The Right Tool for the Right Job

When we discuss analyzing Web site data, we tend to think about doing it with a Web analytics tool. Last time, I described my perfect Web analytics system as allowing me to have fast, reliable access to my core data on the site and to easily and cheaply access and extract the data in different formats. Then I’d be able to analyze the data in different tools if I wanted to.

I’ve always had this analyst toolbox concept in my head. This would be a suite of tools and techniques that can be used and deployed on the right data at the right time. All analytical software have different strengths and weaknesses, so, if possible, it’s good to have more than one tool in your toolbox. You don’t want to try to bang a nail with a screwdriver.

I saw an example of this recently with a global media company client. It wants to be able measure and report on the hundreds of sites it manages across the world. Typically in a situation like this, there are potential tensions between a need for a high-level view of key metrics across all sites and the local need for depth and detail on a small number of sites.

As a rule (and no doubt, I’ll get email from vendors challenging me on this), I find Web analytics systems are better at handling detail on a relatively small number of sites and less good at reporting a small number of metrics across large numbers of sites in a digestible way. The ability to drill from a high-level view to a low-level one in a way that’s found in corporate reporting systems using business intelligence tools is one example of this.

I recommended the client extract summarized data from its Web analytics tool on each site on a regular basis and put them into a separate database. Data could then be reported using a business intelligence tool such as Business Objects or Cognos. This would also enable the client to add more data (such as marketing expenditure or cost data) about the sites into the database and report that alongside the site data.

In this instance, the client needed to be able to aggregate and consolidate data so higher-level trends could be observed and sites benchmarked against each other. At the other end of the spectrum, there may be times you want to delve deep into data and look for patterns or trends that aren’t obvious in regular reports. This is when you’re likely to analyze data at the visitor or customer level rather than on the site level. It may involve use of more advanced analytical or data mining techniques.

This type of analysis tends to be driven by particular issues or problems you want to understand in greater detail. Often, these problems have several factors you must analyze to understand what may be causing them.

For example, a client has a problem with a high bounce rate on a home page. What’s causing this, and is it specific to certain types of traffic? Potential factors I might analyze include the type of referral or campaign the visit originated from, whether this was the first visit, and even the time and day of the visit.

Though it’s possible to filter data in Web analytic systems, the challenge can often be to iterate through various hypotheses quickly and easily enough to get at the nub of the matter. Maybe the problem isn’t dependent on just one factor, but on a combination. Maybe the home-page bounce-rate problem is particularly bad amongst first-time visitors arriving from Google on weekends.

As we look to continually optimize the site and visitor experience, we must delve deeper and deeper into data to understand the nuances and subtleties of visitor behavior. Our Web analytics tools can take us a long way, but from time to time it may be necessary to look at the data in different ways, using different techniques. Next, I’ll take a look at some specialist data analysis tools and techniques and how they can be deployed on Web data.

Till then…

Related reading

tencent_emily-ma_featured-image
kenneth_ning_emarsys_featured-image
bounce-370x229
site search hp
<