depends greatly (actually HUGELY) on the site...

On 16:52:50 23 February 2005 mekanoid wrote:

What percentage of visits to your sites are "directly referred URL’s"?

In any analysis i’ve done this figure seems to be 50% and over. Do you accept that these are genuine visits that did not originate from either a search engine, email, external site or are a significant amount of these visits classified in this way simply because the referrer value was lost from their HTTP header.

Also, I commonly see these visits classed as bookmarks yet it is probably a small minority of these visits that could be correctly classed as Bookmarks.

Has anyone seen any research on this matter?

Hi Scott

We haven't done any (quantitatively meaningful) research on this. However, I know what you are talking about and from those we've talked to (and for this site) it would seem that the majority of these referrals are from e-mails sent out by the site owner.

It depends, of course, on how many e-mails you send and there will be spikes when e-mails are sent out, but certainly e-mail tends to account for a higher proportion of these visits than, say, bookmaks or directly typed in URLs.

As Jon says, this varies a lot, though. For example, by customer segment. It's unlikely you'll be getting an e-mail to prompt you to visit a site that you've never visited before (as most e-mails now are to opt-in list members i.e. customers or those who have visited your site and opted in).

In the past we've measured bookmarks by looking for requests for the favicon.ico  file (I believe this is IE specific) rather than looking at the direct referrers you mention.

I'm not sure how often it is the case that the referrer value is lost from the HTTP header but I'm suprised at the 50%+ figure you quote. For us it is nothing like that high (<10%). What kinds of sites is your sample based on?

Ashley 

Hi Matthew,

Scott from HBX here. I agree that this metric is not too reliable, it's near useless and out of date.

What I am trying to do is improve this report so true directly referred is far better represented. Why is the presence of a "HTTP referer" tag in the HTTP header the only way to define this report? I have an idea about how we could  take this forward.

From the list you put together of reasons why the header information might not be available it is fairly clear to me that the the majority of incidents are related to third party services modifying the header or it not being present at all. (r.g. email)

Currently these visits are included in an "unknown" or "directly referred" pot yet these very same visits are registering responses to campaigns in another area of the analytics tool.

If a visit to a site cannot be linked to a referrering URL yet the same visit has registered a response to an email campaign for example, why not take the visit out of the unknown pot and classify this as an Email referred visit. The same theory would work for any campaign where no referrer is available and a response is registered.

We would end up with a list of referrers grouped by Organic & Paid search, email, banner, keyword network etc etc, followed by a more accurate directly referred count and then the list of individual URL's sending traffic to the site.

Or do we really just want a list of URL's forwarding traffic to our sites?

Scott

I don't really follow what you’re saying.

As an example, we have a developed a micro site that has a very large advertising budget and a major email campaign. The purpose of the site is to get users to register and leave their details. Post campaign, we are now trying to see which campaigns were the most effective. I have one report from the media agency saying that it produced 60% of the visits to the site. When I look at the referrer report 80% show up as 'no referrer'. Simplistically, I'm trying to answer the question 'where did my visitors come from?’ To gauge effectiveness I'd like to have a table of registration rates based on advertising source. If this isn't possible with referrer, then what's the best way of doing this?

So if referrer data is meaningless, what report should you look at to get an overall feel for the current traffic drivers to a site?

Sorry if I didn't make myself clear.. it was late last night!!

In reference to your example, 80% of your traffic came from... "unknown", you would have to make assumtions about where they came from. In the case of HBX, the analytics tool I work with we would set all the advertising sources up as campaigns (based on a unique source= query string value being present in the URL) and measure response rates through to registration "conversions".

The referrer report would not be the most accurate way to measure campaign effectiveness for all the reasons Matthew Tod described previously however using the referrer is an option we have available when designing campaigns in HBX.

Considering all of this, we would still be left with a referrer count of 80% unknown. What I was suggesting was that our campaign tracking environment knows that most of these visits actually originated via a campaign response so why not filter down the "unknown" or "directly referred" to a level of true "directly referred" visitors rather than "we didn't see a referrer in the header".

Scott

On 10:43:57 1 March 2005 hobart65 wrote:
>Scott
>
>I don't really follow what you’re saying.
>
>As an example, we have a developed a micro site that has a
>very large advertising budget and a major email campaign.
>The purpose of the site is to get users to register and
>leave their details. Post campaign, we are now trying to
>see which campaigns were the most effective. I have one
>report from the media agency saying that it produced 60%
>of the visits to the site. When I look at the referrer
>report 80% show up as 'no referrer'. Simplistically, I'm
>trying to answer the question 'where did my visitors come
>from?’ To gauge effectiveness I'd like to have a
>table of registration rates based on advertising source.
>If this isn't possible with referrer, then what's the best
>way of doing this?
>
>So if referrer data is meaningless, what report should you
>look at to get an overall feel for the current traffic
>drivers to a site?