Let's say every day I get in my car and drive to work; I have defined KPIs for fuel and engine speed and I have warning triggers on things such as battery charge. In other words, I have a dashboard, or in an analytics metaphor, I have a site dashboard. But my car is sick - what do I do?
I may get an alert to tell me what has happened, but most of the time something will occur without warning - just like with our site dashboard. If our conversion rate falls from 8.4 per cent to 2.1 per cent over the course of a month, we may not realise until the next dashboard is due. The fix requires the diagnosis to proceed to rule out causator events. Hence our reporting and analysis services should be complementary, with careful alignment of both practices."
THE CLIENT: Marianina Chaplin Manning, Head of web analytics and optimisation, rightmove.co.uk
"People describe the internet as an incredibly measurable medium, but a lot of measurement can happen in the wrong way. For example, there is a growing school of thought that interactive advertisers could be misusing display ads by over-emphasising clicks and diluting the power of branding and communication. I think this is because the first is much easier to measure.
The best way to measure online behaviour is to be very clear about what it is you are trying to measure in the first place, and then structure a communications framework around what you're doing that makes it measurable. It is likely that you will end up using a variety of pre- and post-campaign techniques that will include time in the usability lab poring over web analytics, campaign results, online effectiveness studies and surveys, focus groups and campaign data.
There are plenty of stats that can tell us what people are doing, and with time and effort data from different places can be connected in ways that tell us interesting stuff. But there must be clearly defined criteria from the start."
THE AGENCY: Mark Iremonger, Head of digital, Proximity London
"Use the most appropriate metrics to achieve your goals and then segment those metrics to make them actionable. For example, if you want to improve conversion then measuring conversion and the sales funnel for different segments is best.
Segmenting by behaviour based on pages seen previously or products purchased is a good way of identifying the intention of the consumer - past behaviour is the best predictor of future behaviour. If a consumer goes to a particular content group on a website, he is likely to visit it again in the future. It's worth noting that repeat visitors are eight times more likely to buy than first-time visitors, so use clickstream data to send targeted, segmented emails to customers and bring them back to your website.
Measuring consumer engagement (the online equivalent of offline responsiveness) is a good way of identifying the degree to which an individual is likely to do something.
Whatever metrics you use, the greatest pitfall for effective online behavioural analysis is poor data quality. But that's a whole other column."
THE SUPPLIER: Mark Patron, Chief executive, RedEye