Open topic: How long is too long? Three Seconds
This topic is designed to provoke curiosity.
As last mile speeds get faster and faster it seems that customer expectations are moving in tandem. What used to be the “8 Second rule” became the “4 second rule” and is now below three seconds. Granted, you need to take the above study with some grains of salt due to one of the sponsors being Akamai, a service for hosting static content around the internet to improve time to delivery to the client. Just imagine this response time challenge to those who are constantly trying to change the timeout in LoadRunner to more than a 1000 seconds: Something is amiss in such decisions for it is unlikely that even a customer of an internal system will wait such long periods of time (16 minutes, 40 seconds) without picking up the phone and expressing their frustration to the help desk personnel of the “#%^&!^#^!@#%&*@#^@ slow application!”
What is often unsaid in such studies is that these expectations and abandonment rates are for sites where content is highly commoditized & where the cost of switching from one provider to the next is low. For those sites that are abandoned the opportunity cost of such an event is high, for the same user is unlikely to return until other sites also become unresponsive in their eyes. If you are a site with unique content or it is very expensive for your customer to switch you have the luxury of a bit of time. You can observe this in some financial institution web sites where the public site is lightning fast but the customer site is just a tad slower for pulling up accounts and working with them. As an account holder your cost of switching is high and you can’t just pull off and go to another website to look up your financial data. You can observe the same on government websites where they are the only game in town, why pay for the infrastructure to get to 2-3 seconds when the customer has no choice for another location?
How fast is fast enough? How fast is too fast, where the cost of achieving the response time is not worth the investment? How are you setting your customer’s expectations and reinforcing those expectations on responsiveness in your application design?
Replace ineffective offshore contracts, LoadRunnerByTheHour
. Starting @ $19.95/hr USD.
Put us to the test, skilled expertise is less expensive than you might imagine.
Twitter: @LoadRunnerBTH @PerfBytes
Re: Open topic: How long is too long? Three Seconds
We tend to look at it from the other direction: how slow is tolerable? Sounds cynical, but fast costs money -- qualified developers, more/better hardware, bigger network pipes, more time spent in performance testing and in iterative development loops. We *do* have the luxury of not being an Internet business (though we sell some product on the 'net). And we're not oblivious to customers' desires to download MSDSes or get an order invoice quickly. And there's no crime in having an app that works quickly.
But we try to get our clients to give us realistic performance goals. Sometimes we're surprised to find that very heavy transactions which weigh in at 15-25 seconds are "acceptable" -- until we learn that, before this app arrived, the same work was done by hand. Or that employees in some of our more far-flung international locations are used to 8- or 12-second response times. Faster would be great, but who pays the bill for all that? When response time is put in those terms, slower becomes much more acceptable.
Thanks for a thought-provoking topic, James.