A really fast site, and Google don’t care

cmsofast Featured

Ian Lurie Nov 15 2011

I wrote a post, and this time lots of folks taught me good stuff. See the comments after the post for great info I’d totally forgotten on how Google calculates site performance

Google, I’m very… disappointed.

You may have noticed that Conversation Marketing is a wee bit more speedy now. Last week, a page on this site loaded in 5 seconds. Now, pages load, on average, in under 2. And some pages go faster than stink:

load speed report from pingdom

Wow, that’s fast. Load speed report from Tools.Pingdom.com.

I generated that report using tools.pingdom.com. Go give it a try if you want.

So, why don’t you give a crap?

How I sped up my site

If you want your site to go really, really fast, try the steps I used. This gets a bit technical, but it’s worth figuring out. Or just hand this list over to your web developer:

  1. Enable GZIP compression on your server. If you don’t know how, hound your webmaster until they do it. GZIP smushes down all of the files your web site uses, shoots them across the internet, then lets your browser un-smush them. That really speeds page downloads.
  2. Write decent code. If your home page uses more than 10,000 lines of HTML code, something’s wrong. If you don’t know that by now, and you’re supposedly a web ‘pro’, do us all a favor: Sell your copy of ExpressionEngine and learn to code using a text editor.
  3. Remove all embedded javascript (javascript that’s right in your page source) longer than 5-10 lines and put it instead in separate .js include files. That way, your web browser can load those files once and cache them—save them on your local hard drive. If you use those files throughout your site, your browser still only loads ‘em once.
  4. Do the same thing with any embedded CSS of more than 10 lines. Same reason.
  5. Learn to compress your images. Use JPEG compression on photos, and use GIF or PNG for images with fewer colors. If you don’t bother, you suck.
  6. Configure your server to deliver an expires header at least 4 weeks in the future for any ‘static’ files, such as javascript, CSS, or images that you use in your navigation. That expires header will tell visiting browsers, again, to save stuff to the local hard drive, so visitors only load it all once. Again, if you’re a web developer and don’t know this yet… I don’t know. Consider a career in food services. Or, buy a book about speeding up web sites (affiliate link). Or read Yahoo!’s guidelines for speeding up a site.
  7. ‘Minify’ everything. ‘Minification’ removes spaces, carriage returns and comments from your code. If you’re using big javascript libraries like jquery, this can really speed things up.
  8. Resize images in a photo editor. Don’t use HTML to resize an image. That forces browsers to load huge images, even if you resize them to postage-stamp size. If you resize them in an editor first, then visiting browsers download the image size they need, and you’re not trying to ram a beach ball into a wine bottle.
  9. Use a content distribution network (CDN). This was the big one. I’d already done all of the steps above in previous site sweeps. But I set up an account with MaxCDN, then installed their plugin on my blog, and voila: I shaved 2 full seconds off of all load times.

Google doesn’t care

If there’s one disappointment in all this, it’s Google. They have a site performance analyzer in Google Webmaster Tools. It sucks.

According to Pingdom, pages load in 1.35-2 seconds.

And, according to Google Page Speed, my site now scores a 94/100:

google page speed report

Pretty high marks.

But according to Google Webmaster Tools, my site is driven by four tired squirrels, all running in opposite directions:

chart

Why, Google? Why?

I’d love to hear an explanation from the Googlers, if anyone’s reading this.

This shakes my faith a little bit. Yes, faster-loading pages are better. But will my rankings really improve if Google can’t accurately measure load speed? Or is the report itself just inaccurate?

