In an industry as young as Web Analytics, it may be hard to understand how one properly utilizes a Web Analytics solution to its fullest extent. In this article, I attempt to draw on my recent work experience to provide some insight into good principles of practicing the "voodoo magic" that is Web Analytics.

In the interest of full disclosure, this blog post was written by an Omniture employee and will reference their tools in examples. This blog, however, is in no way affiliated with Omniture, Inc.

Introduction

In many companies that want to "stay ahead of the game", one may generally find a sense of pressure to implement some sort of tracking mechanism for the organization's website.

A Mental Mistake

What most organizations who slap on a Web Analytics solution to their website do not realize is that they many in fact be doing themselves more harm than good. Solutions such as Google Analytics, Omniture, and Coremetrics have the ability to provide a binary ton of data for your use. The problem that arises is simply the fact that most executives do not know how to properly interpret the data into decision-making criteria. Thus, if your organization's executives make decisions based on misinterpreted data, that can lead to severe losses and expenses.

Then how do I use Web Analytics?

The key to properly using any Web Analytics solution is questions, QUESTIONS, and even more QUESTIONS! You need to have broad-based goals and questions that can then be simplified down into one or more analytical questions with measurable results. For example, you can have a simple question such as, "How well does our website attract visitors?" Now, you can break this down into multiple facets. You would want to consider what percentage of your site's visitors are newcomers as well as what keywords and search engines drive traffic to your site.

Of course, there are even more options to consider with this broad question, but you can start to see questions that have quantifiable results starting to rise up out of the fog of uncertainty.

A Visit Number report would allow one to see the distribution of visit times across any length of time, and thus you could see what percentage of visitors entered your site for the first time.

Now, simply seeing this number is not enough. To derive real use of this analytical tool, one would need to set a goal on how many first time visitors they would like within a specified time frame. Also, one would also want to consider the ramifications of having too many first-time visitors, as this would suggest people are not interested in visiting your site again. So in this case, perhaps the best goal would be an optimal percentage range to ensure that you are receiving the right amount of first-time visitors to your site.

Having a goal set doesn't do much for you either unless you act upon the data you receive from the report. We can see that for my site at the time of this posting, a vast majority of visitors are here for the first time. As this site has yet to reach critical mass, I feel this number is very reasonable. I can also see that I do have a few committed visitors to the site as well. However, I want to increase the visibility of my site before I start worrying about visitor retention. This deviation from the original thought is a classic sign of scope creep in an analytical sense. If we do not remain focused on the intent of the original question, then we can easily lose ourselves in piles of data--and then we would end up going nowhere.

So do you remember the question? How well are we attracting visitors to our site? For this report, I have decided that I want to see the first-time visitor numbers increase by 200% over the next month. To attract more visitors, I will engage in viral marketing and SEO activities. On a periodic basis, I will return to examine my growth or decline in first-time visits. In this way, I can discover what activities improve or harm my site's growth in traffic.

Trended 1st-time Visitor percentages (in blue)

From this trended view, I can see spikes where I improved new traffic to my site. However, the trend shows a decline from the inception of my site. This is normal to some degree as you develop a following of return visitors; however, it is clear that I can do more to attract new visitors. And this is all from one simple report!

Last but not least, I need to analyze my overall efforts as well as the efficacy of the reports I used in my analysis. Obviously, the use of one report in my overall goal to improve new traffic is a serious flaw, but this was only used as an example. Unfortunately, I have seen too many companies that depend solely upon one report like a crutch. I would recommend at least three differing reports to provide a 3-D perspective on the issue at hand. Only then can you get a glimpse of the whole picture.

Have a plan!

To tweak the common 90-10 adage, Web Analytics work is 90% preparation and 10% implementation! In the real world, as most developers know, the adage oftentimes becomes the 90-90 rule: "The first 90% of the [plan] accounts for the first 90% of the development time. The remaining 10% of the [plan] accounts for the other 90% of the development time." It would be more funny if it weren't so true. All too often do we (in the U.S.) set aside a small fraction of time in preparation to implement any sort of project, web analytics or otherwise.
The ideal concept behind my rule is that if you prepare well enough, you will only need to spend 10% of your time taking care of implementation matters. Seriously brainstorming and mapping out all of your web analytics projects can and should take a large amount of your time, because these plans will have the power to determine the course of your entire online platform. Here's the breakdown:

1. Discover your questions - involve executives and other decision makers in this process
2. Divide and conquer - segregate the questions into groupings by area
3. Map out the plan - one-by-one, go through each grouping of high level questions. Develop specific, measurable questions and goals. Ensure that you do not succumb to scope creep.
4. Double-check yourself - verify the goals set and the plan you have made coincides with the stakeholders' vision.
5. Implement your plan - You should know by this time what reports you will use and what changes you need to make in the tracking code (or other solution) to implement this plan
6. Evaluate - On a periodic basis, examine your progress within your plan. If need be, you can make minor adjustments. When the plan has been "completed", you can analyze and revamp your plan for a second run if you so desire.

One Final Note

So, you now have a good idea on how to make Web Analytics work for you. There is just one final note of caution that I would add in whatever solution you use:

Realize that you have very little control over the exact details of web site tracking.

If you imagine a web site visit in the three components of a browser, the Internet connection that transmits the data, and the server that hosts your website, you only have control over 1/3 of the entire process! What does that mean? As much as you try to maximize browser compatibility and ensure Internet bandwidth, there will always be unknowns and outliers in all the data that is collected. Thus, you should never expect your web analytics data to match up 100%. EVER. PERIOD. The key to this type of data analysis is to identify trends as opposed to looking at the raw numbers. Trends and overall ratio analysis will give you decision-making criteria. If you base decisions on a raw number, then you are not looking at your analytics data from a robust enough viewpoint.

Are you using Web Analytics on your site right now?