New Uzilla Component: Obzerver

The Obzerver module provides an open source implementation of a click stream monitor.  It writes log files to the local hard drive in Mozilla Firefox and the Mozilla suite.

The stored data is in the FLUD format, proposed by NIST in 2002:
loghead 2004/6/15-8:55:39 subject=100 version=v1 time_zone=-5 generator=obzerver
event 2004/7/15-8:55:40.401 1:0 url=http://www.mozilla.org/products/firefox/start/
#OP newpage complete * url=http://www.mozilla.org/products/firefox/start/
logend 2004/6/15-8:56:21

This FLUD sample shows only the page level data.  Low level event data is also partially complete but not yet integrated with logging.

The FLUD format calls for identifying the sources of navigation elements and in the web scenario might allow us to quantify the relative proportions of methods for accessing new content.  C.Soules (a CMU CompSci grad student) crafted the Obzerver implementation and is interested in the source events for a different reason; for the meta data they provide. 

Obzerver is ready for use today for simple url logging and with a bit more effort for more complete event traces.  The Uzilla.net suite provides these rich event traces, a way to model your test protocol, and data analysis. 

If you're interested in helping with the project, check the out the bug list.  Commenting on bugs and filing new ones is a great way to start getting involved. For a meta-view on how this works, check the working paper Usability Discussions in Open Source.
|||
Posted at 8:58

A Better Multi-Select Widget: Open Source DHTML

In a recent usability study, we observed that the intranet user base was unfamiliar with the convention of control-click for selecting multiple options in an HTML menulist widget.

To address this issue, we used checkboxes and DHTML to eliminates the need to  control-click to select multiple items. In addition to creating an affordance multiple selection, this system also provides better system status information.  

An example is shown below in which the user has already selected two options from the first two collections and is working on the third: The design saves clicks and screen real esate by making the checkbox containers collapsible.



The code is available as open source using the business friendly MPL license. More information, a live demo, and the source.
|||
Posted at 16:33

Form Design: Chunking for Efficiency

 
In a recent evaluation with a moderately large sample of 30, the following form was tested.


We observed three different orders of execution in the data.  The rather jumbled nature of the form layout contributed to this problem.  A bigger issue was the lack of grouping of related fields, taking significant additional cognitive processing.

HTML has fieldset and legend tags to solve this problem.  In this case, re-ordering and changing the layout of the fields was required.  Here's an example of a fieldset:

Uzilla Services

Test Design & Logging
Screenshare
Realtime Monitor

We project a more regular pattern, shown at left, in the revised design. More importantly, we expect that the percentage of users skipping a field will decrease dramatically.  Thirty percent of users missed a field and lost time upwards of a minute in some cases.

There is ROI on layout and visual design, in case there was any doubt.


|||
Posted at 21:25, Published in: Prototyping Testing

Analyzing Usability Quantitatively

Jeff Sauro's site measuring usability .com is a great introduction into the issues and rewards of quantitatively measuring usability.  Articles expound upon the intracies of task timing, task order randomization, and a fresh approach to thinking about sample size.  Based upon a initial reading, these articles all express themes in line with the design rationale of Uzilla.net. For instance, Uzilla.net allows partialing out page load time, supports task order randomization, and of course, by easing the preparation and analysis phases, enables a greater number of participants.
|||
Posted at 10:34, Published in: Testing

Open Source Credits and Related Resources

Two new pages, works in progress, offer credits to open source code and related public resources utilized in Uzilla as well as code contributed back to open source.  The related tools page points to other resources for web developers and user experience professionals.
|||
Posted at 21:17

Macintosh OS X for Usability Professionals?

OS X is very popular among internet developers and designers, with the availability of Macromedia and Adobe tools.  Is OS X a reasonable choice for a usability professional?

Uzilla, LLC's tools, the browser-based usability testing suite, uzReview, and uzCardSort all work on OS X.  UsersFirst has recently introduced a video capture system that requires a mac laptop.  MacSHAPA, a longstanding video analysis tool, is coming to OS X.

Hmm, a g5 or a tiBook?


|||
Posted at 13:15

Quantifying Lost in Hyperspace

Smith (1996) proposed a measure of "lostness" in hypertext structures that centers around page revisits.  The lostness measure shown below, from a data table from the Uzilla.net reporting application , depicts the most lost users in a recent test. 

Sorted by lostness ascending, the pictured users exhibit high rates of page revisitation which Smith argues indicates the users are lost.  Efficiency on the other hand takes into account redundant page views, but also deals with the total number of page views versus the minimal (or "optimal") number.  The two measures are correlated but not perfectly, as users with the same lostness may vary in efficiency.

Lower values of lostness indicate more confusion and lower values of efficiency are better.
 


We're not totally convinced this measure captures negative user experiences in all cases.  Certainly, in a shopping scenario a user might revisit a product page repeatedly while comparing it to other options -- and that's a desirable business outcome. In other applications, revisitation may be a clear indicator of confusion. 

We'll report on correlations between this measure of confusion and other measures we've been cultivating in the coming months.

Smith, P.A.: Towards a practical measure of hypertext usability. Interacting with Computers 8 (4) (1996) pp. 365-381 http://citeseer.nj.nec.com/context/898955/0
|||
Posted at 0:2, Published in: Web Usability

Variations on Remote Usability Testing

The latest IBM Systems Journal features a nice writeup titled "Methodology for remote usability activities: A case study".  F.S.H. Krauss classifies remote usability activity into four categories:

1. real-time design walk-throughs and usability tests
2. surveys
3. automated usage tracking
4. handling of user-reported critical incidents

Number 4, user reported critical incidents, is the least explored in commercial settings but was one of the earliest uses of the phrase "remote usability testing" (Hartson, 1998).

The IBM article re-affirms the cost savings of remote testing and expresses value in moderated remote tests and usability focused surveys. Issues of firewalls and participant bandwidth are paramount, but the general conclusion is that the approach is of value.

Uzilla has additional tools for real time remote moderation in development.  Look for an announcement in the coming months.
|||
Posted at 10:25

Card Sorting Developments

Uzilla, LLC's open source card sorting tool is now documented, and supports an export of similarity matrix for analysis in external applications.

We've had some interesting conversations recently and a spec for hierarchical sorts is underway using a zooming metaphor.  Lou Rosenfield asked if it supports open and closed sorts -- of course!  For clarification, this means you can pre-specify the target categories or allow the user to make them up.  Heck, you can even do a mix of both.
|||
Posted at 15:52, Published in: CardSort

Advanced Search, well done.

NPR's advanced search for the audio archives is very well done:



The use of an inline descriptor for the keyword box serves to conserve valuable real estate, easing the impression of effort associated with completing the task.  The first drop down offers a concise set of time periods, growing in range as they get further from the current time.  The key design feature is an understanding of the user's task, resulting in filters which reflect previous experience. 

A little birdie told me that NPR's site is soon to undergo it's first major redesign.  Keep an eye out, it's sure to make a great resource even more useful.

|||
Posted at 20:52, Published in: Web Usability Design