Cool Custom Report: Campaign latency, pt 1

We touched on the question of the residual effects of campaigns in our past post about Scruffy Campaign Attribution, which details the value and peculiarities of the obscure, deprecated WebTrends visitor history data element “WT.vr.ac.”  

In this post, we’ll describe custom reports for WebTrends Analytics that list your campaigns and, for each campaign, the amount of residual traffic, i.e. the number of visits by campaign responders AFTER the campaign-related visit. 

To report on campaign latency, you need these tools

• A marker in the landing page URL.  For purposes of our discussion, the entry page URL must have “WT.mc_id = <campaignname>” as a query parameter.  We know you can use anything to mark a campaign, but we’re going to stick with WT.mc_id in this post because WT.mc_id is the only one that will get a visit’s campaign information saved in Visitor History.
• Visitor History must be turned on for the profile, specifically Campaign History - because the reports we describe depend on extra parameters that are created by the Visitor History processing. 
• Custom Reporting.  You need this capability in your WebTrends.

There are three campaign latency dimensions that we can work with in WebTrends.  They are based on Visitor History parameters:  WT.vr.rac, WT.vr.ac, and WT.vr.fc.  These parameter names are created internally by WebTrends when it sees WT.mc_id.  WT.mc_id is all you need to supply.  You should never put these WT.vr. parameters into your landing page URLs.  

“Most Recent Campaign”

If the current visit is caused by a campaign, the Most Recent Campaign is the one associated with the current visit.  If the current visit is not caused by a campaign, it’s the most recent campaign used by that visitor, i.e. a previous visit, previous campaign. 

By the way, a campaign stops being the most recent campaign after 90 days (or whatever you have configured).   After 90 days, it’s possible for a visitor to have no “most recent campaign” because of this.

This custom dimension is built on the query parameter “WT.vr.rac“.  You probably don’t have to build it; it’s probably in your out-of-the-box WebTrends as “Most Recent Campaign ID”.

“Any still-active campaign”

This is all campaigns that the visitor responded to in the last 90 days.  A visitor can have more than one “active campaign” in their history.

When used in a report, this dimension is a list of various combinations of campaigns, separated by commas.  The report shows how many visitors during that time period had each combination of campaigns in their history.

Build this custom dimension on the query parameter “WT.vr.ac.“  (We referred to this one in the first paragraph and wrote a previous post on it, calling it “scruffy” – check the previous post to see why.)

“Initial campaign”

This is the first campaign the visitor ever responded to.  This might be the current visit, if the current visit is the visitor’s first-ever campaign. 

Initial campaigns do not expire – they are associated with that visitor forever. 

It’s important to note that, however, that being the first-ever campaign doesn’t mean Campaign ABC was responsible for the first-ever visit, which could have been a search engine visit, a type-in, an email from a friend, anything, including a campaign.  But not necessarily a campaign.

Build this dimension on the query parameter “WT.vr.fc“

 All the above are in the WT Analytics documentation, and the reports they produce are what you’d expect.  The problem with those straightforward reports is that you want to see latency, i.e. later visits.  This isn’t what you get.  All of the campaign history parameters count the current visit if the current visit is campaign related.   You’ll see spikes for “Most Recent Campaign” and all the others whenever you do a campaign because of the current-visit issue, and real latent campaign visits will be masked.

So here’s the trick that produces much more interesting reports.  Get rid of visits that are currently coming through campaigns.  Report on just visits that came to the site by type-ins, bookmarks, searches, friend-to-friend emails (not email campaigns), mentions on other sites, Tweets.  By reporting on the previous campaigns just for those non-campaign sources, you’ll have a far better idea of residual campaign effects.

The custom filter that makes it work

“Current visit is a campaign visit”

Build this custom visit filter based on the entry page query parameter “WT.mc_id=*  (WT.mc_id with any value)

Note that this custom filter is a visit filter.  It is built on Entry Page with a query parameter specified.  Not on the simple “query parameter” foundation (which would be a hit filter).   Being a visit filter is critical – you want to ignore entire visits.

You will be using this filter as an “exclude”.

And the final touch, a long sessionizing threshold

We suggest 120 minutes or longer for sessionizing in order to keep data noise to a minimum. We wrote a post on the length of the sessionizing timeout, and we really recommend that you use a long timeout when you are studying campaign latency.  Why?  Because the second half of an interrupted visit looks like a latent visit for the campaign that caused the first half of the visit.  That’s not what you want to be watching if you are interested in true return visits.  So we suggest keeping the number of interrupted visits to a minimum by using a nice long timeout.

Put it all together in custom reports

Three reports.  One of the above dimensions for each report.  The above filter applied as an exclude.

For all the above, measures should be, at minimum, visits and conversions.  The definition of Conversions, of course, will vary from site to site.  This Outsider post on creating custom conversion measures might be of interest.

POSTSCRIPT

Some cynicism about latency

We’re not totally convinced in the first place that the “most recent campaign” deserves much credit for a later visit.  ”First campaign ever” on the other hand might deserve some credit for subsequent activity, because that first campaign may have introduced the visitor to the site, and that’s a very important role.  ”Most recent campaign” makes sense for some web sites and certain kinds of campaigns, not so much for others.  It’s up to you, dear reader, to decide whether you are in the “some” or the “other” category as described above.  

Unreliable data

Just to remind you of this unhappy fact of analytics life … Campaign latency reporting relies on persistent cookies no matter what analytics tool you’re using.  Therefore, campaign history is gonna under-report actual residual behavior.  The first campaign visit often gets disconnected from later visits.  Because people delete cookies.  They use more than one computer.  And so on.  The more time has gone by since a given campaign visit, the more likely it is that the cookie deletion problem has made the connection between a visitor’s visits gone, gone, gone.    

Not just campaigns

Need we mention – all of the above can be done in WT Analytics for Search-related visits.  WebTrends has visitor history parameters for all search, paid search only, and organic search only.  Do a search in the documentation for one of the Visitor History parameters above and you’ll land in the section that lists all the VH parameters.

 

Tags

campaign latency, campaign success, campaigns, latency, latent, latent conversion, latent conversions, recency, residual, visitor behavior, visitor history

Somewhat Related Posts

Cool custom report: scruffy campaign attribution

...

Cool custom measure column: Number of first-time visits

...

Cool custom report: Daily data within a monthly report period

...