testing MA blog

Analytics + survey data = customer insight [1+1=3]

It's very easy to get consumed under the shear volume web analytics data, so all you want to hear from me is, "have you considered measuring this or just try another data source". 

What's more important is getting the balance between data sources and knowing when to use them depending on what it is you want to improve....there's definitely another posting on that topic in the future.


Up until recently I'd always used quantitive data, the only exception being while doing my degree dissertation, (some five years ago now), I conducted the regulatory interviews on my chosen subject: "What are the barriers to growth in the mobile music download business models?", and turned those interviews into something meaningful by doing the following:

Having collected the primary data; email responses received and recorded telephone interview transcribed, a process of ‘abstracting and comparing’ was used to analyse the data. The responses to the questions were studied in detail one at a time, highlighting the most relevant parts of the data and making notes of any additional thoughts during this process. This process was akin to that of coding and memoing, although the exact definition of these terms is themselves debatable:

On the one hand coding is analysis. On the other hand, coding is the specific and concrete activity which starts the analysis. Both are correct, in the sense that coding both begins the analysis, and also goes on at different levels throughout the analysis (Punch, 1998, p.204).

Therefore the term, ‘abstracting and comparing’ was chosen to best describe the method of data analysis undertaken, although within the data analysis process, some coding was carried out in order to turn raw data into higher level concepts. The abstracted data was listed in a table, compared, grouped into like indicators and then the higher order concepts noted; such that an overriding result to each question could be determined. The process of comparing was ‘essential to identifying and categorising concepts’ (Strauss and Corbin, 1990, p.84) and meant that the more abstract concepts could be straightforwardly developed. This process was repeated for each question and a tabular format was designed to show the subsequent results, representing the evolution and logical relationships between: data, indicator, concept and result. 

Little did I know it at the time, but this methodology for analysing qualitative data has recently proved very useful. Trying to understand a site visitors: intent, expectations and wider issues, using only web analytics is never easy. We always have to: imply, create segments, trend etc etc; it is only through well placed surveys ( on and/or off-site),that we can know directly what our customers are feeling about us, sometimes the joy as a service has exceeded their expectations, sometimes unexpected pain, as parts of a product or service, maybe we aren't even in direct control of, are degrading our websites experience. Either way a qualitative source of data around a visitors experience has given me a whole new level of insight into both the expectations of our customers and what we need to improve.

The hard part once you start collecting  a large amount of qualitative data, which is often in the form of pure text comments is how the hell to analyse it!  Yes there are text analyzer tools out there, but will they ever know the intracacies of your business? I still struggle with trying to get away from the true value of reading every comment - the real gold, only then do you either feel the pain or share the love of your customers. Depending on how surveys are structured before any freeform text field, one should be able to try and categorise the issues to some extent, which helps with prioritising, but I am thankful to my degree dissertation for understanding how to analyse the customer comments we get, through being able to pick out, using coding and memoing what we need to improve.

Where next for web analytics tags?

Two stories in the web analytics space over the past couple of months have been running in parellel but strangely there paths haven't; the increasingly complex use of custom analytics javascript tags and the aquisition of analytics vendors.

Lets look at the first issue, "tag management", often analytics analysts are asked to measure something simply because of  poor back end reporting in another legacy system or because the business thinks it needs to know every granular detail of what a visitor is doing on their website.

If this sounds familiar then resist as much as possible, not only will it be a headache to implement but the business will then be expecting exact numbers, something analytics was never meant for, which will either lead to a lack of confidence in the numbers or some serious education required. Whenever custom tags are needed  ask yourself (and the business area asking for them), will data really ever get used? What value is it really going to add? Can the question the business wants answered be done through any existing analytics reports? With these in mind then at least any additional custom tags on a page will only get added which truely add value. 


Managing additional custom tags is not easy and in a whitepaper "When more is not better: Page Tags", one well known brand is said to have 28 different tags on their purchase transaction page, it's most likely that only a portion of these are analytics tags as its common to place any number of these tags on pages as well:

1) Adserving solutions 
2) Site optimization solutions, for example MVT tools
3) Affiliate marketing solutions
4) Search marketing solutions


....and that's without considering that a percentage of companies sometimes run two analytics tools! A solution suggested by Observepoint is to install a Chief Data Officer (obviously using their tools), but that doesn't get to the root of the problem, how as analysts we don't become weighed down by implementing tagging and add value in the area of providing actionable insight. 

Now onto the second topic, analytic's vendors, there has been constant movement in this area recently, with them being bought up by larger companies whose intentions are never quite convincing (well for me anyway): Omniture by Adobe, Coremetrics by IBM and the rumoured MS looking at Webtrends. At the Omniture summit this year I saw a demo of CS5 and how it was going to make it so easy for developers to put analytic's tags in flash but I still struggle with the fundamental fact that flash developers and analytic's just don't mix. 


So trying to be a little more radical I kept thinking there must be a better way, without the need for lots of custom tags to be able to implement analytic's tags without buckets of code, without having to ask developers nicely to embed our tags in their nice creative work (flash or otherwise). The options are either do away with javascript based analytic's altogether in place of log file analysis but this has its own issues, or host the custom tags outside of the page itself through a tagging framework, companies like Sitetagger and Tagman offer this type of service, but it still requires maintenance of the tags even if off page. 


No I want something radical! What options do we have for when the page is built, tags to be intelligently created by the system that builds the pages? Maybe based on the known attributes about the page that are already known by the system, not just elements of a page in the case of where the Adobe acquisition of Omniture is currently at.


What about the core system that constructs the page...the Content Management System? or ECRM? 


Brian Clifton suggests that "Ultimately, I do not feel an IT company, such as IBM, are best placed to move the web analytics industry to the next level", rather analytics should live in marketing, but marketeers are not interested in the tagging implementation, they just want to understand what customers are doing or not as may be the case.

I'd love to know what your thoughts are on this, should we just accept that the future is cumbersome, complicated tagging implementations or be looking further ahead?

Short or long...can we have both?

NB: This is an old post from a former blog that I never got off the ground.....


There's been two very different stories this week both centered around the issue of size (I'll try and avoid any cheap jokes).

In today's ever increasingly fast paced world, where we all struggle to keep up with consuming what ever media takes our fancy; be it new music via Hype Machine or Digital developments via Google Reader, the trend has been to try and shorten everything into   manageable bite size pieces.

The starter for ten obviously leading this trend was Facebook comments, then Twitter, the URL shorteners added to this, competing to get URLs down to as few characters as possible: j.mp now claiming the current prize for providing the shortest URLs (see Marketing Pilgram article for more info).

But where is this all taking us? Is this a good thing that has greater effects aside from enabling us to digest more information?

Earlier in the week a parody tool emerged, called Woofer; done as a side project by Jointhecompany, it calls itself a macro blogging tool, only allowing Woofs over 1,400 characters. Up for trying most things out I gave it a go and after getting over the 'what do I write about' issue I found myself realising:

 - How enjoyable it was actually thinking about what I was writing, rather than the almost throwaway nature of posts on Twitter and simply sharing an article on Google Reader.

- That the art of writing, something I'd not done properly since University about four years ago, is something although not naturally, I do enjoy and miss.

Here's the article I wrote: Why I don't do Apple

So with all the trends towards smaller and faster are we losing certain values? Does it matter? Will the short and long form happily co-exist?

Hopefully this has at least got you thinking about how you write, in whichever channels you choose to communicate and how their size limitations have changed your language and craft.