Header background

Why your pagespeed score matters less than you think

You’ve put a lot of effort into the optimization of your website and you’ve finally earned a decent Google PageSpeed Insights score. You pride yourself on your optimized database queries, fast code, and slick well-designed frontend, but do you really know what kind of experience your customers are having with your site?

To investigate, let’s check out how well two of the top-listed websites on Alexa.com perform under real world conditions (at least as close as we can get with synthetic testing) utilizing www.webpagetest.org.

espn – An optimized site in the real world

The site espn.go.com/nfl/ is ranked number six in the list of top sports sites world-wide. It earns a decent score of 77/100 Google PageSpeed Insights for desktop browsers.

Assuming a user visits this site for the first time with a fast Internet connection (cable), load time is relatively consistent between the US East and US West coasts (they differ by about 2 seconds).

espn-ny
espn-la

What if a user doesn’t have a cable connection and they’re visiting from a different location? What if they’re visiting from Virginia and using a Firefox browser with a 3G connection?

espn-virginia

In this scenario it takes more than three times as long before the content begins rendering (Start Render time) and more than 20 seconds before the document is completely loaded. Not to mention the 33 seconds it takes for the page to load completely.

What if we rerun the same test using IE11 instead of Firefox?

espn-virginia-ie11

This scenario takes an additional 1.2 seconds longer before page content begins rendering!

What if we run this test from the UK, a key ESPN market? Even with cable speed (using IE9) you have to wait 23 seconds before the document is complete:

espn-london

Note though that the Start Render time using IE9 in London is 54% faster than it is for a user in Virginia working with a medium speed connection and browsing with IE11.

Now what about a user connecting from Paris using a “fast” 3G connection and IE8? When will rendering start? What will be the load time? Will it be slower or faster than a London connection? And how will this compare to the connection from Virginia?

Remember, all these tests are based on the same page with the identical Google Page Speed Insights score:

espn-paris

The total Load Time with 3G speed is less than in London with cable (IE9), but the visitor from Paris (IE8) has to wait more than twice the time until the page begins rendering. Total load time in Virginia with the same connection speed as in Paris is 24% longer, but rendering starts 2 seconds earlier (IE11).

AOL ­ Non­optimized site example

America Online is ranked number 157 globally by Alexa and number 48 in the US. The website only earns a mediocre 59 out of 100 point according to Google’s Page Speed Insights. How are users experiencing this Website? Is AOL that much slower for visitors than espn.go.com/nfl/? Let’s compare using the same test we did for ESPN:

aol-ny

aol-la

The West coast out performs the East coast yet again when it comes to load time. However Google Page Speed Insights gives the website a thumbs down. The first content is displayed to the user even earlier than on the higher ranked ESPN page (which is more than twice the size).

Now let’s test from Virginia using a slower connection:

aol-virginia

Now the same location and connection speed using IE11 instead of Firefox (by the way I re­ran this test twice because I couldn’t believe the results):

aol-virginia-ie11

It takes 53% longer in IE11 for content to begin rendering than it does in Firefox, though load time is relatively consistent.

Now let’s see how London performs for locals using a fast Internet connection and IE9:

aol-london

The lesson here is that a visitor from London with a speedy Internet connection (Cable) and IE9 sees content in half the time (Start Render time) compared to a visitor from Virginia who has a slower connection speed (3G) and IE11 browser. The London visitor sees content at about the same time as a Firefox user in Virginia using a 3G connection.

When is content first displayed and the page fully loaded in the case of a visitor from London using Chrome?

aol-london-chrome

Ouch! With this configuration load time feels like ages for the user because page load time takes 9.5 seconds before content renders!

Nothing feels more real than the real world

I’m sure that by now you get the point: How your users experience your website isn’t determined only by how slick your code is, how powerful your servers are, or how many likes you get from Google PageSpeed Insights. Geolocation, connection speed (last mile), device type, and browser type are critical elements of real world Performance. Therefore, to judge the performance of your service, you absolutely need to know what is going on in your visitors’ browsers. After all, your users will judge the quality of your website based on what they actually see in their browsers, right?

With this in mind, consider the value of the Dynatrace Apdex map that shows you how well your site is performing worldwide by country and even by region.

With Dynatrace you no longer need to guess as to the effect that geolocation, connection speed, and browser type have on the performance of your site.

This is just the beginning of the performance insights that Dynatrace gives you. For example, you can look at the performance experienced by ALL users across ALL user actions, including AJAX requests. With Dynatrace you can answer questions like, how long does it take a user from Germany using Safari to log in to his account page using a 3G connection?

Are you able to answer such questions with your current APM solution?

At Dynatrace we’ve personally benefited from the insights provided by Dynatrace. Before Dynatrace went to market in Europe, we assumed that we needed a host in Europe to serve our website. Thanks to Dynatrace real user monitoring we discovered that our performance was even better in Europe than it was in the US where the server was located. We saved a lot of time and money thanks to that insight.

Dynatrace makes it easy to connect the dots and not only keep an eye on server health, network performance, and service response times, but also the often decisive last mile to the user’s browser. When you install OneAgent, real user monitoring happens automatically, showing you your complete performance picture. You not only get data related to speed, you also see failure rates and the impact of 3rd party components. And if you don’t have write access to your root web server you can manually insert a JavaScript tag into your page and begin monitoring immediately.

50k visits for free! No contract obligations! Are you ready to take a “real” look at your site’s performance?