[Sigia-l] ROI/Value of Search Engine Design

[Susan.E.Hensley at WellsFargo.COM]

I thought it was interesting that in all of this thread (or at least all
I've been able to read) no one has pointed out the *reason* why Amazon's
search is probably not going to be as good at finding electronics or apparel
as it is at finding books and DVDs.  

To the best of my knowledge, Amazon's books and dvd/video information
contains extensive Metadata based on the Baker & Taylor service CIP and MARC
records which are supplied to libraries also (this is the original market
for the services of book jobbers like B&T: they supply libraries with books
as well as the records that go into their catalogs).  The records B&T
creates are based on library standards (MARC) for creating extensive
descriptive information about a given known item.  This is why the book
search can work so well - it's based on structured Metadata that is
extensible and standardized, has a demonstrable ROI and a large market. :-)


Amazon search works extremely well for know items for "book-like" items
because "categorization" was done up front: the Metadata creation is current
and intensive.  There is no infrastructure of services, or set of
interntational widely used standards for other kinds of products (that I am
aware of) as there is for books and other formats traditionally processed by
libraries.  I assume the PC site referred to earlier has excellent
descriptive Metadata for its products - and that is why their search works
better for those kinds of known items.

I don't understand the "either/or" dichotomy of categorization vs. search.
My user testing has always revealed that users *want* the painless search
interface of one box they can type a "ballpark" keword and find results that
will point them to what they are looking for. Using categories involves
reading and thinking. ;-)  But of course we should provide the best of both,
providing good pathways to information based on how we know our users think
about what they are doing (which will never reduce itself to a nice
controlled vocabulary).

This also illustrates Karl's point about the usefulness of the "semantic
layer" that Info Science is so good at and how some web sites are
demonstrating the model for that missing presentation layer that is a more
intuitive and a better experience than library catalogs (e.g. why are we
forcing users to think about which "field" their information will be in?)

Just thought I'd remind folks that usually categorization and search are
part of the same strategy.

(I have no affiliation with Amazon, and my info about B&T comes from being
on the receiving end of one of their sales pitches.)

Susan Hensley
Information Architect
Web Research & Design
Internet Services Group
Wells Fargo Corp.
415-396-6403
susan.e.hensley at wellsfargo.com