Latency Guidelines

12 May 2004 - 8:02am
9 years ago
10 replies
723 reads
Ben Speaks
2004

Comrades,

I am looking for human factors guidelines related to latency thresholds.
Specifically, for hardware input devices that interact with graphic
based interfaces (for example the mouse to a computer screen). When a
user interacts with an interface what are their expectations in
milliseconds for feedback?

I have found a lot of published stuff from manufactures where they
provide latency guidelines for their own devices but I am having less
luck finding the studies and research that support such guidelines.

Does anyone know where I could go to find such information?

Thanks,
Ben Speaks

Comments

12 May 2004 - 8:34am
Josh Seiden
2003

Have you looked at http://www.hcibib.org?

JS

>
> Comrades,
>
> I am looking for human factors guidelines related to
latency
> thresholds. [snip]

12 May 2004 - 12:17pm
Kristoffer Åberg
2003

Some related nielseniana: http://www.useit.com/papers/responsetime.html

You could also try your luck with this one:
http://tim.griffins.ca/writings/usability_body.html

HTH,

/Kristoffer

> Comrades,
>
> I am looking for human factors guidelines related to latency thresholds.
> Specifically, for hardware input devices that interact with graphic
> based interfaces (for example the mouse to a computer screen). When a
> user interacts with an interface what are their expectations in
> milliseconds for feedback?
>
> I have found a lot of published stuff from manufactures where they
> provide latency guidelines for their own devices but I am having less
> luck finding the studies and research that support such guidelines.
>
> Does anyone know where I could go to find such information?
>
> Thanks,
> Ben Speaks

12 May 2004 - 12:35pm
Andrei Herasimchuk
2004

On May 12, 2004, at 6:02 AM, Ben Speaks wrote:

> I am looking for human factors guidelines related to latency
> thresholds.
> Specifically, for hardware input devices that interact with graphic
> based interfaces (for example the mouse to a computer screen). When a
> user interacts with an interface what are their expectations in
> milliseconds for feedback?

Not to sound smarmy, but isn't the answer always "instantaneous?"

You are talking expectations. 8^)

Are you looking for what they'll tolerate, instead of what they expect?
And if so, for specific kinds of behaviors? Does context play a big
role here? For example, in Apple Mail, I click the Account menu in a
new message, and sometimes I get the color wheel progress cursor while
Mail is doing whatever the heck its doing, then four seconds later, it
gives me control of the menu and I can switch accounts. This drives me
batty to no end. I want it to be instant. But I'm tolerating 4 seconds.
However, I'm not willing to switch to Entourage, which I can't stand on
the Mac. So, I tolerate it.

But if the menu switch took upwards of 10 seconds or so every time I
hit it, I would probably dump Mail and search for a new email client.
Heck, I'd even use a browser based mail program instead!

Andrei Herasimchuk
andrei at adobe.com

work: http://www.adobe.com
personal: http://www.designbyfire.com

12 May 2004 - 1:40pm
tomchi at ok-ca...
2004

Here's an excerpt from an OK/Cancel article:

thegoodtomchi wrote (07 Oct 03) :
Yes, you make the important distinction between faster vs. better for users. The topic really deserves an entire essay, but in brief, it's true that faster is not necessarily better. The continuum goes something like this: (all numbers are approximate)
**
0-10 ms : very fast. Anything in this range feels instantaneous (though 10ms is about 10 million instructions these days). Any UI optimization to speed things up faster than 10ms goes pretty much unnoticed.

10-100 ms : fast. A lot of apps exist in this range. System feels very responsive and you'd only notice the lag if you were in a situation where you want true realtime (e.g. gaming, playing a software synth, etc).

100ms - 2sec : medium. These can feel a little slow, but not slow enough for you to think about going to do other things while it executes. Plenty of apps are in here too.

2-40sec : crappy. This is terrible range because it is a long time to wait for an action, but not enough time for you to switch and do something meaningful while you wait. Most of the web is here. Argh.

40sec-5min : Slow... but now a funny thing starts to happen. When tasks get this long, you no longer think of them in terms of response/unresponsive. You just fire them off and get back to them sometime later. Examples of this are using "Find" to find a file in the system, or Replicating a big chunk of email. Because you are not waiting for them, they actually are less of a mental hassle than 2-40sec zone.

