Plus signs appearing in keyphrase strings after GA.js update

[danlefree]

Over the course of my efforts to find long-tail keywords, I've run into an interesting issue with the GA.js update for Google Analytics.

The profile in question still has one subdirectory reporting with the old urchin.js script, which may play a part in the issue, but I do not believe so - the subdirectory in question receives virtually no keyword-driven traffic.

When browsing through the Traffic Sources > Keywords reporting, I am seeing entries like these:

keyword+keyword2

. . .

keyword keyword2

I have reviewed past stats to confirm that the plus sign delimiter does appear for periods prior to the GA.js update, however, the entries are nowhere near as frequent (right now, the plus sign-delimited phrases outnumber the space character-delimited phrases).

Any ideas on this one from our resident Google Analytics pros?

[incrediblehelp]

Could this be referrer SPAM?

[danlefree]

It's possible, however, the individuals would have to be running the spam over a browser which is parsing the GA.js script and forwarding the results on to Google Analytics.

I've never seen an implementation like that, as it just seems woefully inefficient (unless it's being run across a FunWebProducts-style botnet).

The site in question is not a blog site, we don't publish referrer statistics anywhere, and the bounce rate for keywords with a plus delimiter is low (8.76%, with an average time on page of 00:06:39) - and Google Analytics reports revenue for the plus sign-delimited keywords.

There is no increase in referral traffic which coincides with the increase in plus sign-delimited keyword entries, however, it does appear as though the plus sign-delimited keywords are cannibalizing the space-delimited keyword volume.

I should add that this could be an issue with the handoff from the GA.js portion of the site (which covers the portion of the site which is marketed) to the shopping cart running on the urchin.js script (perhaps GA.js makes use of the plus sign delimiter in the session cookie and urchin.js is looking for a different character and then choking on it?).

[wige]

I have noticed that Google Analytics has started an effort to be more intuitive in detecting search queries, specifically, finding searches that are done in embedded search (ie someone has used the API code to embed Google, Yahoo, MSN or a third party search into their site) and these searches can show up in the keywords. For example, I have two similar sites, one of which is just a place to test new designs and layouts. Both sites have on-site search, with the search results page on the experimental site pointing the user to the appropriate page on my main site. Analytics records traffic from searches from this non-branded internal search page, and shows the keywords in the keyword assesment area, in most cases with the + delimiter included. This may be what is happening in your case. I have also noticed that keywords entered in the internal search function on the main site itself are sometimes recorded in the keywords listing.

It should be possible to view traffic by search engine, and then check the keyword list for each of the search engines and see if one of the lists has all the results with the delimiter - at least it may give you an idea of the source of the problem.

[danlefree]

Follow-up: Turns out it boils down to the aforementioned issue with the GA.js reporting versus the urchin.js reporting.


Typical urchin.js _utmz cookie value:

... utmctr=keyword+keyword2|utmcmd=organic

Typical GA.js _utmz cookie value:

... utmctr=keyword%2Bkeyword2|utmcmd=organic

urchin.js just passes the raw plus sign delimiter, but GA.js encodes it (thence, when the urchin.js script picks up the cookie, it passes the encoded plus sign and Google happily stores the keyword+keyword2 value as one long keyword).