Forum Moderators: open
6 years ago - from Top Ten Mistakes of Web Design(useit.com 1996) [useit.com]
10 seconds as the maximum response time before users lose interest
In the 1999 update of that article, the item was still a problem.
Although speed of connections may be increasing, the amount of time that people are willing to wait for a page to load or the time it takes to find something is decreasing or staying at those short time limits. Maybe people are busier or bored with bleeding edge designs?
I ran across a web site that states the top 100 or so retail sites online had pages that loaded up in under 5 seconds, and if your page took up to 15 seconds to load, over half your visitors or close to it, left the web site.
Thoughts?
[xat.com...]
We run our sites through it just to see what happens.
It's MS Windows.
zooloo
I do recall Yahoo! page load times of 6 seconds [years ago]. My switch to Google was based on an even quicker load than Yahoo!
It would be interesting to see some studies or stats on page loads and user preferences. Though I do know my own well enough.
General rule of thumb I take is that you do not want the pages to be bigger then 40-50k (I am too hungover to do the math on that one, but I think you are looking at under 10 seconds for a decent dialup connection).
And that size includes images for me, so I usually check total size by using Netscape 4.X, which shows you the total page size in the properties, including the images.
Obviously it really depends on your customer base though. If it is a homepage for a broadband gamers site, I think it is safe to be alot larger then that.
I do seem to recall some ISP in Aus having a homepage of more then 1 mb (due to an imbedded sound file - honestly how obnoxious *shudders*), that was quite amusing considering they were a DU service provider ;)
I don't wait more than 15 seconds. I give it 10 seconds if nothing starts to appear right away. I don't stare at a white screen for more than 10 seconds.
Period.
I am with you on this one, but equally frustrating is loading up a web site like weather.com or numerous others lately, and waiting for some odd flash thing to load up completely then cover all the accessible options to the information you need to see or use.
It forces the user to wait, until it is loaded, wait again to see if it self destructs and vanishes, or wait until the user decides it is OK to click on the ad or [x] to close that window if they left one visible.
I work with a variety of multi media and web design, so I do not feel bad about the use of Flash. However, I can see Flash getting a nasty reputation if this sort of thing continues. It is NOT neat to have your work flow or directions imposed upon or time wasted. I feel it will cost both the web sites using these types of options or gimmicks, and also cost more people to feel negative about Flash formats. Flash is a great tool and app when used for better reasons. How many people really read these ads or follow them? Is it worth it? Multimedia production is costly and I can think of better ways to re route that business investment.
I cannot see any disadvantage to having a web page load fast. None.
And then, if an image is in jpeg format, the browser still needs to decompress it after the whole file has arrived - and with a high level of compression, that can add a few more seconds.
One last culprit in slow loading pages can be dynamic page calls to an overloaded server, or a slow to respond database, or lots of server side scripting. So the time to render a page may involve more than just file sizes.
A similar page that takes exactly the same length of time to load, but renders-while-u-wait will be percieved to take less time to load, and will therefore be better loved. Really the lesson is focus on the user, not the technology. Mind control works.
This is what I like when you're able to use xhtml/css designs - something happens almost instantaneously as the html is very light. If the sylesheet takes a few moments to render the page the way it looks, no problem.
I think the 'white screen wait' comes from sitting everything in one big table, the browser won't show it till everything has loaded and it's calculated the columns and rows right?
Of course there are ways around that: Just use a seperate small table for your header. This way the user gets something pretty quick whilst the rest of the page downloads and renders...
Nick
One last culprit in slow loading pages can be dynamic page calls to an overloaded server, or a slow to respond database, or lots of server side scripting.
Photo gallery pages[they are going to take longer to load.] It amazes me the difference in times it takes to load up a full page of thumbnails from one gallery to the next.[first visits, images were not cached on my end]. Some gallery image serving applications may be quicker than others.
TallTroll
A similar page that takes exactly the same length of time to load, but renders-while-u-wait will be percieved to take less time to load, and will therefore be better loved. Really the lesson is focus on the user, not the technology. Mind control works.
Unless, of course, users are waiting for the important items like content or information to be loaded.
In the case of full page Flash, though, I think the only option is to wait or stare at a countdown screen. Yet it is important to have that countdown screen to let the user know something is happening. So, maybe those designers or developers can sneak content into the load screen. Small slogans, important points, summaries of services or prodcuts?
Nick
I think the 'white screen wait' comes from sitting everything in one big table, the browser won't show it till everything has loaded and it's calculated the columns and rows right?
That is true, though nested tables can make pages take even longer to configure, especially if there are images and table dimensions that have not been given a size value[again this could be an item with generated galleries of images with various sizes, too]. So a designer and developer has to have a plan, what do I want to load up right away?
---
In the end though, I believe that commercial or retail sites do not afford the luxury of taking a lot time to load or having users lost in features on the page that do not seem to get them to point B fast.
In the brick and mortar world... retail places hot items or items that need to be found right in the buyer's view - up front, main aisles, and eye level are used. Sure they distract them with a few others or associated items along the way. A commercial web site is often built similar to the retail model. If someone has to walk around looking too long, waiting in lines too long, or uncovering items like a detective - they may end up going to another store and even spending more on the item. Time is important, whether it is online or off.
Too many clicks, too many distractions, and too many seconds mean less revisits or people giving up.
Again, I am wondering about the stats involving users at web sites. 100+ top retail load in much less than 10 seconds - I wonder if Load is linked to that fully usable state or when they believe they see enough to stick around?
Do we have any current numbers out there?
holly
Try WebSpeed Simulator - you can set it to different modem speeds to see how your page loads.
[xat.com...]GetPageSize! [q42.nl] - bookmarklet tool.
Right click on the link at the page and put it on the toolbar or into favorites in IE5.5 and up.
When a page is loaded in the IE browser, click on this bookmarklet and it will return total size of page with all associated images, stylesheets, scripts, background-images, embeds and objects.
... it works nicely.The speed/timing calculation is set for a 56k modem connection. Those working with scripting may be able to make a few alternatives that will calculate other connection speeds.
Not current, but in 1999 Zona Research did a study that found 30 per cent of users would abandon a slow-loading page within 8 seconds, and 70 per cent would leave within 12 seconds. For full details of this research, do a search for 'Zona Need for Speed'. Its available as a PDF on the Keynote web site.