Web analytics, web metrics, and how to use them to track and improve conversions

Picked up a new client recently and used Omniture Site Catalyst to determine how much Social Media Traffic this site had.  Turned out it did not have much (.15% counting blogs and 1% if we include traffic from Forums).

Anyway, without getting into site specifics - here's a list of general steps I took to determine Social Media Traffic to a site; I'm not saying these are the only steps - there's probably more that can be done - but here's a start.

1. Using WikiPedia's list of Social Networks as a good starting point - I set up a match-string filter to pull out all the referring urls containing the following -

blogspot technorati myspace 43things facebook flickr linkedin LiveJournal meetup orkut stumbleupon reunion tribe xanga care2 blurty twitter xml feedburner rss forum

Besides blogspot and LiveJournal, which had about equal the amount of traffic - traffic from Forum sites was much more substantial - and I bet, more commerce based, believe it or not; this will vary depending on the type of site.

2. I took the Forum traffic and broke it down by URL string and isolated each Forum domain (took me a little bit of work in Excel to do that) and counted the number of instances of the Forum domain plus the overall number of visits from each Forum Domain.  I ended up with 5 domains sending 65% of the Forum traffic.  My idea was to hone in on Forums that were more active and report back on it. 

If I had more time maybe I'd look at the content on the Forums - or if I had the tools, do Semantic Analysis of the content on Forums - but I don't have those tools at my disposal right now - so I'm not going there - just yet (clearly - that is something to do - since the URLs that are talking about a site and linking to it - can be categorized based on content and what they're saying....about the site).

3. I used SEO Spyglass to check out the indexed blog, messageboard and forum traffic to my client and found it to be a fraction of what was contained in their Site Referral logs (via Site Catalyst).  That doesn't surprise me as most tools can only query and get the first 1000 URLs from any search engine - no more is provided and many of the Social Media content is in the Long Tail - in the other millions of URLs that are not in the first 1000 URLs reported back.  

I was fine with it and only used the search engine indexing in Google, Yahoo and MSN as an additional perspective; it's the data in Omniture that really mattered, in this case.

4. I data mined Meetup Groups (just a little - don't have the time to join all of those that linked to my client just to see who is a member and what they're saying) - I see huge potential in exploring Meetup groups for all kinds of insights - and just to monitor the conversation - if nothing else.

5. I pulled overall popular pages of the clients site using Site Catalyst- but I ran into a snag.  I wanted to create a custom segment of all the Social Media URLs that I identified using the filter in step 1 - but my client's Omniture account does not have Discover 2.0 enabled - so I was unable to compare visitors from Social Media sites vs. overall visitors on things like:

•  length of visit
•  pageviews per visit
•  most popular pages
•  conversions
•  most popular paths
•  registers for a login
•  number of logins to clients "my portal" page

I usually start my analysis of a site with a premise that I want to prove.  In this case, I want to prove that Social Media traffic is more likey to convert (buy something) or is more engaged with the site (because Social Media is part of a conversation - so you'd expect more engaged activity) - but in order to really prove it in a timely fashion, using Omniture - I needed Discover 2.0 enabled.

Now, maybe there were ways to get that data for all the URLs (blogspot technorati myspace 43things facebook flickr linkedin LiveJournal meetup orkut stumbleupon reunion tribe xanga care2 blurty twitter xml feedburner rss forum) in real time without Discover 2.0 - treating them all as one segment (ie: Social Media Traffic) but if that was in the Omniture 12 release (minus Discover 2.0) I did not see it - and I looked all over the place.

So here's when you start needing to go all the way with Analytics.  Say I could prove that all the traffic (in blue - my filter) did interact more with my client's site - and on average, per visit, bought more product.  My case is made - Social Networks here we come!   But I needed the full power of Omniture to prove it - and it's a pity I did not have it - as I wanted to explore this idea more fully - which is what makes it fun to be a Web Analyst.

6.  When I was done with all that - I was in a pretty good position to say the client only got 1% of its traffic from Social Media.  But then the question becomes ....is 1% good or bad?  Well, it depends what kind of site is and what kind of Social Media traffic the client's competitors are getting.

Well, here's another snag .. I could not easily find any published stats showing what "average" site gets in "Social Media Traffic".  I could look at sites that I have access to the analytics and back end data - and if I had the backend access to Compete.com's data warehouse or even a full blown HitWise account - I'd be in a position to say something like:

"most sites in your sector get 6% of their traffic, on average, from Social Media sites buy yours gets 1%".

But I could not say that because I did not have the data to prove what the other sites get.

However, it would help a lot, in Social Media Analysis, if Compete would publish a breakdown of where people came from, not just the number of people that came to the site (which is still under counted by 25% in my client's case).  For example:

• number / percent traffic from Social Media Sites (blogs, message boards, forums, video and audio sites), etc
• number / percent traffic from Search Engines (and specific numbers / percentages if they have the data broken down well enough)
• number of "no referrers" (or "Direct" referrals) who bookmarked or typed in the URL.

I hope Compete will  be doing something like a Social Media categorization of each site's traffic in the future (Hint...Hint ... Compete - can you start working on this if your not doing it already?).

7. I ended my analysis by suggesting some Social Media Resources that could be used to enable the community that already comes to the clients' site to communicate with each other and the client (which is how it would be "Social Media" in the first place).

• www.Kickapps.com  for Social Network building (or magnify.net, or ning.com)
• www.combinedstory.com for Second Life development - they're right around the corner from my office and I took a Second Life Immersion Training from CombinedStory - it was excellent and I highly recommend it.
• V-Tracker for Second Life Web Analytics - I wrote about this a lot lately in V-Tracker - A full blown Web Analytics Package for Second Life from Code4Software.com.

I like to recommend companies, services and ideas that I have personally looked at and reviewed - that's what I did above.

To end this post - I want to add that my client does not have any Social Media content on their site - no blogs, no forums connect directly to the site content (but plenty that were going on from other sites coming in) and really - not much to dig into that they produced or enabled themselves.  

I could have gone further into my analysis if there was more Social Media to analyze - but in this case - ie: number of postings, citations, etc - and you'd expect that kind of data should also be part of the analysis of the Social Media traffic to / on the site - but again, in this case, there was not much to sink your teeth into - so I went with analysis pulled out of Site Referral traffic.  

Even common sense suggests 1% is a low number given what's going on in Social Media today (if I knew nothing else - it be easy to say 1% is low, in other words). 

I proved my case anyway - Social Media Network and tools - here we come. 

And that just goes back to underline Web Analytics starts with an idea (same as Art - and Science, by the way) - a hypothesis - which your pulling data and analyzing the data to prove or disprove your idea.  

I could have easily pulled data and found the client had a lot more Social Media traffic (but they did not) - in which case - I'd have to re-examine my hypothesis.

This post is long enough - time to stop.