Dispatches from Searchnomics: Google's Avinash Kaushik on Advanced Web Analytics
[Things have been busy lately; I'm finally catching up on some older posts I've been meaning to get to ...]
I've written before (see: Google Analytics - take two , Bounce Rate ) about Avinash Kaushik - Author, Blogger and Google Evangelist. He's a recognized authority in the relatively new field of Web Analytics; his blog, Occam's Razor, is one of the highest-rated blogs on this topic. In his trademark style, Kaushik gave a highly entertaining and very informative presentation on Advanced Web Analytics at the recent Searchnomics Conference.
In his talk, he focused on Seven Tips that you can use Now! to implement and improve web analytics for your web site.
1. Join PALM! (People Against Lonely Metrics)
Kaushik
began with the key point that one should never look at the data or
chart for a single metric in isolation; when studying analytics data,
one should always look at multiple metrics together [otherwise the
metric gets lonely!]. For example, when looking at how many users
viewed a single page or how many bounced - you want to look at those
together, along with the site average, delta changes and indexed
performance. Indexing metrics, especially, highlights the items that
are working particularly well or badly. In this way, you can turn the
metrics into actionable information.
2. Give your data Context
Associated
with the first point, Kaushik suggested that you should look at all
your data in context. You can compare any given metric or set of
metrics to either the average values for your site as a whole, or to
the same metrics during a prior time period. He used Google Trends as
an example, showing how you can compare two companies at different
points in time.
3. Segmentation Rocks!
He
pointed out that a web site audience is rarely completely homogeneous;
instead, it is composed of many users who can be broken down into
different segments. It is important to partition the data into segments
according to the type of user. For example, segments can help you
compare adword results vs organic results; if the bounce rate for
adwords for a given segment is high, you are effectively losing money!
Most important: web site improvement actions are always implemented for specific segments, never for aggregates.
4. Enter the Matrix: It's a Multi-Dimensional World
Since
real-world marketing proceeds along multiple dimensions - not just web
site traffic, but also Emails, Ad words and Campaigns - it's important
to analyze the data using multiple metrics along all these dimensions,
in order to optimize your spending and conversion rates.
5. Think Looong Tail. Everything.
Kaushik
explained that all of these Analytics graphs - Clicks, Visitors and
Budget Spend - have something in common: a large head and a long tail.
Similarly, when looking at Keywords: the expensive, branded keywords
are all in the large head, while keywords characterized as
generic/category/"early bird" are all in the long tail.
Typically, according to Kaushik, a lot of the AdWords spend goes into the head; but if you have good SEO implemented, you should be highly-ranked in the organic results anyway, so there should be no need to bid on it. On the other hand, long tail keywords, which are usually a lot less expensive, are used by end users very early in their buying process; if you want to focus on new customer acquisition, these are precisely the users you would want to target.
A critical insight: Web Analytics should not just provide reports, it should provide solutions to business problems.
6. Web Analytics 2.0 is Qualitative
The
key questions Web Analytics should be trying to answer, for each user
segment, are: What is the user here to do? How easy or hard was it for
her to do it? And, finally, was the user's problem solved?
The bottom line is to consider the Customer Intent, by looking at a variety of metrics. Looking at Conversion Rate alone will not be sufficient to do that.
7. Web Analytics is Testing!
Experimentation
and Testing is an integral part of Web Analytics; otherwise,
decision-making degenerates to using HIPPO - the "Highest-Paid Person's
Opinion". E & T is your one powerful weapon against HIPPOs: you
take the best ideas everyone has, and run them through tests. Let the
data show you which ones are the most effective and which ones are not.
Kaushik pointed out that Google has a Multi-Variate Testing tool called the Google Website Optimizer, which allows you to test multiple ideas at the same time without the need for any IT changes.
----
Related Reading
Search meets Web Analytics at Searchnomics Conference by Jeremiah Owyang, where Jeremiah does an excellent job of covering an equally useful Web Analytics presentation by Eric Peterson of Web Analytics Demystified.
