Enhanced reporting for numerical custom variables
03/11/2010 | Written by | Categories: Analytics Philosophy

Whichever web analytics tool you are using, you can only capture so much information within the page name, standard variables, ecommerce code, etc.  For everything else, you need to use custom variables, whether these are the SiteCatalyst eVars or s.props, Google Analytics custom variables, Sitestat labels or equivalent with other tools.

Usually custom variables are used to store text values – examples include:

  • The site section for the page being viewed
  • The page type of the page being viewed
  • The method used to access product page
  • The filters which have been applied to search results

But sometimes the custom variables are used to capture and record numerical values with reasons for this including:

  • The number of search results returned
  • The number of days between today and a reservation date
  • The number of items in a basket when commencing the checkout process
  • The length of a reservation
  • The number of images viewed

These numerical custom variables can provide really valuable information and insights into your customer behaviour, allowing the analyst to answer questions like:

  • Are our visitors looking to make a booking well in advance or just before a holiday?
  • Does the conversion rate vary depending on how far in advance they are looking to make a booking (serious intent to book vs research)
  • Do you have two or three distinct segments of visitors to your website that you should be treating differently instead of catering to the “average” visitor?

The problem is, it’s not easy to get at these numbers in web analytics tools.  The numbers are treated as text fields, dimensions like any other variable, so reports won’t be that useful.  The information is great but difficult to present in a simple way for managers to understand and for business actions to be taken.  It can be enhanced using any web analytics tool (that allows for segmentation) by extracting the data into Excel and manipulating it there but it would be great if there was an easier way.

The purpose of this blog post is a feature request sent out to any/all web analytics vendors, possibly even to those companies building add-on tools for web analytics based on data extracted via APIs.  The feature I would like to have access to is to be able to identify numerical variables within the web analytics configuration and from that, to be able to treat them like the numbers they are.  This would enable two sets of features for these variables.

The first is for various statistics calculations to be automatically performed and the results to be made available as metrics.  I don’t have a definite list for this but I expect it would include:

  • mean
  • median
  • maximum
  • 1st quartile
  • 3rd quartile

In web analytics reports, it would be the same as choosing Bounce Rate or Conversion Rate as metrics.  How good would it be to report on the average Book to Rent window by traffic source without having to run numerous segments and extract all the data into Excel?

The second feature would be the ability to create groupings, similar to segments, which are applied to the variable.  For example, you might set up groupings for the book to rent custom variable (that is set during searches, not on for completed orders) of

  • 1 day
  • 2 to 7 days
  • 8 to 14 days
  • 15 to 28 days
  • 29+ days

The % of either searches or visits that fall into each grouping would be available as both metrics and dimensions within the web analytics tool.  What could this give you – maybe the knowledge that non-brand organic search terms have 65% of searches at 29+ days with a conversion rate of 1.2% compared to brand paid search terms with 50% of searches for bookings in the next week, converting at 6.8%.  One group is primarily about potential future bookings while the other are the people who are either going to book now or be lost to competitors.

This knowledge enables you to make smarter business decisions which is the point of using web analytics.  It would be presented in an easily accessible and understandable manner for all levels of the organisation, without a reliance on data junkies to crunch numbers and spreadsheets.  I’m not saying it would be easy to create but the people who create web analytics tools appear to be pretty clever folk…

5 responses to “Enhanced reporting for numerical custom variables”

  1. Bob Mitchell says:


    While there isn’t a clean GUI in the product for all of what you ask for at the moment, I know that IBM/Unica NetInsight has this right now.

    I have the opportunity reasonably often to create metrics such as ‘average rating’ based on a ‘rating-value’ recorded in a tag (that would otherwise end-up being a dimension – and it’s the work of minutes.

    1st/3rd quartile would be more challenging, but certainly not beyond the bounds of possibility.

    I’m genuinely surprised that other tools don’t have this. Or do they?


  2. Peter says:

    Hi Bob,

    Thanks for the feedback, I haven’t used NetInsight so I wasn’t aware it could do this. However I do think calculating the mean is the easiest option, getting the median as well (assuming not a normal distribution) and then the groupings would be even more valuable. I have been trying to think through the other tools I am familiar with and I don’t think they can even calculate the mean in most cases.

    SiteCatalyst offers calculated metrics but the variables needs to be metrics first. It might be possible to create a calculated metric for the mean if you have two variables, one for the sum and another for the count but with a single variable. You could always predefine the groupings and capture both the actual number of days and which grouping it belongs to but I would prefer a solution through the configuration (without using segmentation).

    Sitestat ReportBuilder is quite powerful and you could possibly calculate the mean there as well. They are also able to create custom reports so if reports were designed using the dimension as a metric or grouping values, they could create and give access to.

    Google Analytics doesn’t allow for calculated metrics so can’t even calculate the mean. Given the work people are doing with creating products off the API, I might send this as a challenge to a couple of people – since we can do it all by extracting data into Excel, maybe they could build a tool that does it all automatically.

    Let’s see if we get more feedback from the vendors


    • An excellent vendor solution: Next Analytics for Excel is an Excel add-in that can read direct from the Google Analytics API (or any CSV file you export from another source) and can interpret any field as text, numeric or date. It includes a scripted analytics engine so aggregations (like mean and median), segmentation (like grouping values), transformations and comparisons can be done in memory before the data appears in Excel, saving you from spreadsheet formula hell. We give away a large number of innovative Excel ‘dashboard’ workbooks demonstrating the capability of the product that you are free to modify and use.

      The product is designed as a personal productivity tool and is priced for the individual but is frequently deployed in agency environments. Download to working solution in minutes on your personal computer or laptop. No commitment free trial is readily available.

      • Peter O'Neill says:

        @Mike Not quite what I was thinking of but definitely a potential solution. I would still prefer to have the data grouped in the web analytics tool but if this can all be automated within the scripting engine, would still enable users to quickly and easily get to the data they need. Thanks for the suggestion.

Leave a Reply

Your email address will not be published. Required fields are marked *

Looking to improve your digital analytics or conversion rate optimisation?
Contact us today on +44 (0)20 7148 5970

Alternatively, come visit us in London and we’ll buy the coffee!
Get in touch
Leave us a message
Use the form below to leave us a message.

Lindsey Street