We have to check monthly reports for server response time on Analytics, and I noticed something that makes very little sense to me.

When looking at a monthly report for last month (April 2017), we get a server response time of 2.10 secs, due to a spike from a single user. When looking at that same month on the spectrum of the site's entire life (from Jan 2014 to May 2017) April 2017 is displayed as 1.17secs.

In fact, looking at a 4 year spectrum, the values look different all around.

What exactly should I trust on the server response time calculation? Can anyone explain to me how they calculate these values and why they would be different depending on the time period we extrapolate them from?

  • I used to be in the SLA game for a global Telecom. There are some things you have to consider. One is that you cannot control the Internet or the responsive nature if any request made. That is simply beyond your control. Instead, a sites response time can only be measured on the server itself - from when the request is received and then responded to. You need to measure your metrics that way only and write your SLA accordingly. Otherwise, expectation of response times to remain predictable across the net is wildly unrealistic.
    – closetnoc
    May 12, 2017 at 3:25
  • Thank you, @closetnoc! I think I am going to use this response of yours to put my case up to their SLA team and try to get a new solution for this. May 16, 2017 at 10:37
  • I thought that this might be due to data sampling. Google Analytics doesn't use all the data to show stats. It may base stats on a certain percentage of visitors. Jun 12, 2017 at 8:30

1 Answer 1


The problem with the Analytics server response metrics is they are very subjective.

The best thing about the Analytics server response metrics is they are very subjective.

These measurements are from users with different internet connections in different places, if your user-base is small then expect wild variation when someone from very far away on a satellite connection connects vs someone local on fibre. This isn't just measuring the time it takes for your server to wake up and send HTML - it's the time it takes for that initial packet to get there.

This could be the sign of an issue - or it could not. As long as things seem healthy from your end, don't worry too much. If you are running your own server infrastructure or using a small provider then I would encourage you to do a geographical breakdown just in case there are network issues.

  • The issue here is we have SLAs standing with clients that force us to be under a certain response time. If I check from different time spectrums and I get different values, we are risking our responses to said clients. May 11, 2017 at 9:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.