These days, half the Web seems to consist of blogs, discussion groups, and other visitor-generated content.
These small-scale sources of visits can really add up and they sometimes can be interesting as well. Furthermore, if we’re sharp, they can be turned to our advantage. But we have to know they’re there, which is the job of analytics.
Examples from just last week:
- On a mom-oriented discussion group, one of the participants had offered a link to one of our client’s product pages that she thought could be a solution to another discussant’s problem. Our WebTrends reports showed dozens of visits suddenly coming from that one link. Once it was noticed, our client had the opportunity to go to the discussion group and make an even better solution suggestion. The moms benefited from more ideas and information and our client’s site benefited from more visits, some nice comments, and an insight about the importance of the mom community. And, by the way, the moms seemed to think it was great that the manufacturer had dropped in. There could be a future marketing campaign there.
- Someone was selling our clients’ used products on CraigsList, saying positive things and providing links back to the brand site. It was very useful to see how real users were “selling” the brand, because the product characteristics they touted were actually somewhat different from the official selling points the brand usually uses. We had a fresh insight into the end user’s perception of the product and what they considered valuable about the brand. CraigsList has often been useful in this way.
It was essential, of course, that we noticed these small-scale traffic sources almost immediately. Otherwise the client wouldn’t have been able to post a response to the discussion while it was still active. And the CraigsList listing would have expired and disappeared, leaving us none the wiser about the interesting brand perceptions.
That’s why we created some reports that shows small-scale referrer information at a glance, that bubble-up these referrers in a focused way that doesn’t take a lot of sifting through.
As a side benefit, these reports help us feel more on top of the fluctuations we see. We sometimes feel kinda helpless or ignorant when we see one or two specific products jump 5% while the other few dozen products stay steady. On sites with hundreds of thousands of visits, we’re sure that a 5% change can hardly be a simple fluke. Now, with these reports available, we’ve gained faith that something systematic is going on.
The world is not random, we just need better analytics!
Here’s a rough idea of the reports we made.
- A custom report on the dimension “referring page (initial per visit)”
- A custom report on the dimension “entry page” .
- And, most importantly, a two-dimension report combining both the above, as follows:
- The primary dimension is the entry page
- The secondary (nested) dimension is the referring page
- The measure is “visit” with intervals turned on
- Filtering is as described below
The two-dimension report is the important one. It gives, for each entry page, a very specific, clickable link to the exact pages where those visits came from. For CraigsList, for example, the link goes to the actual seller page. For a discussion group, the link usually goes to the specific discussion topic.
You’re probably already thinking that a 2D report like the one we describe would be enormous, in terms of database space and time required for the user to go through it. We agree. We wanted something that would be more focused and easy to read.
This is why the filters are important:
- For the entry page dimension we decided to allow only visits that enter on a product page or on certain “advice” or information pages, including FAQ pages. We’ve learned that links to these kinds of pages tend to be a lot more interesting than links to the home page, so we’ve decided to just concentrate on deep internal pages.
- For the referring page dimension we use Exclude Filters for everything we can think of that we don’t care about. We exclude
- search engine referrers
- sister site referrers (if any)
- pay-per-click visits (which are of course set up to be identifiable through special parameters in the landing page URL) (It seems redundant to exclude both pay-per-click traffic and search traffic, but pay-per-click can come from non-search domains.)
- email traffic coming from webmail sites (like mail.yahoo.com)
- traffic that comes from our banners, which are all marked with special parameters in the landing page URL
- site-specific filters – for example, a filter to exclude referrals from dealers’ sites
There could be 20 or 30 of these filters in all. They pare down the report to something that, for our sites at least, is a manageable size.
The result is a report that shows visits broken down by entry page, only for entry pages of interest. For each entry page there’s a list of referrers, but only referrers that are left over after the boring ones have been removed.
Because all the referring page items are, in WebTrends, hot-linked to the exact source page, we only have to do a bit of clicking and reading on the origination page to find the interesting referrers. Usually we don’t go further down than the top five or ten referring pages for a given entry page.
A few tips:
- Measures should be “visits” as well as KPIs for all three reports. And “collect interval data” must be turned on in order to have the time-trend graph available. Go here and here for our all-time favorite way to display KPIs as measures
- Because we have the one-dimension referrer report available, we can use the cross-reference link to jump from the entry for the referrer in the two-dimension report to a trend line for that individual referring page.
- This report is most useful if it’s scanned often. This isn’t a report to look at once a month.