Comments on: Does Page Load Time Really Affect Bounce Rate?

By: James Finlay

James Finlay — Fri, 27 Jul 2018 00:22:36 +0000

Hello! This is a very interesting study. We find this data incredibly useful as we try to provide meaningful insight into our engineering efforts.

I’m wondering whether your analysis had included any investigation into the bounce rate between mobile & desktop devices. We’ve seen some analysis that mobile users are less patient than desktop users, but Akamai recently published a contradictory study. I would be very interested in reading what your analysis shows regarding bounce rate over time.

By: Felix Pecarski

Felix Pecarski — Tue, 23 Jan 2018 09:51:57 +0000

In reply to Cyrus Shepard.

Hi Cyrus,

Thanks. For this article, we defined bounce rate as a visitor landing on a page and then leaving the website without any further progress on the site in question. We collated data from 35,000+ e-commerce sites that we monitor over the last 6 months.

Yes, we did mean John – thanks for pointing it out. 🙂 We’ve edited that typo now.

By: Felix Pecarski

Felix Pecarski — Tue, 23 Jan 2018 09:39:58 +0000

In reply to Thorsten Funke. Fixed! Thanks for the correction. :)

By: Thorsten Funke

Thorsten Funke — Sun, 21 Jan 2018 18:00:52 +0000

It’s John not George Müller 😉

By: Cyrus Shepard

Cyrus Shepard — Sun, 21 Jan 2018 03:03:16 +0000

Great study. Very useful to show clients interested in speed performance improvement. A couple of questions, if you could answer. 1) What defines bounce rate, for the purpose of this experiment? (visiting only one page, performing a task, a set time on page, or a combination or something else entirely) 2) How many sites/visits were included in the study? And what type of sites?

(one clarification: I believe you meant John Mueller?)

By: Rick Viscomi

Rick Viscomi — Thu, 18 Jan 2018 21:06:06 +0000

Thanks for posting! I have a few questions about the data.

What do you define as page load time? It’s an overloaded term that could mean lots of things (onload, time to [consistently] interactive, first [contentful, meaningful] paint, etc).

Do you also have median page weight stats? Averages tend to be heavily skewed by outliers and are not always representative.

“Does A affect B” is a question of causality, which requires a fair amount of experimental rigor. These results certainly show a strong correlation, but I wouldn’t necessarily say bounce rate is *affected* by page load time.

Lastly, I’m curious if you’ve compared your stats against the HTTP Archive. Would be interesting to see a comparison of real user and synthetic data!