> Time-based metrics have utility in context of user needs, site objectives,
> and goals. As a stand-alone metric for comparing overall performance or
> engagement across different sites that fulfill different informational,
> navigational, or transactional needs, I think it's misleading and easily
> fooled. On many sites rich media or AJAX functionality are only small
> components of a site's overall web experience.
>
> More importantly, I also don't immediately see how "total time online"
> *directly* ties to revenue, like the page view.

Judah, man, you know I love ya and think you're a very smart guy, and I
agree with what you are saying - at the site level. But I don't think
that's the issue.

Let me suggest we all push the reset button on this topic for a second and
look at it from the perspective of the Nielsen / comScore / agency complex.
From the point of view of *their* business model, just like you would look
at the analytics for your own site based on your own model.

Kick this discussion up to a macro perspective.

Right at the top, let's be clear that we are talking about a change in the
way web sites are **ranked*, correct? This has nothing to do with actual
execution, it is simply a ranking system. All they are saying is it makes
sense to them - and does anybody think this was done without supporting data
of some kind? - to rank web sites by duration of visit as opposed to page
views. That's it. They're not saying anything about how to analyze a site,
how you should analyze your site, etc.

They are changing a ranking system, and they have decided it makes more
sense to rank sites by "quality" as opposed to "quantity", by "engagement
with the media" as opposed to "velocity of the media". On the face of it,
if I am a buyer of display media, this makes sense, because I want to buy
"weight", not velocity. I can't get weight from a 3 second visit, or if 50%
of the visits are 1 page visits. Think about it. Frequency or velocity
misrepresents the value of the media.

So let's drill into what this idea looks like in the real world. If I am
buying media for a global packaged goods company, I'm not thinking about
sites except for a very few instances. I'm buying tonnage across networks
online and offline - TV, radio, magazine, web.

I need a thesis of some kind when I'm doing this to have any rigor at all -
I want to reach a certain type of audience, with a certain level of
repetition, etc. I reach certain people with certain repetition and I
achieve "weight" that translates into awareness / trial that translates into
sales.

Whether you believe in that model or not, that's the way it is done, because
they can only work with the data they have. They don't have granular
information, and they don't really need it, because their own models show
that **it works**, over and over. You buy X audience weight, you get Y
results. You can confirm that with market tests and scanner data. The
model works.

So you're this buyer, and you are looking at the web. You know there is
audience on the web you want to reach. But the current media buying model
doesn't work, your media mix models tell you that when you buy weight on the
web *as it is currently defined** nothing happens - there are no incremental
sales. Awareness, OK maybe, (if there is a way to prove that isn't really
coming from the TV), but no sales.

So the display ad buying model that works everywhere else in electronic
media is broken for the web. You start to think about it, talk to people
about it, and you think, hey wait a minute, in every other electronic media,
we look at duration. Duration is a - admittedly weak - proxy for
engagement, and we all know it's not "accurate", but it is precise. Precise
because it is always wrong in exactly the same way.

Then you start thinking about the behavior, visits to web sites. And it
strikes you that a lot of this activity is simply not really conducive to
the idea of "weight", because there is no substance to it. It's a lot of
thrashing around and so forth, and in many cases, people don't even see the
ads because they are not really "engaged with the page", if you know what I
mean.

And that's probably why we can't get any weight on the web, because the
whole Frequency side of the equation is screwed up. Frequency on the web is
not the same as Frequency on TV or Radio - it's different. The Web is
different, as the chant always goes, right? So let's admit that?

Frequency on the web is busted, it's not the same as Frequency on TV or
Radio. So we have to compensate for that, we have to weed out the thrash
that is low value. One way to do that is to use duration as a proxy for
Frequency, a different way to measure engagement.

We know people are generally "engaged" listening to radio or watching TV at
some level. Sure, there is a ton of noise in that, but like I said, it is
always wrong in the same way. It's consistent. The average percentage of
people going to get a beer during the TV ads is always the same. The fact
they do this doesn't really matter, only the consistency of outcome for the
media buyer and his or her boss matters.

I buy X weight, I get Y sales.

This idea of using duration as a "screen" is fundamentally no different than
excluding one page visits in key metrics, or excluding robots and spiders,
or any of the other things we do in web site analysis to get cleaner data
out of a really dirty dataset. It's a filter.

Now, you start to think about the future, and you think about mobile, and
in-game, and all the social networks that think they are going survive off
display ad revenue, and whatever else is coming down the line. And you
think to yourself, look, the more interactive the environment, the more this
matters, because there can be all kinds of thrashing around in these other
media too.

All kinds of opportunity for an "impression" that isn't really an
impression, because the visit is too quick, or it is too focused on a task.
These quick hits are like 1 page visits, they lack weight, they shouldn't be
counted.

But now back to the reality of executing. There is no way I can analyze all
of this, I can't possibly manage a network on a global basis that is looking
at all these reasons a web site analyst can come up with why I'm not
measuring things correctly. I need a damn standard, something I can use
globally at the macro level across all these platforms that gives me some
sense of "weight" that I don't get with Frequency in an interactive medium.
And the easiest and most precise - though incredibly inaccurate - standard I
can come up with to do that is duration.

It is universal, it is stable.

And I need the most stable thing I can get, because I need to plug it into
models and try to figure out how to buy weight, how to impact sales.

I need a least common denominator representation of engagement, to tell me
that there is at least a chance in hell visitors saw my ad. And I can use
this LCD to create a ranking system that - on a macro basis, with a lot of
noise, but consistent error - tells we where I am most likely to find
"weight".

Does the above make any sense at all to you folks?

Jim Novo
jim@...
Web Site: http://www.jimnovo.com
Blog: http://blog.jimnovo.com/