Library Reference Transaction Patterns: Visualizing Library Usage Data


Table of Contents


Libraries reference staff provide invaluable assistance to patrons, however, trends in reference services often go untracked. Analysis of trends in reference staff's transactions with patrons can reveal patterns that help libraries make staffing decisions and better understand their users' needs, as well as understand the changing face of reference transactions over time. In Spring 2010, an analysis of patron transactions with reference staff at NCSU's D.H. Hill Library was undertaken. The analysis considered the medium of transactions (in person, phone, email, direct reference or chat), the type of transaction (reference, computing, directional, print-related), as well as overall trends in transactions over the years, the semester, and the course of the day. The analysis also compared transactions over the course of the day to other measures of library usage in order to gain a better understanding of the holistic use of the Library.

Back to top

Research Questions

  • What longitudinal trends do we see in library reference? What trends do we see over the course of semesters and days?
  • What trends do we see specifically in library chat reference?
  • What patterns exist within different types or different media of reference?
  • Do trends in library reference correspond in any way to trends in other library usage patterns, such as gate counts or computer usage?

Back to top


Data for email transactions and direct reference date from 2001 and the extent of granularity extends only to monthly totals. Data for chat transactions dates from 2001 as well, but spans several different chat systems and formats. Level of granularity for chat transaction data extends only to monthly totals for chat references from 2001-2006. In 2006, the current chat tool, Library H3lp, was implemented. From 2006 to the present, chat data granularity extends to date, time, reference queue, and full-text. Dating from 2009, chat data also includes referring URL. Only sampling of data exists for in-person and phone reference prior to 2006. From June 2006 to the present, in person and phone transactions are tracked by day of the year, hour of the day, medium of transaction, and content-type. Data can only be examined in great granularity, therefore, for chat transactions after 2007 and for in person and phone reference post-2006.

'Academic years' are interpreted as 12-month fiscal cycles, spanning July-June. Where a chart lists the academic year as '2006-2007,' this signifies 'July 2006-June 2007.' It was impossible to break down total transaction counts below the month level, as there are no statistics for email transactions other than a monthly total. Therefore it was not possible to interpret academic years or semester based on actual academic calendar lines, which fall mid-month. At points in the report, data is analyzed on the actual semester dates. This is made clear in footnotes when it occurs, and is done only where a limited set of data was analyzed that had granular date information.

Chat transaction data discounts all chats that lasted 10 seconds or less. This was done in an attempt to avoid inflating chat numbers by double counting transactions in which one of the two chat parties closed the chat window at the end of conversation before the other chat partner had made a final statement, such as "bye" or "thanks." In such circumstances, the final statement appears as a new chat, when it is in reality part of a longer previous chat. Chat transaction data also discounts chats that lasted longer than an hour. It was assumed that such transactions represented chat windows that had been accidentally left open.

For charts that depict chat volume by semester week, semester week boundaries are calculated from Monday to Sunday, beginning with the Monday of the week in which classes began and ending with the Sunday of the week exams ended.


Back to top


Chat queue
The Library H3lp chat client that records chat transactions is able to identify the Web pages on which a user is currently located when they initiate a chat session. A 'queue' represents a certain set of pages from which a user initiates a chat. While the overwhelming majority of Web pages are funneled into a single queue for general chats (ref-desk), several important queues include: chatters initiating from Library Tools pages (library-tools queue) and chatters originating from the NCSU Libraries catalog (ncsu-catalog queue).
Refers to the subject matter of the transaction. Content-type is only tracked for in person and phone transactions. The four content categories tracked are defined below.
A request for information, assistance or instruction on computers and computer software.
A question involving directions to a physical location (such as the restrooms, the Digital Media Lab, etc.)
A question involving instructions or location assistance on printing and making copies.
A question related to seeking information for research purposes.
Direct reference
Transactions that occur when a patron contacts a reference staff member directly, such as in a direct email, direct phone call, or approaches the staff member directly at a time when they are not at the reference desk. Staff are responsible for tracking these transactions themselves, and reporting monthly totals at the end of each semester.
Gate count
As of January 2010, a new gate counter tracks the number of people entering the library each half hour of every day. Gate count data used in this analysis refers to averaged data available at the time the report was written, from January 2010-March 2010.
Refers to the medium with which a patron performed a transaction. The four media categories are: in person, phone, email, and chat.

Back to top

Last updated: July 7, 2010