Benchmark for website speed optimization

Question

I working on website speed optimization. I mostly used 3 tools for analyzing speed of optimization.

Speed analyzing Tools:

• Google pagespeed tool
• Yslow Firefox extenstion
• Web Page Performance Test

I am measuring performance using above tool and benchmark result as below like before and after.

Before optimization :

• Google PageSpeed Insights score : 53/100
• Web Page Performance Test : 55/100 (First View : 10.710s, Repeat view : 6.387s )
• Yahoo Overall performance score : 68

Stage 1 After optimization :

• Google PageSpeed Insights score : 88/100
• Web Page Performance Test : 88/100 (First View : 6.733s, Repeat view : 1.908s )
• Yahoo Overall performance score : 80

My question is ?

• Am i doing correct way ?
• What is the best way of benchmark for speed optimization ?
• Is there any standard ?
• Is there any much better tool for analyzing speed ?

Answer 1

No, there is no standard.

If you measure the same things in a predictable environment (say, your development or staging servers, being the only person with access during the test time frame) you can optimize to a certain extent and at least compare the results and the efforts (like you do). The tools you named provide you with a good load of recommendations (DNS, separating servers for static and dynamic content, CSS sprites, ...). Also take a look at the Chrome Developer tools, especially the Network and Timeline tab.

You can reduce the work and workload for a browser (files to request, memory for all assets, complexity for the rendering, ...) and for the server (the latter can be benchmarked with tools like Apache Benchmark, for example).

On a production site other aspects play a role, like the network setup and situation of your provider, DNS response times (DNS round-trip times), Load-balancing, HTTP and HTTPS, your client's situation and configuration, things like utilizing a content delivery network.

Answer 2

I use pretty much the same tools (and pingdom) so it sounds like you are doing the right things.

That said, an optimized first view time of almost 7 seconds is something that would cause me to do some serious reconsidering. I am a fairly new developer and I focus a lot on mobile compatibility so maybe 7 seconds isn't all that crazy.

However, if I click on something and it takes 7 seconds to load I might hit 'back' before it opens up. And if I don't, I am unlikely to visit another page expecting to wait 7 seconds each time.

Steve Souders is a great source of info on this topic. Personally, I Base64 encode images and embed them in my CSS and I conditionally (and async) load resources with Modernizr.