How to measure the impact of Fasterize

Modified on Tue, 02 Jul 2019 at 10:33 PM

You asked us, here it is: an article that explains in detail and simply how to test the effects of Fasterize


1. Estimate the Fasterize capital gain before connection 


The solutions that we will discuss in this part will allow you to have only an indication on the potential improvement of your site. 


To estimate the benefit that Fasterize would bring to your site, you can compare your site without Fasterize and with Fasterize (using the previous preview mode) on the following tools:



These 5 latest tools will allow you to test the technical metrics of your web pages (page weight, number of queries, etc.) and to check that web performance best practices are implemented on the fasterizée version.


For these tools, the ideal is to launch several tests (a dozen) and take the median results.


3. Evaluate the real technical and marketing impacts on your site


For that, no magic solution, no crystal ball. The only solution to know the real gain of our optimizations is to connect!


But do not panic, a month is offered to make your first impressions.


  1. plug in your site
  2. reassemble the information in Google Analytics
  3. analyze technical metrics (with Real User Monitoring (RUM) tools such as Basilic or Newrelic , Pingdom ) and business metrics in Google Analytics!


We automatically implement an AB test (90% optimized / 10% non-optimized) when you connect your site to Fasterize. This will allow you to compare the results of your site with and without optimization during the free trial period.

Some details, however: 


  • Sampling  : sampling or sampling consists of taking only a portion of the data points to establish measurements. In our case, it intervenes in two places:
    • for measuring load times in Google Analytics: Google measures loading times of just a few pages among all the pages viewed by your users. In the A / B test, you may have measures on a page for optimized users and no metrics on that page for non-optimized users, making the comparison obsolete.
    • if your audience is large, Google keeps only a fraction of the measures to perform its conversion rate calculations (the limit is 250000 sessions), which can lead to aberrations. The simplest in this case is to aggregate even its data. 
  • Loading time:  You should look at the median rather than the average when you analyze your loading time. Indeed, the median is not influenced by the extreme and minority values, whereas the average is strongly influenced by it. The median is therefore more representative of the reality of the majority of Internet users. You can also study the 80, 90 and 95 percentiles (the median is the 50th percentile). 
    For example: imagine that 9 people load a site in 5 seconds but one person loads the same site in 100 seconds (for a reason X). 
    The average is 14.5 seconds, while the median remains at 5 seconds.
  • The duration of the test: your test must be carried out over a sufficient period of time. This will stabilize and clearly identify populations:
    • new users receiving the optimized site;
    • new users receiving the non-optimized site;
    • the returning visitors receiving the optimized site;
    • the returning visitors receiving the non-optimized site.


At the start of the test, some visitors are placed in the "optimized" category while they had previous experience with a non-optimized version. It takes time for these populations to stabilize and be clearly separated.


This time is variable depending on the site being measured. It depends on the audience and the decision-making time before a user converts: we do not take the same time to think about buying a garment or a computer.


  • Sample size:  To know if your A / B test is relevant and therefore to determine if there is a correlation between the acceleration of the site and the increase of the conversions for example, you must do the test of the Khi2 . The question is whether there is enough data to verify that the A / B test is reliable and that the result is well correlated to the change that has been made (in our case, site optimization). This test is achievable through an online calculator .


Was this article helpful?

That’s Great!

Thank you for your feedback

Sorry! We couldn't be helpful

Thank you for your feedback

Let us know how can we improve this article!

Select atleast one of the reasons

Feedback sent

We appreciate your effort and will try to fix the article