I’ll keep you all posted.

tags : conversation marketing

10 Comments

  1. Greg

    Try measuring your bounce rate and the time spent on your site.. How do they compare to pre-optimisation?… Those factors are a better indicator of improved user engagement… And surely that will help drive SE traffic in the long term…

  2. Hi Ian,

    I too worked on making my site faster by doing much of what you did above. I am also using WordPress and W3 Total Cache, in conjunction with MaxCDN. The interesting part is that I currently have MaxCDN turned OFF within W3 Total Cache as I am doing a bunch of redirects, etc right now, yet I was still able to get my speed down below 2 seconds (under 1 second for many pages). The one issue you didn’t touch on, and that made a HUGE difference for me, was to activate APC (alternative PHP caching – using your server’s memory, instead of disk, to do caching). This had a tremendous impact on my site speed.

    Can’t wait until I turn back on MaxCDN to see how much more my site speed is improved…

  3. Afaik page-speed is measured by users’ Google Toolbar. Probably the reason why nothing changed so much. Don’t count on pingdom since they have a T1 connection. Makes the test somewhat pointless.

  4. eywu

    Pretty interesting stuff. Given that the PageSpeed metric is based on toolbar data, many times you’ll have to consider the audience. But I don’t think it would effect your site like others.

    For instance, on a site where the majority of the audience is using older browsers, they tend to also have slower internet connections which is a contributing factor. Even if you’re site is super optimized, if a large number of users are still hitting your site with dial-up or slow DSL / ISDN lines that can affect the load time displayed in GWT.

    Also, if you have a good amount of international traffic and your CDNs aren’t pushing out to edge servers near them, that would also be a culprit.

    I’ve also noticed that sometimes GWT lags in updating their stats. Sometimes I’ve observed them do “re-calculations” so when I came back the numbers were different.

    I’m not sure how long it’s been since you’ve made the optimization, but you might see something different if you “wait”.

    The last thing to consider is that you have a number of third party calls being made from your site. Many times ad servers or 3rd party plugins can make additional calls which can seriously lag for your visitors. These things also would contribute to a slower PageSpeed time.

    Looking forward to some of the comments and any future updates. It’s definitely a topic that I enjoy ^_^

  5. As someone who has gone on this journey myself, I can understand both the highs (the obvious simple speed increases) with the lows (Google ignoring your efforts). But you need to understand that there’s a distinct disconnect between the WMT data on site speed and actual measured performance – I’ll point you to a similar question (and it’s answer) on StackExchange – http://webmasters.stackexchange.com/questions/17137/how-exactly-is-google-webmaster-tools-measuring-site-performance

  6. Matt

    Matt Cutts addressed this issue following his keynote address at Pubcon. It was to a small audience that hung around afterwards, but I wish more had heard it.

    Here’s the gist of what he said: Page Speed is a small, very small ranking factor. It mostly comes into play if you are within the slowest 20%, which indicates that it is closer to a penalty than a ranking factor. He specifically said that page speed should be one of the last things that an SEO should focus on.

    Some of this may be due to the wildly inaccurate data that Google is trying to normalize. If I had to guess, CTR and bounce rate data weighs more heavily, given Matt’s emphasis on user experience. User experience could include page speed, but some other factors would seem more important there.

  7. Great post and even better comments. Thanks to all! I’m an seo/content marketing guy with little to no programming background. I have to rely on my php/html guys to cross my T’s and dot my I’s when it comes to these issues. Its a huge weakness for me and I always feel like my pants are down around my ankles. Any feedback on what tools I can use to run against my site so that I can feel confident that my programmers are following best practices.

  8. Lance

    I recommend checking out CloudFlare as they can help with 4-5 of the above list.

    I found ads and social media icons are a big problem. Google Analytics page speed (more accurate in my experience) measures after 2 seconds of no network activity. Plugins, especially the Google ads ironically, can take minutes to load even though the page is visible to the visitor. The only way I can explain some pages reporting as > 1000 seconds.

  9. Google loves google….our awesome redesign didn’t get much attention from the crawlers until we bumped up or Adwords spend….suddenly we found our site on page one.

  10. Hi Ian,

    I’ve also experienced the very same thing.

    Whilst Google does more or less everything else remarkably well, it seems that page speed tools they are quite content to just leave you in limbo with.

Comments are closed.