[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Journal usage statistics

You should ask yourself- how much is a usage report worth to you? not
much, it seems. A publisher asks the same questions- How much is this
additional service worth to my customers? How much will it cost to
implement? How much will it cost to maintain and support?

I have made the argument elsewhere that the passive usage statistics
produced by service providers are worse than worthless. They mislead you
into thinking you know something, when really you don't learn anything.
They are sweet but provide no nourishment. OK, so you could learn that you
get less usage during vacation, and you could learn you have a lot of
usage at 3AM, but you should know these things already.

A while back, a customer asked about statistics and I wanted to respond
"We'll help you do this the right way, but if you want us to do it the
wrong way, we'll charge extra."  That sounded a bit arrogant, so I didn't
say it (it was my 10 years at Bell Labs trying to talk), but I started
writing up how I think the information community should pursue the
statistics problem.

The problem is that there are no good measuring sticks. What is a unit of
usage? It's easy to count hits- but a single page view might generate 1
hit or it may generate 50 hits. You could count clicks, but which is more
usage- a user who clicks 50 times in a fruitless search or a user who
clicks once, gets to the article they want, and they're done. Accurate
measurements require measurement protocols, which start with a
determination of what question is to be answered.

The answer to this problem is almost here.

The next generation of internet based information services has started to
arrive, and customization is pervasive. The people paying the money
(libraries) need to be putting their "brands" on things so that the people
who support them will see where their information is coming from. To help
libraries do this, information services are adding the ability to attach a
library logo, to add links to the library's holdings, to add OpenURL
links, even to use the library's custom web design.

Once you have access to these customization tools, you can really start to
use them to CENTRALIZE your accumulation of usage statistics. For example,
if a service lets you embed a graphic on a web page, you can use that
embedded graphic to track usage by using a coded graphic. I call these
"measuring stick gifs" because you can make controlled measurements with
them (unlike reports you get from vendors). For example, you can have
firstsearch display the graphic
"http://stats.university.edu/logo.gif?firstsearch"  Then every time a user
loads this page, you get a hit on the stats.university.edu server, and the
logs will show the user IP address, the referring page, basically more
than you really should be knowing. It turns out that there a lot of really
good, cheap or free programs that you can use to analyze server log files.
We use one called "Analog".

Looking further into the future, you will see services with the ability to
add OpenURL metadata onto these tracking gifs. I know this will happen
because we have implemented it, and features get copied in competitive

Looking even further into the future, I imagine the development of
intelligent logging and awareness systems- ones that spot problems long
before users get the chance to complain about them, ones that adaptively
help users find resources relevant to their current tasks, ones that
anticipate what the chairman of the physics department will ask for

So what I'm sketching is a strategy for centralizing the accumulation and
analysis of usage statistics on library web servers. You can eliminate
vender-to-vendor variability, and decide for yourself how to summarize


>Sept. 16, 2002
>Currently, I am negotiating with a commercial publisher to include usage
>statistics in the Library's online only journal subscription at no
>additional cost.  The publisher wants to charge an additional percentage
>more on top of the published subscription price to obtain usage
>statistics. In comparison with other commercial publishers, I have not
>encountered this situation since they include online user statistics with
>their subscriptions.
>I can not justify paying for this service.  Has anyone had a similar
>experience?  All replies will be kept confidential.
>Thank you,
>Elizabeth Lorbeer, EdM, MLS