on target

Web Analytics: Grain Matters!

It is funny to discover that some web analytic applications, despite their name, are not "analytic enough".

Time ago a friend asked for some hints on his landing page. He told me that he had read many landing page optimization papers and was also tracking his site through the Web Analytic application I suggested.

That application reported an average of about 22 seconds on landing page: too little, in his opinion, to read and understand the page. He thought that increasing the time on page would improve conversion.

I told him that he could not base any conclusion on an “average” and I asked him a set individual observation. He replied he had no idea on how to get such data and gave me the password to access his analytic application. After few attempts I realized I was also unable to get data at desired detail level.

My friend seemed a little disappointed with the application I suggested, so I reminded him his main constraint: “simple and cheap”. My suggestion was not only an easy easy and free solution, but also the most widely currently used!

He asked me: “And you use it?”. My honest reply was: “Well ... yes and no: we definitely look at data from our tracking log”. Feeling a little guilty for my suggestion, I adapted tracking code to his page and we got some measurements. The distribution of results (in seconds) is shown below:

> summary(timeOnPage)
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
  1.545   2.715   3.053  22.220   3.443 348.600

A very strange distribution! Can you figure what's happening?

We are looking at two very distinct groups of visitors: those that immediately decide that the page is not interesting and bounce away (A) and those that read the page (B).

Group A is (and will probably continue being) the largest group: curious, competitors and people driven to the page by his ads that immediately realize it is not what they figured:

> summary(time_A)
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
  1.545   2.688   3.007   2.994   3.336   4.472


Group B, on the other side, stay on page about 4 minutes (241 / 60), what is very reasonable.

> summary(time_B)
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
   77.8   206.2   237.9   241.6   279.8   348.6 

The strategy, consequently can possibly be to improve visitors filtering and not time on page.

“So what 22 seconds reported by the application mean?” asked my friend. “Honestly, just nothing!” I replied: in fact it is too high for first group and too low for the second.

My friend added: “There is something wrong: I don't understand how the supplier could produce such a poor application, and how it became so popular”.

The fact is the application is absolutely not poor, and be sure data exists, but it is not made available to user. Users only see aggregates, possibly for the same reason they don't get IP number (“privacy”  as explained by supplier).

The lesson is simple: if you really need analytics (web or any other kind) you shall get access to fine grain facts. Like we have seen in previous post, in many situations averages are not enough.

Optimizing landing page is a common task and many similar cases exist. Certain analytic application are perfect if you only need a “trend dashboard”  but, if you need to understand what really happens be sure to also dispose of fine grain data.

Interested in partnership?

Would like to try or coach this technique with your clients, do you need any additional or technical detail? Please let me know! Component Bases Solutions has great interest in partnership with consultants. We can help you automate your proposed solutions in a very short time. We can also help to increase your visibility through links from many management tools we make freely available on the Web. Please contact for more detail.

Did you like this post?

Please, spend a few seconds and click on button below to socially bookmark it or tweet it. It will help to promote this blog and improve its content, Thank you!

Posted at 12:00AM Mar 30, 2010 by admin in IT  |