What’s the difference and why does it matter?
There’s one thing worse than a crime and that’s a crime that serves no purpose even to the perpetrator.
It bears repeating, again, that failing to analyze without segmenting is a crime (Thanks Avinash) and a waste of time serving no purpose.
What’s the average salary of an NHL hockey player?
I would think Sidney Crosby asks more pointed questions.
Is “What’s your site’s Conversion Rate?” a meaningful question or just cocktail party chit-chat? Your bounce rate went up from 70 to 90% – what does that mean? It sounds bad but not if you are running the store, and the visits from elsewhere on the site to the store increased from 2% to 3%
At the eMetrics Summit in Toronto this April, Jim Novo made the point (I’m paraphrasing) that being a good web analytics practitioner requires being a detective.
This contours up images of the magnifying glass through which one views one’s metrics.
Looking closer for Insight
Placing the Conversion Rate under the magnifying glass by segmenting visits and looking at the Conversion Rate for Repeat Visits may be more relevant to you site. Perhaps Geo-location is more important. If you have a winter sports site in Vancouver, your traffic and bounce rate may increase as we get closer to the Olympics but the geographic source of the traffic will probably show visits from locations well outside of your marketing area.
If one makes a change to one’s site and notices a drop in Conversion Rate, is one to conclude the change had a negative effect?
If one were to look at Conversion Rate for only Repeat Visitors or by a Custom Segment (User Defined Value in GA) of Registered Users one may see changes in that focused Conversion Rate that is quite different to what may be an irrelevant site-wide average change.
Some follow-up insights arise. Changes to traffic volume will have some impact on the site’s conversion rate but will not be direct.
That, in turn, will have an even more indirect effect on the conversion rates for Repeat conversion rates and even more indirect for Registered. So, if one were to plan a campaign to increase conversion, would one drive more traffic or aim for more registrations? I guess we’ll be driving for more registrations. What drives more registrations? Now we’re asking the right questions!
Vertical vs. Horizontal Segmentation
Vertical refers to a hierarchy through which visitors are promoted.
On the bottom are the casual visitors. They get promoted as they “achieve” our loftier goals:
- PDF Downloaders; followed by
- Newsletter Subscribers (which requires some greater level of commitment)
- Registered Users
- Quote Requestors
- 1st time Purchasers
- Repeat Purchasers
In this particular example, visitors begining by Requesting a Quote advance immediately. If they subsequently download a PDF they don’t lose their status.
Our previous post explained a limitation of the UDV that would result in only the first of the above conversion showing up against the relevant UDV.
Assuming the first step a visitor takes is signing up for a newsletter, that first value would brand the visit as being one by “Newsletter Subscribers”.
The following visit begins with the UDV cookie having a value and branding the next visit as being one by a “Newsletter Subscriber”, which is correct.
However, requesting a Quote during this 2nd visit would not change the visit’s attributes mid-visit. That change in the __utmv cookie would only be reflected in the reports in the following visit.
That should not be a problem since, over time, one can still view KPI’s by these segments.
If a site’s segments were many and visitors who progressed did so in single or quickly succeeding visits, the UDV’s limitations may disqualify this as a solution.
With such “vertical” segments, there is no need to combine values. Multiple-value variables are appropriate where 2 or more segments are simultaneously relevant to a visitor.
“Lead|Member|Purchaser” does not make sense since each value denotes a segment that includes the previous segment (unless one can be a Purchaser without being a Member). “Newsletter-Subscriber|Purchaser” makes sense since visitors can be either or both.
John Henson of Lunametrics wrote the first post I know of on this concept and coined that phrase.
Using a “supersetvar” is appropriate to satisfy the following requirements:
- Simultaneous Relevance
- Profiles have more than short-term relevance, lasting adequately beyond the visit in which any of the “sub-values” is set.
- Missing the visits in which a “sub-value” is set, should not significantly impact the metrics and trending.
“Newsletter-Subscriber|Purchaser” on sites where few visitors buy more than once, may not provide much actionable data to improve conversion. However the data would be helpful to understand content consumption and whether it is found to be important for pre-sales or post-sales support.
Sites relying on repeat sales would do well to use such “SuperSetVars” to track purchasers and to go even further to identify “1st Visit Purchaser|PurchaserN” where N is an incrementing count or a range, depending on site.
Keep in mind while the __utmv cookie only stores a single value, an analytics implementation can store many values in many cookies. The restriction is only on the single value accepted by GA reports and the way in which visits are “branded”.
One of the best suited requirements for SuperSetVars was a social networking site serving content, games, quizes, competitions and other such sticky enticements to increase pageviews and ad impressions.
Visitor profiles, whether tracked or not, were made up of membership attributes such as age group and gender and by activities such as the content areas (sports, entertainment, quizzes, etc) visitors frequented, the number of groups they joined and the popularity of the groups they created.
Although these were very dynamic attributes that would only be reflected in GA in subsequent visits, the longevity of visitors’ relationships with the site make that limitation insignificant in the trends over time.
While other tools may have made implementation easier, their costs were prohibitive. Making the most of creative solutions became the solution of choice.
The single value was made up of individual, unique mnemonics, each occupying its own position
- Gender: umf: (unknown/not disclosed, male, female)
- Age Group: 8, 9, 0-7 (18, 19, 20 … 26, 27+, based on date of birth. For non-registered visitors the position is left blank)
- Visitor Status: vrI (casual Visitor, Registered, Internal staff and content writers)
- Attachment: nfcph – None, Friend (was invited), Connected (has 5+ friends), Popular (10+ friends), Hub (20+ friends)
- Activities: xiQs referring to eXpanded profile, Invited a guest, took a Quiz, Sent an e-card
- u8rn|__Q_: Segment of Registered visitors of undisclosed gender claiming to 18 years old and who, as at the previous visit, had less than 5 friends and had taken a quiz
- u8rc|x_Q_s: A similar segment but who, as at the previous visit, had between 5 and 9 friends, expanded their profiles and sent an eCard
Yes, there will be many permutations. However, the technique involves in-line, Advanced Segments and/or profile filtering to limit metrics to individual attributes or to certain specific value combinations.
Multiple profiles are not always necessary since segments can be created as trends identify the most valuable visitor segments. AdvSegs can be used to dissect and drill down further.
Content sites such as this, earning revenue from page views and consequently impressions, will gain financially rewarding insight from a better insight into content consumption on their sites.
Hierarchical Super SetVars
Visitor Status and Attachment are hierarchical. Once a member has acquired 5 friends, they were not demoted if 1 friend breaks off the friendship.
One of the important aspects of creating segments, or better still, having your users create their own segments, is that you are likely to learn what you may otherwise have never expected.
There are many ways to swing a cat. If it pays to do it, do it.