The Economic Value of Rapid Response Time (1982)

System response time can be further divided into:

When online systems first began to spread throughout the business world, psychologists such as Robert B. Miller, then of IBM’s Poughkeepsie laboratory, argued that two seconds was the longest a person should wait for a response from the computer. In a pioneering article, inspired by Doherty’s work, Arvind J. Thadhani, of IBM’s San Jose Laboratory, suggests that the number of transactions a programmer completes in an hour increases noticeably as system response time falls, and rises dramatically once system response time falls below one second. The system and user cost for this time were estimated at $900,000 monthly (Figure 6), 15 times the incremental cost of a new processor capable of providing sub-second response time to 500 simultaneous users.

Source: jlelliotton.blogspot.com