I introduce you to the moving average.  The concept is simple, replace each data point with the average of the last X data points (including that data point).  No more looking at graphs like the following and trying to understand the overall trend...

Instead, use a moving average graph like the following twelve-month moving average to filter out seasonal noise and get a clear picture of what is going on...

Omniture SiteCatalyst has a built in feature to do this.

I have a strong warning to give about an ugly mistake you can only make once - Make sure you have enough data points preceding your starting point.  For example, in the above graph, the first data point with a calculated moving average is September.  For this graph to be correct, the preceding eleven months need to have complete data.  For this data set, the month preceding the first October is when tracking began, so the twelve-month moving average can only be calculated twelve months from October..  If I had started my moving average at the first June, then the moving average would be very low because I am missing three months of data in my computation (Jul-Sep).

How about a little insight into web analysts' behaviour instead of visitor behaviour today? 

The web analytics vendor (Omniture) we use provides us with four different tools that offer varying degrees of complexity.  From a "big picture" tool, ClickMap, all the way to a deep dive tool, Data Warehouse.

In discussion with a couple of other Sun web analysts recently I made the comment that there are "too many tools.  I guess you can't build a house with just a hammer."  The tool you need really depends on how much precision, data confidence and expense you want.  We do this decision making continually and think nothing of it.

It is very important to consider your confidence level with each tool's output.  Most of my work is done somewhere in between with SiteCatalyst and to some extent, Discover.  I just don't trust ClickMap very much.  The reason I hardly use Data Warehouse is because of time constraints.  This is another factor web analysts consider when choosing a tool.  I usually don't want to wait days for information.  So I usually prefer a lower level of confidence over a long wait time.

Another factor is cost.  The smaller the level of confidence, the shorter the wait and the smaller the cost.  Cost is really a factor of wait time and is not determined by the level of confidence.  The level of confidence is a by-product of the wait time.  Quick is cheap and dirty.  Lengthy waits beget precision and are more expensive.  You can see that precision and level of confidence are intertwined.  Cost begets precision which begets confidence level.

What I find interesting is that we usually skip the first step mentally - Cost.  We just automatically assume that we can't wait and can't spend money.  While the latter is usually true, wait time isn't always as expensive as we think it is.  In this fast-paced world, faster is assumed to be better even if precision is lost, so long as our level of confidence doesn't run dry.  I guess this is just how business works.  We strive for a balance between precision and confidence level to get answers as quickly as possible at the smallest cost.

I'm curious about how much you think about this on a day to day basis.  Do you prefer low cost, high precision or high confidence levels in your analysis?  In which situations would you prefer one over the others?

Happy analyzing your analysis behaviour!

Did I get your attention?

I've spent the last twelve months promoting minamilist web analytics reports focused on a handful of KPIs that were supposed to lead me to the low-hanging fruit.  You know what?  It's not there!  Well, it might be there, but it's not one large insight.  It's made up of a bunch of seeds of insight which represent a bunch of small changes.  I'm not saying that minimalist dashboards don't have their place.  I wouldn't advise taking them away from your executives.  You just need to know your audience and what you're there for.

[sidenote:  I came to the realization today that many of our web analytics thought leaders (and I love them all dearly, because I attribute so much of my knowledge to them) forget the trench practitioners (like myself) because they spend the bulk of their time advising the mucky-mucks and their web analysts.  Who can blame them?  That's where the money is.  After all, it's not the trench practitioners who make up the majority of their financial support.  As a result, much of their discussions relate more to higher-level web analytics practices.  I think they should start their discussions by mentioning the intended audience.  Of course, when taking free advice, it is the advice taker's responsibility to assess its usefulness and appropriateness to their situation, especially if the intended is not stated.  Now there's something they don't teach you in college!]

Here are some of the pitfalls of searching for low-hanging fruit: 

Pitfall #1 - If it's low-hanging, shouldn't it be so obvious that you don't need a dashboard to find it?  In fact, it probably doesn't even need a web analyst to be found.

Pitfall #2 - If you believe that low-hanging fruit is where it's at, then what will you do when you've found it all?