5-60min : These tasks are interesting because they are long enough for you do to some meaningful thinking in the time they are running. When I did astrophysical research I would run these batched modelling processes, and in the 20 minutes it took to run, I'd be reviewing and analyzing previous results. Sometimes I think they even made me more productive because it was like a 20 minute egg-timer telling me that I should have finished 20 minutes of work - If it finished before I could review the last results, I felt like I needed to speed up.

60min+ : well you get the idea.

---------- Original Message ----------------------------------
From: Andrei Herasimchuk <andrei at adobe.com>
Date: Wed, 12 May 2004 10:35:37 -0700

>On May 12, 2004, at 6:02 AM, Ben Speaks wrote:
>
>> I am looking for human factors guidelines related to latency
>> thresholds.
>> Specifically, for hardware input devices that interact with graphic
>> based interfaces (for example the mouse to a computer screen). When a
>> user interacts with an interface what are their expectations in
>> milliseconds for feedback?
>
>Not to sound smarmy, but isn't the answer always "instantaneous?"
>
>You are talking expectations. 8^)
>
>Are you looking for what they'll tolerate, instead of what they expect?
>And if so, for specific kinds of behaviors? Does context play a big
>role here? For example, in Apple Mail, I click the Account menu in a
>new message, and sometimes I get the color wheel progress cursor while
>Mail is doing whatever the heck its doing, then four seconds later, it
>gives me control of the menu and I can switch accounts. This drives me
>batty to no end. I want it to be instant. But I'm tolerating 4 seconds.
>However, I'm not willing to switch to Entourage, which I can't stand on
>the Mac. So, I tolerate it.
>
>But if the menu switch took upwards of 10 seconds or so every time I
>hit it, I would probably dump Mail and search for a new email client.
>Heck, I'd even use a browser based mail program instead!
>
>Andrei Herasimchuk
>andrei at adobe.com
>
>work: http://www.adobe.com
>personal: http://www.designbyfire.com
>
>_______________________________________________
>Interaction Design Discussion List
>discuss at interactiondesigners.com
>--
>to change your options (unsubscribe or set digest): http://discuss.interactiondesigners.com
>--
>Questions: lists at interactiondesigners.com
>--
>Announcement Online List (discussion list members get announcements already)
>http://interactiondesigners.com/announceList/
>--
>http://interactiondesigners.com/
>

________________________________________________________________
$0 Bannerless Web Hosting, 10 POP and Web Email Accounts, & more
Get It Now At www.doteasy.com

12 May 2004 - 2:23pm
Andrei Herasimchuk
2004

This is great information, Tom, but I still wonder if the real question
is: If given a choice when working with a computer, would you/users
prefer instantaneous response or not?

I know searches can take a long time, but now I'm spoiled by Google
search speeds -- and that's over a pipeline, not even in resident RAM
on my machine. I expect everything to be lickety split now.

The original post posed the question of "user expectation." My
expectations are I want things to be instantaneous. I'm still wondering
is the original question is more of what will users tolerate and in
what contexts.

Andrei Herasimchuk
andrei at adobe.com

work: http://www.adobe.com
personal: http://www.designbyfire.com

12 May 2004 - 3:09pm
tomchi at ok-ca...
2004

Instantaneous is nice in a lot of instances, but two things:

1) Depending on the system you are building and the technology you are using, instantaneous is not always possible. Google raises the bar and shows what can be done, but remember their speedy performance is due to a ton of investment in both algorithms and hardware. Not all projects have the resources to pour into these spaces.

2) For some tasks, instantenous may not be best. Human minds can only operate up to a certain speed, and managing something that responds very fast could be a very stressful situation. For example, if you worked an assembly line, the machines could give you a new part to tweak the instant that you finish the last one. Psychologically this might induce stress because you never feel you are finished, and the rapid response of the system puts inhuman expectations on your performance.

Granted, for most jobs and tasks in the world we still have to wait for the machine (esp. in the 2-40 second range). But my spectrum of response times is meant to argue that there are multiple sweet spots that you can design your system to operate in. I think designing a sweet spot is more important than setting a user expectation of latency and meeting it.

