I came across an interesting eConsultancy interview with John Squire of Coremetrics on how he differentiates from other vendors.

I have a lot of respect for Coremetrics as a web analytics vendor that tries to be different. Rather than using the common sales hyperbole of saying “our tool is better because we have 10,000 features more than yours” (no-one cares about features that don’t provide insight, right?), he describes a number of areas that Coremetrics has worked on. I will let you read the interview in full, though I wanted to comment on one particular aspect…

According to Squire, Coremetrics is differentiating its technology by focusing on individual customer data in a multichannel environment. Linus Gregoriadis (eConsultancy interviewer and all-round good guy) makes the comment:

"This focus on website visitors as individuals makes sense for web analytics vendors
because a significant limitation of Google Analytics is the inability
to tie up analytics with data at an individual customer level."

This is perfectly true, Google made the decision not to track individuals in 2005 during the acquisition of Urchin Software Inc, the company behind Google Analytics. At that time, individual visitor tracking was in beta testing and in fact went on to become part of the Urchin Software suite – the server-side web analytics product that is also provided by Google.

I have discussed my take on the issues of privacy and Google before. To summarise, the view at Google (from the very top of the organisation i.e. Larry, Sergey, Eric), was that individual tracking, even anonymously, is a step too far and the web public does not want this. That view remains today, and I am a strong advocate of this.

Leaving aside the issue of privacy, is it valid to track visitors as individuals?

From a marketer’s perspective, tracking individuals sounds great in theory – you understand your customers better right? But if you receive 10,000 visitors per day and have weekly marketing performance meetings, that equals 70,000 data points to discuss? Best practice is to consider longer time frames in order to mitigate against calendar anomalies i.e. weekends v weekdays, holidays, the weather, force majeure etc… So for one month that could be 280,000 data points.

Of course no-one is going to look at each visitor session, so the first thing you do is either:

  • Aggregate i.e. group visitor types together. For example, visits that add-to-cart, visits lasting longer than 60 seconds, visits that viewed the special offers section, returning visitors who have previously purchased etc.)
  • Sample a smaller selection of “representative” visits for analysis and the potential quagmire of what that means. For a discussion of sampling considerations on the Internet, read the excellent sampling review paper of Nigel Bradley (University of Westminster)

Aggregation obviously defeats the point of tracking individuals in the first place.

Sampling, unless done with scientific rigour each time, introduces its own errors. So for me, tracking individuals (even anonymously and with visitor knowledge/consent) is not so relevant for the marketer or web analyst.

What do you think?