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

[Mike.Steckel at SEMATECH.Org]

I am sceptical of this on my first read. How is it determined that Albert is a casual searcher in the system? How is he "assigned" as a casual searcher? Also, following the berrypicking model (http://www.gseis.ucla.edu/faculty/bates/berrypicking.html) there is a good chance that Albert would increasingly move toward being an Interested Layperson during the same search session, especially as he narrows the search to something more specific, and then possibly back to Casual Searcher again as he talks to the person requesting a book. All during the same search. I just don't see the real world falling into these neatly defined categories. I also, don't see people being able to effectively assign themselves to the categories.

-----Original Message-----
From: Scott McDaniel [mailto:scottmcd at cognetics.com]
Sent: Tuesday, February 18, 2003 4:08 PM
To: sigia-l at asis.org
Subject: RE: [Sigia-l] ROI/Value of Search Engine Design - Resources?


At 10:56 AM 2/18/2003 -0500, Ken Bryson wrote:

At this point I'm surprise nobody has brought up the issue (albeit
librarian-centric) of  known item vs. subject searching. That is, depending
on whether you know exactly what you're looking for (Linksys Wifi) or just
the general category/subject of items (wireless routers), you will likely
take a different route to finding your information.

Debating the usage of category browsing vs. searching based primarily on how
people use ecommerce or online news sites, seems to limit the debate a bit
too much.

I don't have an MLS, but my wife does.  We recently co-authored an article that appeared in the second volume of UPA's User Experience magazine.  The complete reference is:

McDaniel, S. and McDaniel, M. (2002) The Big Dig: Mining Nuggets of Value. User Experience, 1:2, 20-29.

In the article we note that the current heuristics about how to produce a good search interface were so general as to be next to useless.  Since the answer to specific questions always seems to be "it depends," we proposed three dimensions upon which "it" could depend.  Most of the heuristics came from current literature (NNG's report on search, Jared's book, LIS textbooks, Tog on Interface, etc).  Some are proposed heuristics without research support.  If you'd like the complete set of heuristics, see the article if you have access to the magazine or contact me off list.

The basic idea is that you determine where your users fall on each of these three dimensions:

Type of User (How familiar is the user with the information being searched?)
  -- Casual Searcher (have a passing interest in the material.  Most e-commerce shoppers.)
  -- Interested Layperson (understands basic jargon and wants to learn more)
  -- Subject Matter Expert (thorough understanding of the material)

Search Experience (How much experience or training has the user had in searching?)
  -- Novice Searcher (typically enter a search word and press enter)
  -- Intermediate Searcher (understand basic Boolean searching)
  -- Advanced Searcher (are trained search professionals)

Goal of the Search (What type of information is the object of the search?)
  -- Precision Search (locates a specific, known item)
  -- Recall Search (locates all items that meet given criteria)
  -- Some Good Items (locates a small set of items that best meet given criteria)

Each value on these dimensions has a small set of associated heuristics.  By determining where your user falls on each scale, you derive a full set of heuristics for the search interface.  I'm including a brief excerpt from the article below.

I've seen a couple of Jared's presentations (most recently the Scent of Information, given to the UPA DC Metro chapter).  My impression is that most of the people he tests would be:

  Casual Searchers or Interested Laypersons
  Novice Searchers
  Precision or Some Good Items

I'd hesitate to generalize findings for these people to all Search situations.  (Jared -- Are my impressions above correct?)

    Scott McD.


Here's the specific excerpt I mentioned earlier:

Let's take the case of Albert the Bookstore Associate to illustrate how this works.

Albert is receiving many requests for books about Alzheimer's.  Because Albert is not an expert on Alzheimer's or particularly motivated beyond a customer's request, he is best classified as an Casual Searcher, and because he, like most bookstore employees, have limited search experience, he is best classified as a Novice searcher.  Since his customers don't have specific books in mind, Albert must perform a "some good items" search to recommend a few choices.  Combining these results, leads to the following set of guidelines:

§       Offer a simple search box on the home page and on each page throughout the site.
§       Present information in the search results that allows users to assess relevance.
§       Allow Casual Searchers in an e-commerce setting to search for things other than products. 
§       Allow Casual Searchers to search for items with their own vocabulary.
§       Consider alternatives to a  search.
§       Global search is better than a scoped search.
§       Lead Novice searchers through the search dialogue.
§       Present the search results clearly and simply.
§       Make it easy to evaluate individual result items. 
§       Allow users to save and resume Recall searches. (Also applies to "some good items")
§       Make it easy to expand searches as well as narrow them.
§       Search globally. 
§       Accept word variants, word-stemming, and synonyms for search terms.
§       Make a simplified reference interview an optional part of the search process. 
§       Support the task of creating a short list of results.

Inspecting the guidelines reveals some redundancies, such as "Global search is better than scoped search" and "Search globally."  These are easily consolidated.  But what happens if the guidelines conflict?  Short of flipping a coin, the best way to make a decision is to look at data from your user analysis.  Give more weight to the needs of your primary users.  You might also consider whether you in fact need two different user interfaces (text entry search vs. a category browse), or even two different products.





~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Scott McDaniel
Cognetics Corporation, Designer


1320 Fenwick Lane, Suite 209, Silver Spring, MD 20910
301.587.7549   --  scottmcd at cognetics.com  --  fax 301.562.8267
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~