Understanding Funnel Reports
Matt Hopkins posted in Web Analytics on March 26th, 2007
If you're new here, you may want to subscribe to my RSS feed. Thanks for visiting!
A funnel report or scenario funnel is a report that looks closely at a single path that visitors take through your website. This path could focus on a particular website proves such as newsletter signup or a checkout process for an e-commerce website.
A typical graph shown on a funnel report is shown below.

Elements of a funnel report
Funnel reports are defined by identifying each stage within the funnel. This identification is normally based on profiling your website visitors, possibly by the pages they visit but it could be by any online behaviour.
For example in a checkout process you might have 3 stages:
- Entering the checkout
- Entering credit card and personal details
- Confirming the delivery address
It happens that these 3 stages are based on page views within the checkout process.
Another element of the funnel report is the visitor drop off value. This value is shown for each stage within the funnel and shows how many visitors have dropped out of the process at each stage.
The final element of the basic funnel report is the conversion rate metric. This metric is shown for each stage of the funnel except for the first stage and shows the percentage of visitors that have converted or reached the stage from the previous steps.
Common misunderstandings
Funnel reports are not always as they seem. Lets take the example of analysing a web form with multiple fields. You may wish to know where visitors are dropping off of the form.
The funnel report shows the number of visitors who have completed all stages within the funnel. Importantly there is no explicit order in which these stages have to be completed and the visitor may have gone to other pages within the site between stages of the funnel.
So perhaps a funnel is not the best visualisation for this kind of report. If we use the example of the web form then lets use the data visualisation shown in the Venn diagram below.

The above diagram shows all the people who filled out field 1 of the form in red, visitors who filled out field 2 in green, visitors who filled out field 3 in blue, visitors who filled out field 1 + 2 orange, visitors who filled out field 2 + 3 in grey, visitors who filled out field 1 + 2 + 3 in black.
From this diagram you can see every possibility of the 3 form fields being filled out but the funnel report will only show data in red, orange and black. This is assuming that the funnel is ordered by the field numbers.
So why don’t the web analytics tools show this visualisation? My excuse would be the complications involved if there are over 3 stages as the diagram would have to be in 3D to show all possibilities. A 2D version of a Venn diagram is only ever possible when you have a prime number of stages.
My personal feeling on the best type of visualisation for analysing form fields would be a line graph as shown below.

You can see that there are 9 fields and that people drop off the form at field 2 and 4. If you can think of some better visualisations then please post a comment below.
What are the uses of this funnel report information then?
Despite the downsides to funnel reports you can still get a lot of useful information from them and use that information to make accurate business decisions.
As long as you use funnel reports to track exact paths such as website processes where steps cannot be bypassed the visitor drop off and conversion rate metrics are very useful in working out what steps of the process need improving. From this data you can investigate further to see if there are commonalities with the visitor drop offs and and work out how to improve the offending steps.









March 27th, 2007 at 5:28 pm
It should seem that this problem would lead to a Markov matrix. Improving the process would lead to increasing or decreasing the transition probabilities.
These types of problems were initially tackled with graph analysis but I’ve seem analysts try all sorts of oversimplified approaches since then. Part of it may come from their lack of comfort with a graph approach
March 27th, 2007 at 6:56 pm
Adelino,
I agree that over simplifiying the analysis is not a good idea if you are trying to do any sort of detailed analysis such as form usage or perhaps multivariate analysis.
However I think the tools try and strike a balance between being powerful yet providing easy to understand information.
Perhaps future versions of web analytics tools will add more statistical functionality.
Would any vendors out there like to comment?
April 11th, 2007 at 12:57 pm
Matt,
Thanks for a very clear and concise explanation of funnel reports. I have also often found that Venn diagrams, although offering a more complete view of the process, are more difficult for some users to understand; my work with data mining applications has proved that. I am presenting at a DMA workshop on online marketing in London on 19th April, would you mind if I used your diagram above and referred to your site?
Best regards,
LEE
July 19th, 2007 at 11:18 am
Thanks Matt, very clear article on funnels!
Can you recommend any software (except Clicktracks) that tracks those funnel from webserver logs?
Thanks
July 19th, 2007 at 6:06 pm
Hi,
Yes I know that Unica NetTracker does funnels very well and Google Analytics does an attempt at funnels.