LCP is much higher with tool than for users

Hello all! We recently ran a test and found the LCP for our web page with the tool is reporting at 56 sec! Looking at the waterfall, we found the culprit to be several images that are taking > 30 sec to load and are all < 1 mb. The most extreme example is a 700 kb image taking 50 sec. The interesting thing is that when loading the web page from a fresh browser, all the images load in < 8 sec.

Is there some discrepancy when using the tool that I’m not aware of? Some performance piece when serving images to bots that I’m missing? I’m hoping someone can give me a pointer as to what could be causing the difference in experience vs reported time.

Hey Caleb! Do you have a test result URL you can share?

Yup! It’s here: WebPageTest Test Result - Virginia USA - E...tore.steelcase.com - 06/29/21 15:52:40

Thanks!

It looks like there’s a bunch of resources getting queued up at once, and then competing for bandwidth. The connection view makes that a little easier to see.

That’s how it more or less should work, so that by itself isn’t the issue. BUT, several of those images are very large. The annotated screenshot below only calls out the images that are over 1MB.

As a result, you can see the page load process is using up all the available bandwidth for pretty much the entirety of the page load process.

It looks like several of the images are significantly larger (in terms of height/width dimensions) than necessary, so if you’re able to provide them in a smaller size you should see some improvements here.

Thank you for the info! We are actively looking to better size these images to improve the performance of the page. I missed some of those larger images when reviewing the results but I think I follow your points. The 5000 kbps bandwidth would explain some of the difference in the tool’s load time compared to our test users’ loads.

I’m still surprised that the total time is so long given the total load size of 30,838 KB. I would have thought that this would still result in <10 sec load time through the tool. Not our goal necessarily but it still seems like 50 sec is especially slow.

30 Mega Bytes at 5 Mega BITS per second is (30 * 8) / 5 = 48 seconds. That’s the theorhetical fastest those bytes could be delivered with no ramp-up or delays.

30MB is Huge for a web page.

:man_facepalming: I should’ve realized that, that makes total sense. We should have some reduced images soon and see some improvements but those numbers lining up really helps us consider our resource sizes more carefully. Thanks again for your help, I really appreciate it :slight_smile: