Blackbeaks Blog….All things Analytics

There has been some discussion in Web Analytics Demystified around the article that Shane Atchison from ZAAZ wrote describing a method called the hub and spoke method, designed to measure or at least attempt to measure non linear conversion processes.

Hugo Calvo responded by suggesting that Markov chains may be used in web analytics in the future. I’d never heard of a Markov Chain before so I checked it out. Wiki had this to say about it;

“A simple way to visualise a specific type of Markov chain is through a finite state machine. If you are at state y at time n, then the probability that you will move on to state x at time n + 1 does not depend on n, and only depends on the current state y that you are in. Hence at any time n, a finite Markov chain can be characterized by a matrix of probabilities whose x, y element is given by and is independent of the time index n. These kinds of discrete finite Markov chains can also be described by a directed graph, where the edges are labeled by the probabilities of going from one state to the other state that are on either end of the directed edge.”

Clear as the mud at the bottom of the drink huh?

What I think this means is that web analytics tools might start to get more sophisticated in the future. They will start by recording what happens in the first instance (but unlike the Markov chain, will record it), via a cookie as they do now. Then via some kind of new reporting mechanism (for instance a new form of segmentation method) that the analytics vendors come up with we’ll be able to see what they have then come back to do later. We’ll then be able to see the paths taken by users who left the site, came back later, to complete an action (or didn’t).

I know that sounds like simple segmentation of repeat visitors, but the difference would be that we’d be measuring scenarios of repeat visitors, who left from a particular place, who then returned at a particular place to complete a micro or macro action of some nature, and following the advice offered by Shane, could be based around a variety of different inbound and outbound micro conversions.

More importantly though we’ll also then be able to start predicting what scenarios to develop because of this new information and therefore the areas in which we need to work on in order to produce more conversions. Combine this with Persona based modelling like Persuasion Architecture and you’re getting really scientific. Not only would you be able to see how your visitors are acting and predict their next moves, but you would also be able plan your communication for them based on who you know they are because you’ve already researched devloped their persona’s.

In this way we might be able to start answering the big question posed in Shane’s article.
Is the linear conversion funnel dead? (or dying?)

I already know that the guys at Future Now have taken considerable steps with predictave modelling with PA and the combined use of web analytics. The question I’d like answered is are all web analytics systems likely to follow suit?

I’d be very interested to hear opinions on this and whether I’m barking up the wrong tree.