Thread

How much would you say the website response time influences visitor behaviour?

Website I’m working on sits on a server in London UK. Average response time (time it takes to download homepage HTML) is 864 ms.

The “problem” is that audience is quite international and yet website response time varies from 500ms (London) to 1200ms (Los Angeles).

I’m trying to predict the benefit of moving the website to say Rackspace cloud where response time sits at constant(ish) 700ms(ish) from across the world.

Any thoughts on relation between response time and visitor stats (e.g. pages-per-visit, bounce-back rate)?

Any pointers?

Thanks!

A second is fine. After three I am beginning to lose patience…

Hey Milan,

This article might be of interest: http://www.useit.com/alertbox/response-times.html

Also, I was in a the same boat a short while ago where the average response times from MediaTemple was around 1300ms which I found totally unacceptable. After exchanging quite a few emails where I felt like I was talking to a brick wall, I switched to EngineHosting and I haven’t regretted it.

I’ve attached my pingdom report of the last 3 months showing you what the results were like (bear in mind that I’ve set it to ping from all global servers so it really is a good idea of average response time worldwide).

The average response time in the period that I’ve been with EH only is around 620ms.

Tony, thanks for the article - it was an interesting read.

Your website average response times is not averaged across your visitor base, but across Pingdom test servers.

In my case average is 864ms, but majority of our readers are in US, where average is 1100ms and 1200ms from east coast (not as snappy as I’d like; as your article points out - snappy website = lots of conversions).

Interestingly, Bloomberg website has average of around 70ms (ranging 13-150) across the world.

Sadly we don’t have resources to compete with them…

Now I just need to figure out how to reduce the load times without spending a fortune 😊

Your website average response times is not averaged across your visitor base, but across Pingdom test servers.

I know, that’s why I was careful to point out that I’ve set pingdom to use all the global servers 😊 Since the majority of my visitors are from North America, I can rest knowing that their ping is even lower than the average ping (which I’m quite comfortable with all things considered).

I think the response time drops (or rises) in relation to the “need” for the particular content and/or the options that the consumer has.

For instance, if you are searching for some experiences of other other people regarding a rare disease you have - you would wait longer than if you were looking for a stock quote.

I am in general agreement with the idea of less than 2-3 seconds for a complete page to render, but that becomes more true if the user needs to visit a lot of pages - it also depends on the users expectations. If they have a relatively slow connect (256K DSL, etc.), then they wait longer for a lot of content.

I think the real answer is “it depends” or “as fast as possible”.

I have found that my visitors (very targeted) will put up with quite a lot of lag…I found this out by accident when we had problems with our server and it bogged down (open mysql ports, etc.).

Back in the 56K modem days, the threshold might have been 20-30 seconds or even one minute! So response time is certainly a moving target. Google sorta moved the bar on this issue…..

(FYI, It took about 3 seconds for this to post - I was starting to think it was very slow)

I agree - website I’m working on is for a niche market (news, blogs and technical papers).

Visitors are unlikely to be able find this information elsewhere so are likely to tolerate “longer” response time.

On the other hand, my assumption is that they would respond positively to a more snappy website (so more pages-per-visit, less bounce-backs, more returns, more purchases).

I’m trying to predict the benefit of shaving off say 300ms (800ms to 500ms) from response times.

So far it seems that my only option is to experiment… and that’s going to be time/money consuming 😊

Good to keep in mind that Pingdom reports the time for loading _all_ the resources for a page—as if you had a browser with no cache.

This is markedly longer than it typically takes to load a real page, after the viewer has first landed on your site, because then most/all of the background css, javascript, set imagery, etc. is then cached on the browser.

Nontheless, Pingdom does indicate truths, and I have have had just the same good experience in switching to EngineHosting, even on their lowest cost option. A little monitoring over a day shows how much differently they act during high use time compared to other shared hosting - there is real evidence of managed load sharing going on. I don’t know the specifics, but it works.

Regards,
Clive

Good to keep in mind that Pingdom reports the time for loading _all_ the resources for a page—as if you had a browser with no cache.

I was talking about the actual test (as opposed to Pingdom Tools) which report time it took to download the HTML code only.

i think handyman is right. it depends on what you are looking for. but of course, the less time, the better the surfing experience. e.g. i was looking for green card informations, because a friend of mine is thinking about moving to the states and most of the non-official pages were hosted on very lame servers which made my short research an odissey 😉

Google published an article a few years back that showed for them that the faster a page loaded, the more likely a user was to interact with the page (i.e., continue browsing, clicking, etc.)

A couple things for you to consider:
1) http://www.webpagetest.org/ - lets you simulate page downloads at different speeds from around the world. I use pingdom too, but this allows more control if you need it.
2) Consider a CDN and offload static content (images, CSS, JS) to the CDN so it can be served geographically closer to the user.

I’ve got to admit I find this very interesting as I never knew it played such an important factor! The article was a very interesting read and I will be following up on the tips given. Hopefully this will improve my progress dramatically!!!

Also, I was in a the same boat a short while ago where the average response times from MediaTemple was around 1300ms which I found totally unacceptable. After exchanging quite a few emails where I felt like I was talking to a brick wall, I switched to EngineHosting and I haven’t regretted it.

Same thing, only I switched to Linode VPS and also haven’t regretted 😊 Response time is very important, if you’r serious about your site.