latency and expectation

12 Apr 2005 - 8:14pm
484 reads
Tom Hobbs

Hi all,

I know this is a subject that has been addressed to different degrees
on this list and elsewhere, but I have some specific problems I need
solid data regarding refresh expectation and response times. I was
wondering if any one can point me to something concise that
specifically deals with the context of this problem (I have some but it
isn't enough ammunition yet.)

It is dealing with unproven technology that is not too responsive but
as it is engineered to become more responsive, team members are finding
it difficult to distinguish between simply 'much better' and what users
are likely to find 'expectable' in relation to their expectations
(including myself.) I don't want to be the person who becomes
responsible when tons of money has been wasted on user testing and
product development to find out something is unusable when the data
already exists that shows it is probably not a workable solution.

To give some context of the interaction, this characterizes the key

i) locally based actions (these actions are not requested across a
ii) action to response is approx. 900 ms (until user sees a change on
screen – this is before process time becomes an issue)
iii) this is a consumer product aimed at tech savvy audience.
iv) users have viable alternatives to achieve the tasks/needs, only
advantage is this is mobile.

The interactions in question are:

i) scroll list — move selection
ii) text entry of keyword to phrase — time it takes character to appear
on the screen
iii) page turn/scroll — move from connected content across two pages
(i.e. paragraph starts on one page and moves to another.)

Remember, all this takes about 900 ms!! e.g. user presses 'a' and 900
ms later it appears on the screen (a user can type more letters and
multiple can appear at the same time, 900 ms later.) [Yes, I know how
frustrating it was typing remotely into PINE in email circa 1993!)

Looking at this in relation to accepted latency standards the answer
may seem obvious (i.e. 100ms-2secs is slow enough for people to notice
it is happening and pause) but that is not specific data to the problem
so is not seen as particularly relevant. It does have to be specific to
the above tasks and context.

thanks in advance,



Syndicate content Get the feed