By Brian Strauss, Director of Sales Engineering —
In the crowded field of web analytics and website optimization, companies are queuing up like contestants at an American Idol casting call to peddle their trendy offerings in “real-time” or “ad-hoc” segmentation. While it might be clear after the first few bars of a demonstration which offerings are merely “good intentions” and which are “serious business,” it can be quite time-consuming to parse the latter group for solutions that offer the real game-changing value: actionable data. Let's face it, most of us don't have Steven Tyler and an audience of millions to help us choose the winner.
Image courtesy of: www.hollywoodreporter.com
What is this ad-hoc segmentation business anyway?
Web analytics is a reporting world. You think up the reports you want, determine which segments you want to monitor, tag your pages, and then get down to the business of counting things. In the past, this has been an important source of information when it comes to monitoring a web site, but it's a fairly static environment. In the modern world of analytics, we rely on our data not only to provide metrics—how many visitors, abandonment and conversion rates. Yet we also want to use the insights we gain from the data to drive site optimization—to justify changes to the website that positively impact the business.
In the crowded field of web analytics and website optimization, companies are queuing up like contestants at an American Idol casting call to peddle their trendy offerings in “real-time” or “ad-hoc” segmentation. While it might be clear after the first few bars of a demonstration which offerings are merely “good intentions” and which are “serious business,” it can be quite time-consuming to parse the latter group for solutions that offer the real game-changing value: actionable data. Let's face it, most of us don't have Steven Tyler and an audience of millions to help us choose the winner.
Image courtesy of: www.hollywoodreporter.com
What is this ad-hoc segmentation business anyway?
Web analytics is a reporting world. You think up the reports you want, determine which segments you want to monitor, tag your pages, and then get down to the business of counting things. In the past, this has been an important source of information when it comes to monitoring a web site, but it's a fairly static environment. In the modern world of analytics, we rely on our data not only to provide metrics—how many visitors, abandonment and conversion rates. Yet we also want to use the insights we gain from the data to drive site optimization—to justify changes to the website that positively impact the business.
In the traditional web analytics world, if changes in the metrics you've been observing raise questions for you or make you hungry for more information, testing your hypotheses can be a long and tedious process. In the best of cases, this involves more coding to get more tags into pages so that you can gather additional data points. Now, there are major problems with this. For example, waiting can be costly – the longer you have to wait to get actionable data that allows you to optimize your site, the more money the lack of optimization causes the business. Waiting can also make it hard to hold the thread. Our days are busy with lots of competing priorities. If we have to wait days or weeks to solve problems, we test the limits of our attention span, which means those problems may never get fixed. Ad-hoc segmentation promises to solve this challenge by allowing the analyst to combine data from different dimensions in our repository to create new segments on the fly and allowing for more rapid hypothesis testing. In short, you are supposed to be able to come up with actionable insights far more rapidly.
Are there solutions that really deliver?
Modern web analytics products employ a form of ad-hoc segmentation and promise an ability to “discover” insights rapidly by providing a framework where you can drag and drop different dimensions from your data pool to aid in analysis. This sounds promising because it allows you to go from a state of general awareness of increased abandonment to understanding that this trend is largely contained to, say, first-time visitors from a particular campaign–effectively allowing you to zero-in on the actual business problem. That's a good first step but in the end, with this brand of ad-hoc segmentation you are still limited to diagnosing the cause of the issue with the same high-level aggregation that you tagged for in your last build. While you may have a solid framework for identifying an underperforming segment, you don't necessarily have any tools to help you determine why that segment is actually underperforming. That's because there are some fundamental problems with doing optimization with web analytics data:
- While you can see a business trend or realize an underperforming segment, the data is too high-level to tell if the segment is having trouble due to marketing problems, application issues, or design flaws.
- What if the data you need isn't in your tagged data set?
- How can you really discover new insights when your data set requires you to predefine everything you want to analyze?
In short, while the version of ad-hoc segmentation in leading web analytics solutions helps to provide some very keen awareness, your ability to act on this data is limited — “Good to know I have a problem, but what is the true source of the problem?”
Stay tuned for part two of this series, where I will discuss more effective approaches to ad-hoc segmentation. In the meantime, what are your thoughts on website analytics approaches? Have any solutions delivered on their promise of being truly ad-hoc?
Comments