System Performance Objectives of the User Experience


Author: Karl Vochatzer
Last Update: 02/28/2000

The goal is to maximize the user's control of the system (or perception thereof) through appropriate response times of system actions and events. Additionally, where there are current technological constraints to achieving a desired goal a response time goal will be set that may be attainable through optimizing on the interaction of the variables that go into the equation .

Take page downloading as an example of an event that has many variables that figure into the response time. Near instantaneous download times would certainly be desirable but hardware limitations currently restrict this from happening. Given the current technology constraints and hardware platform chosen, some of the constraints are that modems and phone lines max out at 56k so downloads above that or even approaching that are not attainable today. Therefore a goal for this example might be to download pages to the client within 10 seconds of the initiating user action. Accomplishing this goal requires investigating and adjusting the known variables or constraints to fast download times (e.g., page size, bandwidth, processing power, video graphics power, server factors, etc.). Decreasing and/or increasing the size, rates, power and speed of the variables may allow us to achieve or come closer to the desired performance goal.

The set of actions and events listed below are taken from Marc's architecture meeting on Tuesday February 22, 2000. It was suggested that a few of them were prioritized as affecting the user experience. I would content that all of them should be considered as directly impacting the user experience with the exception of logging for our benefit. I have proposed the following response time goal after each action or event in accordance to published usability guidelines for giving the user the sense that they are controlling the system and not the other way around.

  • Logging: not a consideration of the user, but whatever the response time is it should not impact the user in any way
  • Cached Content Access: < 100 ms to give the perception of instantaneous results
  • Dial-out Delay (connection time): < 10 sec. - the point at which users start losing attention rapidly. Visual feedback should be provided to keep the user informed of the progress. An informed user is more forgiving of longer wait times than an uninformed user. Basically, the current wait time of nearly 60 seconds to set the modem, dial-out, connect, and receive data needs to be greatly reduced to approach the 10 second goal.
  • Line Speed (also progressive load of HTML): current hardware/line limitation - as close to true 56k as possible
  • DNS Lookup: < 1 sec. - to allow the user's flow of thought to remain uninterrupted. At the completion of the lookup the user should begin to see the progressive download of the page requested and be complete in less than 10 seconds total time.
  • Navigational Speed: 100 ms for cached content and less than 1 second for server side content navigation (keeping in mind that the complete download of the server side page should occur in less than 10 seconds, the Page Download and Display Time).
  • Time-out Values: time-out can occur at 15 minutes after the last keyboard or mouse action. Although I don't have any usability data to support this time-out goal, it should be based on the time from the last user initiated action and not just on the fact that the user has not initiated a server side download. This should help to minimize the frustration experienced during a reconnection.
  • Hang-up Response Time: < 1 sec. - a slight noticeable delay is acceptable. Currently it appears to take about 2 seconds to get the connection light on the i-opener to turn off. Users have commented that they think that they have to hit the key twice to make it hang-up when it is really the fact that it takes over a second to get any feedback that the key press took.
  • Content Update Time: we will likely never figure out the best time to deliver content updates for the wide variety of users situations and user's needs in our target market. With that in mind, it is recommended that the event be user configurable so that the user gets a sense of control for the updates.
  • Multiple Redirects: not recommended, especially if it means that each redirect page is flashed before the screen each time. See the current implementation (02/28/2000) of the "going to the web page" redirect prior to the display of the "off-line... please wait" HTML message.
  • Application Performance (e.g., mail): < 1 second, but as close to 100 ms as possible so that perceived delays are minimized
  • Screen Saver Initialization (not in Marc's list, but it should be recorded): based on the disconnect event at 15 minutes of idle time, it is appropriate to initialize the screen saver at the same time. However, it would be in the interest of the user to actually initialize a little prior to the disconnection time. Doing this may alert the user and get them back on task. However, the earliest recommended time would be at 10 minutes of idle time - any sooner and it could begin to disrupt some users that really are still looking at the current screen (maybe they read really slowly!).

Based on Jakob Nielsen's book Designing Web Usability (pages 42-53).



Back to portfolio page

e-mail: kvochatzer@yahoo.com