Pitfall #3 - Would you rather have four large achievements scattered throughout the year or would you rather have numerous small achievements every month to keep you going?

I have yet to run across that one magical factor within someone's conversion funnel that makes them oil-rich overnight.  You know what it is?  It's a bunch of small factors, that, when combined make a noticeable difference in the success of the website.

Why is it so great focusing on the small things?

Bonus #1 - It's easy to get commitment from others.  Seriously, who's going to have commitment issues over a ten minute site update?

Bonus #2 - You won't need for approval from upper management.  If they have time to fuss about the meta tags on your downloads page, they have too much free time.

Bonus #3 - Small changes require very little time investement.  And if doesn't work as expected, you haven't lost much time.

Bonus #4 - You'll have tons of fodder for your employee review.

Bonus #5 - Since commitment is easy to acquire, more people will become involved.  They will all feel more closely attached to the success of the website.  You will also have a TON of advocates.  You will become a star.  You will be young forever.  Ok, I made that last one up.

So given today's enlightenment, how do I feel about my recently updated web analytics process?  I still like it very much.  I had a lot of fun developing and implementing it.  I think I will probably do a lot more intensive analysis and be careful about who I recommend the rest of the process to.  I also think that when using the full process, I will make sure that I don't just report the KPIs, but make sure that I am always digging deep beneath them.  I won't provide advice in such broad terms, such as, "We need to do more SEO to improve our search engine traffic," or, "We need to increase our referrals to training."  I will followup up with a more detailed analysis that suggests some small changes and get commitment.  I'm already discovering that commitment follows without requests or hesitation.  I must have heard the phrase, "I'll take care of that," a handful of time this week.  I don't remember anyone saying, "I'll work on improving our SEO."

Ok, so I've ranted late into the night and as I read back through this I'm wondering if I really explained myself well.  I think what I'm really getting at is that high-level dashboards are not a substitute for deep analysis and insights.  They are a means to and end, and without the end you are just making laser pointer dots on a screen.  Do I feel that I've been misinstructed by web analytics experts?  NO, experience and trying it out have made me realize that it was me that made certain points bigger than they were (or were they really presented in a fashion that made them seem bigger?)

Comments???  Am I a loon for dissing low-hanging fruit?

After having developed (with the help of numerous blogs and books) a my web analytics strategy over six months ago, I am faced with a new challenge - the reality that a one-size-fits-all strategy doesn't always fit.  The "bodies" are not always in "alignment" in order for a goal-based strategy to operate.  And sometimes, pages/sites need intensive analysis before a goal-based strategy can be valuable.

Now I will contradict myself.  A one-size-fits-all strategy doesn't always fit unless it is a living, breathing organism that you are willing to restructure as you go along.  By adding a new "step" to my current strategy process, it will now fit all of my current challenges, thus making it a one-size-fits-all solution once again.

So what's this "new" step?  One-time, thorough, intensive analysis; or, for short, intensive analysis.  I have placed this optional step at the beginning and end of my current process:

Ok, great, so what is inside this new box?  I put together a preliminary draft of what I'm including in my intensive analysis.  Of course, once I start using it, I expect the list and it's order to change drastically as I'm sure I will realize some obvious territory that I neglected to include.  Here is my intensive analysis list with some scribbled notes on the content of each:

1. Visitor Survey Analysis - Numbers can be picked apart, but when the customers speak, they're clear.  I pattern my surveys after the 4Q survey (Thanks, Avinash Kaushik).  It's important to get your survey up right away.  You don't want all other areas of analysis to be over and then have to wait weeks for ample responses to be collected.  Try connecting visitor feedback to metrics for stronger recommendations.  Multiple recommendations may come out of this analysis.
   
2. Heuristic Analysis - This basically means to learn about the page/site from the creator and the visitors perspective.  One area I recommend looking into is page size compared to visitor browser size and desktop size.  Another idea is to gather a list of best practices and compare the site to them.  You may not come up with any recommendations with this analysis.  The main point is for you to wrap your head around the intended purpose and architecture of the site and what the visitor sees.  It's a learning experience to prepare you for the remainder of your analysis.  Most Importantly, many ideas will pop into your head of potential areas of analysis you should perform.  Jot these ideas down and see if they fit into any of the following analysis categories.
   