---------- Original Message ----------------------------------
From: Andrei Herasimchuk <andrei at adobe.com>
Date: Wed, 12 May 2004 12:23:08 -0700

>This is great information, Tom, but I still wonder if the real question
>is: If given a choice when working with a computer, would you/users
>prefer instantaneous response or not?
>
>I know searches can take a long time, but now I'm spoiled by Google
>search speeds -- and that's over a pipeline, not even in resident RAM
>on my machine. I expect everything to be lickety split now.
>
>The original post posed the question of "user expectation." My
>expectations are I want things to be instantaneous. I'm still wondering
>is the original question is more of what will users tolerate and in
>what contexts.
>
>Andrei Herasimchuk
>andrei at adobe.com
>
>work: http://www.adobe.com
>personal: http://www.designbyfire.com
>
>_______________________________________________
>Interaction Design Discussion List
>discuss at interactiondesigners.com
>--
>to change your options (unsubscribe or set digest): http://discuss.interactiondesigners.com
>--
>Questions: lists at interactiondesigners.com
>--
>Announcement Online List (discussion list members get announcements already)
>http://interactiondesigners.com/announceList/
>--
>http://interactiondesigners.com/
>

________________________________________________________________
$0 Bannerless Web Hosting, 10 POP and Web Email Accounts, & more
Get It Now At www.doteasy.com

12 May 2004 - 4:09pm
Dave Collins
2004

>Instantaneous is nice in a lot of instances, but two things:
>
>- instantaneous is not always possible...
>- instantaneous may not be best...

On that second note, system response time should take into account the
user feedback loop. Visual, and sometimes audible feedback helps the
user know the system is behaving. It is good for the user's piece of
mind to have the system let them know they did something before moving
on.

It becomes important where some parts of the system can possibly behave
less than instantly. Consider a case where 99% of your interface is
instant, and then 1 screen gets stuck for a fraction of time. The user,
used to the instant reaction, thinks suddenly that they didn't press
hard enough, and continues to press - with arbitrarily bad consequences
for system and/or user.

On the other hand, a micro-brief delay while displaying feedback says to
the user "Yes, I heard you." - which, on a delayed screen, implies "Yes,
I heard you. Hang on a sec."

Dave

12 May 2004 - 5:09pm
John Vaughan - ...
2004

Zackly so, Tom Chi. And then there's "perceptual threshold".

i.e. Animation, films and television "work" exactly because us humans don't
perceive the fact that the illusion of "lifelike" motion is created from a
succession of quickly changing still images. Digital images - which were
laughably crude, chunky and clunky only a few years ago - now far outstrip
our perceptual threshold. One of our first object lessons came with the
making of computer-generated spaceflight sequences for "Star Wars". The
visually crisp imagery moved right, but didn't look right until Industrial
Light & Magic realized that they needed to smear the spaceships a little
(kind of like the blurry moving image artifact you get in a still frame).
Only then did us humans accept that the image was real.

And lets's not forget the other positive aspects of human frailty.

I've always enjoyed this story: People were terribly upset with the
slowness of the elevator service at a huge skyscraper. It was a major
problem. The building was gigantic, the technology was already operating at
its absolute limits, it was impossible to reengineer or redesign anything
major, yet within a few days complaints about the slowness of elevator
service disappeared completely. The solution: Floor-to-ceiling mirrors
were installed in the lobby..... Never underestimate the power of human
curiosity, self-absorbtion, misdirection, vanity, etc. In the usability
game, perception is everything.

John Vaughan
email: vaughan1 at optonline.net
website: http://www.jcvtcs.com
115 Minnehaha Blvd
Lake Hiawatha, NJ 07034
Voice: (973) 316-1182
Fax: (973) 316-1183

----- Original Message -----
From: "Tom Chi" <thegoodtomchi at ok-cancel.com>
To: "Interaction Discussion"
<discuss-interactiondesigners.com at lists.interactiondesigners.com>; "Andrei
Herasimchuk" <andrei at adobe.com>
Sent: Wednesday, May 12, 2004 4:09 PM
Subject: Re: [ID Discuss] Latency Guidelines

