How to determine the best metrics for a website

By Mark Kimathi

What metrics are the best for a particular website is not always obvious. But succinctly speaking, they should always lead to increase in website performance, whatever the expected performance, be it more sales, more page views, more subscriptions or even more visitors.

As pointed here before, there is really no particular metric that can be pontificated upon. Similarly pointed out, is the need to start tracking, analysing and testing one’s website immediately.

It is tricky as you need to track yet do not know exactly what to start tracking. The key to unlocking this stalemate is just to start, probably with general best practices, what are often agreed upon as Key Performance Indicators.

For example, track conversions through your persuasion funnel, or track Bounce Rate for the few most visited pages. This will get you creating an idea of what your visitors are doing on your site vis-a-vis what you want them to do.

It will also raise some questions that can give you further direction for investigating and hence get the ball rolling. The following outlines a strategy for identifying and retiring metrics along the never ending quest to understanding what happens on your website.

Metric life cycle

Avinash Kaushick a respectable web analyst calls it the "metric life cycle". The cycle is borrowed from a business management system called Sigma Six.

The system entails two methodologies each with five steps. The first, Design, Measure, Analyse, Improve and Control (DMAIC), is used to improve an existing business process. The second is Define, Measure, Analyse, Design and Verify (DMADV), used to create new products or process designs. Our metric life cycle is a mash of the two. Both have slight variations even in similar steps.

Based on our website proposition (reason the website exists), we define most likely candidates for a good metric. We then go ahead to measure and analyse the relevant data. So far we’ve been fully DMAIC, but from here it gets blurry.

This is because at this point, either the analysis is giving us some direction for action to improve performance or it is not. When it gives direction for action, it often will require designing a test to verify if that action will result to improvement or not. At this point we start borrowing from DMADV.

On the other hand, if the analysis results to the chasing of one’s own tail then you have identified a metric you should not be tracking – at least for now. Retire that metric and find a better one from the insights you have collected thus far.

Consequently our metric life cycle would look like this Define, Measure, Analyse, Action, Improve or Eliminate. Since its perpetual cycle we keep repeating, inching towards ‘perfection’.