in-chief. Currently on the Econsultancy MSc in Internet Retailing.

Matt's posts

Post has shared content
Useful new feature to allow the creation of custom acquisition channels in GA >
Google have silently launched a great update to Acquisition Channels. Haven't found any mention of it online yet....

You can now create additional channel groupings and choose which one to use on the channels report.

Better still - they are retrospective and deletable, so you can now group and analyse your traffic in many different ways historically.


- Separate brand terms out into a separate channel, or groups keywords in other ways
- A single Paid traffic channel which combines all advertising together 
- Split different types of social or referral traffic into different channels e.g. forums

There's also a new option to create private channels so you can experiment with your own ideas without it confusing your client.

You can't share private channels. You can copy existing channels but you can't copy public to private or vice versa, making it a bit cumbersome to test modifications to your existing setup.

If you come up with any great uses for this then please post them :)

Post has attachment

Post has attachment
R's BTYD model looks just the thing for my MSc project CLV data analysis ...

Post has attachment
Found an interesting paper while writing up my dissertation (which is on coupon code websites and why they are shite) which examines CLV model performance. Here's my summary and the paper.

Malthouse (2009) has undertaken one of the few empirical studies to compare the performance of the various CLV models in a project in which 24 teams examined the same 15 year data set. Malthouse (2009) found that all CLV models required substantial skill to build and recorded marked difference in their performance, even within the same model family. He concluded that the complications of calculating CLV meant that the success of models was determined more by the modeller than the model family used. Further, the results also call into question whether CLV can be predicted accurately at an individual level at all, something also noted by Blattberg, Malthouse and Neslin (2009), with the best individual-level model off by a factor of 5.4, which implies that firms using CLV to allocate marketing resources will often invest in the wrong customers. 

Post has attachment
Blimey. That's the second useful thing I've stumbled across. Must be my lucky day...

Post has attachment
This has just come in handy. A very thorough guide to doing more complicated stuff (in relative terms) with robots.txt. 

Crikey. It's been months since I posted anything here...

Anyway, here's a quick bit of code you can use to track zero result searches in Google Analytics.

Stick this on your "no search results" page and it will grab the contents of your HTTP GET request search parameter (q in this case) and assign it to a Google Analytics event so you can track it. 

<script language="JavaScript" type="text/javascript">
function getSearchTerm(name) {
  return decodeURIComponent((new RegExp('[?|&]' + name + '=' + '([^&;]+?)(&|#|;|$)').exec(||[,""])[1].replace(/\+/g, '%20'))||null
_gaq.push(['_trackEvent','Zero result searches', ''+ getSearchTerm('q').toLowerCase()  ]);


Post has attachment
Very good post covering responsive design resources and tools. Wish this had been written a few weeks earlier! 

Post has attachment
One of two that my five year old son, George, caught today! 

Post has shared content
Nice post from +Michael Whitaker on my favourite GA feature - event tracking. The error message tracking one is a beauty. 
Wait while more posts are being loaded