> Instantaneous is nice in a lot of instances, but two things:
>
> 1) Depending on the system you are building and the technology you are
using, instantaneous is not always possible. Google raises the bar and
shows what can be done, but remember their speedy performance is due to a
ton of investment in both algorithms and hardware. Not all projects have
the resources to pour into these spaces.
>
> 2) For some tasks, instantenous may not be best. Human minds can only
operate up to a certain speed, and managing something that responds very
fast could be a very stressful situation. For example, if you worked an
assembly line, the machines could give you a new part to tweak the instant
that you finish the last one. Psychologically this might induce stress
because you never feel you are finished, and the rapid response of the
system puts inhuman expectations on your performance.
>
> Granted, for most jobs and tasks in the world we still have to wait for
the machine (esp. in the 2-40 second range). But my spectrum of response
times is meant to argue that there are multiple sweet spots that you can
design your system to operate in. I think designing a sweet spot is more
important than setting a user expectation of latency and meeting it.
>
> ---------- Original Message ----------------------------------
> From: Andrei Herasimchuk <andrei at adobe.com>
> Date: Wed, 12 May 2004 12:23:08 -0700
>
> >This is great information, Tom, but I still wonder if the real question
> >is: If given a choice when working with a computer, would you/users
> >prefer instantaneous response or not?
> >
> >I know searches can take a long time, but now I'm spoiled by Google
> >search speeds -- and that's over a pipeline, not even in resident RAM
> >on my machine. I expect everything to be lickety split now.
> >
> >The original post posed the question of "user expectation." My
> >expectations are I want things to be instantaneous. I'm still wondering
> >is the original question is more of what will users tolerate and in
> >what contexts.
> >
> >Andrei Herasimchuk
> >andrei at adobe.com
> >
> >work: http://www.adobe.com
> >personal: http://www.designbyfire.com
> >
> >_______________________________________________
> >Interaction Design Discussion List
> >discuss at interactiondesigners.com
> >--
> >to change your options (unsubscribe or set digest):
http://discuss.interactiondesigners.com
> >--
> >Questions: lists at interactiondesigners.com
> >--
> >Announcement Online List (discussion list members get announcements
already)
> >http://interactiondesigners.com/announceList/
> >--
> >http://interactiondesigners.com/
> >
>
>
> ________________________________________________________________
> $0 Bannerless Web Hosting, 10 POP and Web Email Accounts, & more
> Get It Now At www.doteasy.com
>
>
>
> _______________________________________________
> Interaction Design Discussion List
> discuss at interactiondesigners.com
> --
> to change your options (unsubscribe or set digest):
http://discuss.interactiondesigners.com
> --
> Questions: lists at interactiondesigners.com
> --
> Announcement Online List (discussion list members get announcements
already)
> http://interactiondesigners.com/announceList/
> --
> http://interactiondesigners.com/

12 May 2004 - 5:59pm
id at ourbrisba...
2004

Quoting Ben Speaks <Ben.Speaks at hcrest.com>:
> I am looking for human factors guidelines related to latency thresholds.
> Specifically, for hardware input devices that interact with graphic
> based interfaces (for example the mouse to a computer screen). When a
> user interacts with an interface what are their expectations in
> milliseconds for feedback?
>
> I have found a lot of published stuff from manufactures where they
> provide latency guidelines for their own devices but I am having less
> luck finding the studies and research that support such guidelines.
>
> Does anyone know where I could go to find such information?

Scott MacKenzie's done a little of the work here for you:
http://www.yorku.ca/mack/phd-references.html

Best regards,

Ash Donaldson
User Experiennce Designer
"It depends."

10 Jun 2004 - 1:56pm
cfmdesigns
2004

Andrei Herasimchuk <andrei at adobe.com> writes

>On May 12, 2004, at 6:02 AM, Ben Speaks wrote:
>
>>I am looking for human factors guidelines related to latency thresholds.
>>Specifically, for hardware input devices that interact with graphic
>>based interfaces (for example the mouse to a computer screen). When a
>>user interacts with an interface what are their expectations in
>>milliseconds for feedback?
>
>Not to sound smarmy, but isn't the answer always "instantaneous?"
>
>You are talking expectations. 8^)

I don't think you'll find any real users who really expect (vs.
"want") instantaneous.

What is most desirable is something approaching as fast as the eye
can track. After making an action, users "expect" a reaction within
the time it takes them to scan the screen for changes; any longer and
there's a perception of having to wait.

And in fact, "instantaneous" often isn't even desirable. With QE
testing work, I've seen occasional cases where the feedback came (and
went) too quickly, gone before the user could register what had
occurred. We sometimes have to slow things down to where the
feedback can be given and used.
--

----- ----- ----- ----- ----- ----- ----- ----- ----- ----- -----
Jim Drew Seattle, WA jdrew at adobe.com
http://home.earthlink.net/~rubberize/Weblog/index.html (Update: 06/08)

Syndicate content Get the feed