3. Create Visitor Profiles - I thought of this one while halfway through this list.  Try to understand the different types of visitors to your site and create a sample profile of each type.  Keep in mind that this is just another way of being in your visitors' shoes.  Eric Peterson's book, Web Analytics Demystified, has a short section on this topic that provides a nice introduction to the topic.
   
4. Conversion Analysis - Determine what conversion(s) this page/site is concerned with and analyze the performance.  Study the conversion path and make recommendations accordingly.  There is a plethora of information available about conversion analysis.  Refer to it as necessary and then some!
   
5. Revenue Analysis - Determine (or approximate) the revenue stream of this page/site.  Approximate?  Yep, some sites are strictly for the sake of providing content (as with many of the sites I support).  However, there is usually a way to connect them to revenue even if it isn't one of their goals.  As I always tell them, "You're gonna be asked about it sooner or later; and if not, then you'll just make yourself look even better."  It's a win-win!  Execs understand dollar signs (even though they may talk about volume all day long)  ;)
   
6. Incoming Traffic Distribution Analysis - Where does the traffic originate from?  Look for areas of improvement.  Save your digging around in search traffic for the next analysis.  Segment your various reports for typed/bookmarked, website, or search engine entries.  If you want to get really crazy segment various reports for specific websites or search engines.
   
7. Search Engine Analysis - Put on your SEO hat!  Every web analyst has a little SEO in him/her.  Do I even need to detail this?  Nah!
   
8. Content Usage Distribution Analysis - This is something I got from Avinash.  How much of each content type do you have versus which types of content are your visitors spending the most time?  A combined column distribution graph will tell a ton about what you're not providing enough of (or too much of).
   
9. Navigation Analysis - Where are visitors going on your site and how are they getting there?  Is your navigation confusing?  You could spend a week on this!
   
10. Competitive Analysis - This is probably the least plundered area of web analytics.  Just looking at what all the web analytics blogs talk about reveals how little this area has been studied.  We all know about the tools out there, we just need a little nudge to start learning how to apply them (myself included).  This is last because there's no sense in studying the competition if you don't understand yourself.  How does the competition do things?  Can you improve on an idea of theirs?  Be sure to provide an overall idea of how your website is doing compared to the competition.  Everyone has a competitive streak.  :)
    

Many of these items overlap.  You'll have to decide to dig in now or write it down and save it for later.  In nearly all of these steps, you should end up with at least one recommendation.  If you have multiple recommendations, order them by considering importance, ease and ROI.  In the end, provide a complete list of recommendations in the order you think they should be considered.  You should come up with nine or more recommendations.  This intensive analysis could be done in as little as a week (if you have nothing else to do or you aren't "intensive").  Realistically, I expect that these will take me two to four weeks.

I'd love to hear feedback about other areas I missed or if you've put together such a list yourself.

Now that I have started moving through my web analytics strategy development process with a handful of teams, I am starting to realize that securing time at weekly meetings can be quite a challenge.  It seems like everyone wants weekly updates, but they don't have time for discussion and planning.  And something tells me they don't have time for implementing their improvements.

The initial web anlaytics strategy development process takes considerable effort on the team's part - usually hours.  By the time the process is complete, they need to return to items that were neglected to make progress in their web analytics strategy development.

So rather than getting thirty minutes a month in one chunk, I've decided to take an alternate route (at least at the start).  I've begun requesting just five minutes at the outset of weekly meetings to give a short energy burst update on web analytics.  My goal is to skim the KPIs for trouble spots, point them out, and have a quick discussion on resolution.  At the end of five minutes, the discussion will close and can be taken to email throughout the week if necessary.

In this way, every week the team is thinking about web analytics and the role they can play.  Over time, the team may lengthen the five minutes, or they may do so well at email discussion without the need for round-table discussion.  Maybe you are fortunate enough to be working with teams that have ungodly amounts of time to spend on web analytics; or maybe you are like me, thrust into an industry that is fast-paced and reaching for the caffeine, taurine or whatever else makes you faster!