Web metrics that matter

Came across this article on CIO recently:
http://www.cio.com/archive/111502/matter.html

It proposes the most important set of web metrics for the following categories of sites:
- Business-to-Consumer/Retail Sites
- Content Sites
- B2B Sites

Obviously these are pretty broad categories but the notion that actually there are actually only a few identifiable key metrics that matter for a given sector / type of site is an interesting one.

Interesting because two of the biggest challenges with web metrics are 1. What should we actually measure and care about? and 2. How do we avoid information overload with all that *can* be measured?

Although the proposed metrics are somewhat limiting I'd broadly agree with what is put forward. As a content site we look most carefully at:
- subscriber numbers and trends (so we know how much money we're making and how we might do better)
- total unique users in relation to repeat users (to understand our reach, our ability to get people to return and our overall conversion ratios)
- loyalty segments (based on recency, frequency) in relation to subscribers (to understand whether more loyal users are actually more likely to subscribe. You'd think so, but not necessarily true...)
- clickstream (to optimise traffic flows around key pages e.g. homepage, subscribe page, registration process etc.)

I reckon we spend 80% of our metrics time and effort on just these key few metrics. The other 20% is on ad hoc stuff, campaign analysis, search engine marketing related analysis etc.

Anyone else agree / disagree that probably the majority of your web metrics time is best spent on only a few key metrics?

Agree entirely Neil. I've just done some work for a large financial services company on enhancing their web management information and the process/hierarchly worked broadly as follows:

- the big questions (what the business really wants to know)
-- the value drivers (what are the key levers/influencers towards answering the big questions)
--- the reporting & analysis (how are you going to present the MI, to whom, through what channels, when, what analysis is required etc.)
---- the metrics (what fundamental measurement units are required as the building blocks to do the reporting which is often composite metrics)
----- the tools (what tools do you need to help you get the metrics)
------ the data (what data you need, in what format etc.)

Typically I've found a "top-down" approach works much better than a "bottom up" one i.e. you don't start with 'what have we got and what can we measure?' rather you start with 'what do we need to know?'. In practice the "bottom up" approach is still done to some degree as you need to know the parameters and constrainsts you are working within to define the best solution.