We all know that Optimization and testing efforts should be “data driven.” Quantitative and qualitative data should lead you to the problems that testing attempts to solve for your prospects.

But how, exactly, should quantitative web analytics data be used to power your Optimization efforts? How does analysis differ when you’re looking for insights and testing opportunities? Great questions, if I do say so myself 😉

I’m going to answer both in this post as I lay out a basic framework for how to “do” web analytics in the context of Optimization. I compare web data analysis to looking with different “lenses” at the same set of data depending on the context. For example, if I’m doing analysis on “site performance,” I would take a different view of the data than if I were doing “campaign analysis” or “content performance analysis.”

It’s the same for Optimization. You look at the same data as always, but you look through a different, contextual “lens” to suit your needs. As I’m looking at web analytics data for a client, I tend to work my way down through three basic levels, which I’ll explain as I go. The point is that I start “high level” and work my way deeper until I’ve gotten enough insight to plan a prioritized testing road map that guides testing over an extended period of time. The road map is crucial because it gets stakeholders and practitioners on the same page.

High Level Optimization Analysis – Customer Insights

This highest level of analysis should, of course, be done first if time allows. The goal at this level is to understand the target audience that is visiting the site in order to Optimize their experience and achieve your desired business goals.

This level is often aided by qualitative data like surveys, personas, and social media “listening.” Don’t be afraid to bring in qualitative data sources at this level to gain a broad understanding of the “hopes, dreams, and fears” of the target audience.

In terms of analysis I do within a web analytics tool, here are a few of the high level reports I rely on most:

1) Reports with “disqualified” traffic filtered out – I almost always create filters to remove a portion of the site’s traffic that I determine to be “not convert-able.” Examples of data I filter out are visits that hit the “careers” or “investor relations” sections of a site and visits with very low quality entrance keywords. It’s often surprising just how much traffic gets filtered out in this analysis.

2) Entrance Keywords – This is a good high level analysis that I posted about previously. It’s very insightful regarding the mindset of those entering the site from search.

3) High-traffic pages and landing pages – When building a testing road map, I need to know what pages are seen by large numbers of target customers. The larger volumes of visitors and pageviews allow me to iterate quickly through multiple tests and get the most business impact for my clients. I often export “top pages” and “top landing pages” reports, combine them, and use conditional formatting to analyze metrics like bounce rate, page value, and conversion rate.

Mid Level Optimization Analysis – Site Experience Insights

After doing your high level analysis, you’ll understand your target audience at a deeper level, you’ll have segmented their data, and you’ll know what areas of the site are realistically “test-able.”

The next step, at mid level, is to analyze the data specific to site experiences of interest. Some of this will be obvious from the get-go, like your key landing pages and your checkout funnel. Nevertheless, go through the exercise of analyzing your data (remember that you’re looking at a segment of “convert-able” traffic) focusing on these key site experiences.

Other site experiences that you may drill down into might be signing up for a webinar, using some sort of interactive calculator, or referring a friend to a service.

Here are a few of the mid level reports I rely on most:

1) Comparing top landing pages – Once I know which landing pages are test-able and interesting, I will start comparing them against each other. Again, I will likely export this data and use conditional formatting to look for anomalies and patterns. Do my landing pages all have similar time on page? Bounce rate? Do certain landing pages convert much better, or worse, than the average?

2) Funnel analysis – This type of analysis is pretty obvious, but shouldn’t be overlooked. Different web analytics tools provide different flavors of this report, but ultimately you’re looking at funnels on your site and seeing where the friction might be in the process. To take this analysis beyond the basics, try segmenting the data and understanding the page-to-page micro-conversion rate between steps of the funnel.

3) Path Analysis (leading to goals) – Once you’ve analyzed your funnel, it’s worth investigating the paths that lead up to entering your conversion funnel. For example, do prospects move directly from their landing pages into your funnel? Do shoppers visit a few category pages before they drill down and enter your checkout? Google Analytics’ new flow visualization reports can be of help here, or segment by those who enter your funnel and look at the pages they visit outside of the funnel.

Low Level Optimization Analysis – Page Interaction Insights

Referring back to my infographic earlier in the post, we now know our target audience, and their important (and popular) experiences on the site. Next, we need to analyze their more granular interactions with important pages of the site.

“Interactions” is a bit of a broad term, referring to on-page elements that prospects use while trying to accomplishing their goals. Examples might be using radio button selection, using sliders, interacting with flash, etc. And of course now with touch screen ubiquity, the number of interactions such as “swiping,” “pinching,” and “tapping” need to be analyzed as well.

Here are a few data sources I use at this level of analysis:

1) Non-pageview interactions – There isn’t a single report that I rely on for this, but I need to analyze interactions that aren’t captured as pageviews. In Google Analytics, you would likely be looking at Event Tracking reports. Some examples might be interacting with Flash or Ajax calculators, a checkout that doesn’t have unique pages, or watching videos.

2) Touchscreen/mobile usage – If a decent portion of your prospects are attempting to convert on tablets or smartphones, their interactions with your pages need to be analyzed and optimized. While you get some web analytics data on mobile usage, it may be more productive at this point to conduct usability studies on these specific devices to understand how prospects interact with your crucial conversion pages.

3) Heat maps, scroll maps, etc. – Web Analytics tools like ClickTale and CrazyEgg allow you to collect data that is interaction specific. For example, how often visitors click on certain elements, how far down a page they scroll, how their attention is focused, and more. This is just the kind of analysis that can help you uncover hidden conversion challenges and form hypotheses about design changes to be tested. If you haven’t played around with these types of technologies, they’re worth exploring. The trick is to figure out how much time to devote to them vs. your “regular” web analytics reports!

Conclusion

After following this framework and doing high level, mid level, and low level analysis, you should have found plenty of insights to drive testing. Remember that analysis finds problems, and hypotheses are potential solutions that need to be vetted. Start by knowing your target audience. Then, drill down to understand their core experiences. Finally, analyze crucial interactions. A quality testing road map won’t be far behind!

Are there other reports you think belong in the 3 various levels? Let me know…