A good friend on “teh client side” emailed a reply. I get a lot of email replies these days as the commenting system on Blogger isn’t so nice.

“Actually CB, i would look for sources of traffic that convert, and then advertise there more. Also question why and are there any other sites like that out there that have just never heard of me and advertise there too.

For ecoms, this is low hanging fruit.

But, changing topics slightly, how do you judge if all that bookmarked, direct dial traffic is good? That is my specially urk these days.”

On the first point – I agree. If I was noticing a big conversion lift from a source, such as a forum, I would absolutely invest more time seeding that group. I’d also ask myself, eventually, what else is similar. So, if your product/site tends to resonate with the 4Chan crowd, maybe you’d try the Something Awful forums too as a next step.

Such decisions based on inference and hypothesis testing is what web analytics should really be about. Sadly, that special combination of talent and opportunity to actually test hypothesis is so rare. I think we have such a tendency of breaking down tasks into very, very simplistic, repeatable tasks – maybe a left over from the evolution of businesses over the past 250 years – but I find the act of “inferencing” to be so much harder to break into specific steps. I mean, how would you write an employee manual for ‘inferencing’ anyway? To be sure, I’m not saying that it would be impossible to do so. Merely, I don’t know how I would go teach somebody ‘curiosity’ or ‘hypothesizing’ or ‘questioning conventional wisdom’. How do you teach that? I know how to foster that kind of a social environment where brilliance is valued – I don’t know how I would go about busting down the human spirit into a replicatable equation.

Onto the next point:

Bookmarked traffic!

Well, at least in GA, not all “Direct Traffic” is “Bookmarked”. Frequently, with bookmarks blending into that URL line (at least in Firefox 3), this kind of distinction between somebody typing in a URL directly into the line, and somebody selecting a bookmark from a list, is kind of fuzzy. When it comes right down to it, is that distinction really important?

If somebody is visiting your site from a bookmark, it assumes that their experience was good enough the first time they visited to actually save the link. And now they’re returning. I guess maybe a great question would be “where is direct traffic landing to?” – ie, what is RETURNING traffic findin the most valuable. This, of course, assumes that all direct traffic is bookmarked.

But it isn’t.

You have people who see your URL on print or TV, or maybe some friend told them. With a site like CNN.COM, or CBC.CA, you’ll frequently get people who will just type it in and not have it bookmarked.

I guess, then, I’d make this key distinction:

Traffic that is RETURNING and DIRECT should be analyzed or segmented differently than Traffic that is FIRST TIME and DIRECT.

It’s gonna be dirty because of those cookie deleters out there.

For all intents and purposes, I’d almost treat the Returning and Direct people as being your bookmarkers, since that fuzzy distinction between typing a URL and selecting from a drop down list isn’t as important as the actual act of returning or remembering to return.

Parenthetically: (The difference between a ‘mental bookmark’ and a ‘physical bookmark’ isn’t as important from a bottom line perspective – or rather, I’m not at that point in my life where I’m willing to invest a massive amount of mental or emotional capital into such a fight. Moreover, I don’t believe there’s a method in GA to distinguish between a mental bookmarker and a physical bookmarker anyway. So it’s moot. MOOT!)

I’d treat your first time direct visitors as almost being a proxy for pure offline or Non-Digital Word of Mouth (ND-WOM) attribution, deflated by 20% because of cookie deletion. Ie. “Hay Christopher – there’s some freaky deecky stuff that Iggy’s doing, check out CBC”, and off I go to the URL bar.

If your website is based on out and out conversion – say, the act of actually buying something – I’d pull out Jim Novo’s arguement here and look at the typical product lifecycle of what you’re offering. Is that product really expensive and requires research? If so, barring unique session logging, which very few of web analysts are doing these days – then the distinction between how multiple-visitor folks are behaving on the site and first time visitors is important. Setting up a LifeCycle model and anlayzing the session on session predictors would become a priority for you at that point.

If your website is based on a Sukmanowski Pageview revenue model, then you’d set conversion at say, “5 pageviews per visit”, and you’d be interested in the relative relationship between the First Visit Direct Traffic and the Multiple Visit Direct Traffic. If your website is super high on the First Visit Direct Traffic, and you’re not getting a lot of Multiple Visit Direct Traffic – well, I’d say that you have a problem.

People are evidently not finding a hell of a lot of value from your site, and they’re not returning.

Based on what I’ve seen over the years, much first time direct traffic is targetted at the home page. It’s typically just the pure URL. Is the home page optimized to demonstrate value based on the overall goals of the site? I’d argue that in most organizations – it isn’t. Most home page pages remain a mess, and people tend to vote with their mouses – bookmarking deep links and going there, or, spending minimal time on the home page as they go directly to their destination. Where are Direct Returning Traffic visitors landing? That might be a fairly important question ask.


5 thoughts on “Responses to “Referring Site Metrics”

  1. For direct traffic, you also have to take a couple of things into account: redirects dropping the referrer, and more interestingly, the growing phenomenon of people using Google just to get to a URL they already know (makes life simpler, since possible typos don’t get you to a 404).

    All these can screw the “authentic” direct traffic.

  2. Jacques is right about the google-as-nav function. its very common.

    but hmmm. looking at deeplinking as a clue to what should be on your home page: that is a very simple and brilliant hypothesis.

    I will see what I can do.

  3. Although i wouldn’t use a 5 page-view depth as a conversion point. I was leaning more toward a frequently returning visitor. 5 visits per monthly unique (more than once a week). It depends on your content type. If you’re say, cnn.com, then that is probably a good way to segment your news junkies. But if you’re abc.com, that might work for your news site, but be less relevant for your entertainment section. And then there is the obvious “so what” question. What do you do with the information.

  4. @Jacques:

    You’re absolutely right.

    A) I exhibit the Google as URL behavior from time to time. I wonder aloud whether or not a “Returning Organic Search” would be something worth considering. I frequently see this sort of behavior in our pre-click analyses, with the top ranked searches revolving around brand terms.

    B) Redirects continue to be an ongoing issue. 🙁


    For you, a Sukmanowski model, one that is based on the Pageviews = Revenue model, might not work for you. I think you’re right on with the frequency and recency measures as conversion goals. They’re what probably matters to your organization.

  5. Redirects are bad… I mean they still continue to be a big issue. Brand terms I see all the time, and they get in the way of more relevant terms one wants to optimize for.

Comments are closed.