Anyone who owns a website, web application, or any mobile app is also probably using web analytics tracking code with it. However, implementing these systems isn’t a straightforward process, and, often we find flaws in how sites have been instrumented.
At TUG, we consider this a missed opportunity. In this era of web analytics, tools have become highly sophisticated in tracking your users, yet are simple enough for interpreting collected data.
Data collected from web analytics tools can be put to a variety of uses, like:
- Filling holes in research data. Almost invariably, the user research we do is not all-encompassing because of time or budgetary constraints. In my experience, in such situations, data from web analytics perfectly complements the research data for creating personas, scenarios, and knowing more about our users.
- Acting as a feedback mechanism for implemented design changes. We can see in both real time and over an extended time how a particular UX design change is functioning and then make adjustments to get the desired effect.
So, to help you get started using web analytics data to complement your user research, I am going to illustrate a hierarchical structure. (Before we begin, be sure you are well acquainted with the process of implementing web analytics.)
Base: Investigate and Adjust Data Quality
Accuracy of data is the primary concern in web analytics tools because inaccurate data can result in facile interpretations. So, a thorough review and sanitizing needs to happen before you can build on the underlying data. This level is fairly technical—you may want to work with your IT team to make sure you are addressing these issues appropriately.
Some of the most common traps to watch for are:
Double counting / undercounting traffic: Improper implementation of web analytics’ script, improper HTTP redirects (http://www.iis.net/
In-site search data: Most of the time, the query parameters of in-site search are not recorded. This happens mostly because of incongruity between the in-site search engine’s query syntax and web analytics tool’s indigenous query syntax. In these cases, it’s essential to work with your search team and integrate the in-site search data with your web analytics tool.
Other blind spots: This is always a gut-check. I usually look for extremely high/low conversion rates, goal funnels, and behavioral info to determine if there are other holes in web analytics configuration.
Level 1: Play to Data’s Strength
At this point, you can start making use of the data by answering questions, proving/disproving hypotheses, and segmenting your audiences for granular analysis:
Ask web analytics what actually happened: Mature web analytics tools like Google Analytics store and present a lot of useful information about your audiences. These include (but are not limited to) user behavior, referral data, user characteristics, and e-commerce data. Additionally, Google Analytics has bundled-up a lot of analysis tools, which can be used to answer questions like:
- How much time users do spend on my website?
- What are the key content pieces users look at?
- What do users search for?
- Which products outsell other products?
- How many user interactions happen before a purchase is made?
- Which marketing techniques work better?
A full-blown analysis of your digital product’s web analytics data will reveal both anomalies and areas that are functioning well.
Confirm or disprove a hypothesis: Data can be used to confirm/disprove either certain behavioral characteristics of your users or website’s existential reasons. These results can then be used during your website’s redesign or re-architecture.
Here’s an example of how this might look:
Chain together factual attributes to see a lifelike slice of what really happens with a certain segment: Especially for your target audience, create filters that separate this user group so that you can dissect their interactions with your digital product.
Level 2: Be Opportunistic
Design feedback: One way to test your design implementations for any digital product is via usability tests. But often times they are too expensive and slow to provide feedback. In these cases, you can use web analytics benchmarked data to assess the impact of the change you just made. Within a day, you would have enough data to decide if further research is needed or if everything is going as expected.
Generate hypotheses: To model desired user behavior for your digital product, you need to continuously generate new hypotheses and then make design decisions based on outcomes. For example, if users are not interacting with the social widget on the homepage, then is it because of its position or structure or meaning? Generate hypotheses for each of these assumptions and test them.
Level 3: The New Normal
After you get past level two, analytics can help sharpen your organization’s strategic vision. Each objective of your strategy should have a set of Key Performance Indicators [KPIs], with a target assigned to each. Applicable data from web analytics should be fed into the reports that measure these KPIs, so that the management can get a holistic view of their product and take strategic decisions accordingly. For example, if one of the objectives is “Revenue from China” and the corresponding KPI is e-commerce conversion rate, if this KPI’s target is not met, then some changes need to happen in the website.
So by now, you should have a fair idea of the major levels that need to be traversed. With this information, you can plan, maintain, and use web analytics in your individual organizations. But if you’re a web analytics evangelist and want to know more about the aforementioned steps, stayed tuned, as we will unpack each of these dense levels in upcoming posts.