Re: Question on time spent visiting

From: John Marshall <jmarshall_at_clicktracks.com>
Date: Fri 04 May 2007 21:16:27 -0500

Sandy,

I may well be one of the people behind the movement towards using Average
Time On Site (ATOS) as a benchmark. At least I'm willing to take the blame.

I'm the founder of a web analytics firm, and I'm also pretty technical. I've
been involved in the design of our tool, and I do a fair amount of public
speaking on web analytics.

Our tool does ROI/ROAS calculation, and handles latent conversions using
cookies, just like all web analytics tools. No surprise there. I noticed
that even though web analytics vendors kept on beating the 'measurable ROI'
drum, a surprising proportion of any given set of marketing campaigns had no
attributed conversions. Almost regardless of the site you'd have the top 10%
of campaigns getting 90% of the attributed conversions. Naive marketers
might elect to kill all but the top 10% of campaigns, but of course if
you've been doing this stuff for a while you know that web analytics cannot
correctly attribute large chunks of conversions. Despite what vendors say,
cookie deletion, shared computers, errors and other realities of life cause
many campaigns to produce zero conversions, even though they actually do
produce some.

One might therefore think the solution is to make conversions more
definitively trackable. Switching from 3rd to 1st party cookies, tweaking
the site, using better heuristics. All good stuff definitely, but my
experience leads me to believe that the harder you try to nail this stuff
down, the more you simply move the problem from a known place to an unknown
place. You take a spotlight and see the problem, pound it with a hammer and
it shatters into hundreds of smaller pieces that disappear into dark
corners. Crude analogy I'm sure, but there's some truth here.

We wanted a way of measuring the effectiveness of campaigns that simply
avoided the problems of ROI attribution. We knew we were going to lose all
kinds of 'accuracy' type stuff, and it would make CXOs nervous, but they
don't understand how inaccurate ROI is anyway, and we felt we'd gain more
than we lose.

After a lot of thinking we went with ATOS. It's ridiculously simple and
therefore reliable. It has many many faults, but they are outweighed by the
simplicity. It relies on the simple human trait of 'I'm interested in this,
so I'm going to spend time here'. Interest might mean intention to buy a
product, or it might mean noodle around looking at content.

Here's what's good about ATOS

1. Simple, hard to mess up
2. Longer times = more interest = greater probability of conversion
3. Simple integer value, easy to compare across many campaigns/keywords
4. Self normalizing (see below)

Here's what's bad about ATOS

1. Simple.
2. Longer times could mean bad site navigation, slow connections, using site
and watching TV at the same time
3. Cannot compare across sites


Number 3 is the real kicker. You cannot compare ATOS across sites. Site A
has a streamlined nav structure and sells a low priced impulse buy item.
Site B has a dreadful nav structure and sells nuclear power plants.
Obviously they will have a very different ATOS.

Despite this problem, the ATOS compared across campaigns within the same
site is a remarkably accurate predictor of conversion rates, and you get
good data within a day or two of a campaign going live. No need to wait for
latent conversion data 60 days later.

ATOS is also self-normalizing. You can argue that site B will have an
implicitly longer ATOS, and you can argue it's long because the nav
structure is poor, but that problem is very likely to be equal for all
campaigns so ATOS still works.

In short, learn to love ATOS despite its flaws.

Now, at last you answer your question: I doubt there's a need to combine
ATOS with any other metric, because when you look at ATOS across all the
campaigns, the higher value is going to be better. Perhaps you can't see
ATOS for the other campaigns? In that case looking at ATOS isn't useful.


John Marshall
CEO
www.clicktracks.com


Received on Fri May 04 2007 - 21:16:27 CDT