Hi @Liam-Shopify - any progress/insight on this issue? We are now at the 30 day mark until we loose our BFS status, so if it is not resolved shortly it is likely our average many not get back below the 2.5s mark.
I would note that just like Sam_Smith, we also reverted a bunch of code over the last 2-3 weeks to try and address this prior to posting here and could see locally LCP scores btw 1.2-1.8s, but zero impact on the actual score. It almost looks to be somehow taking into account the actual load time of the shopify admin page (if that is possible?). Hopefully you guys are close to finding the root cause here.
I’m not sure this is contributing (or has even changed), but I was looking at the app-bridge.js CDN script and seeing how long that might load from various regions in the world.
The only thing that stood out to me was the Google Distributed Cloud header “X-Dc” which had a value of “gcp-us-central1,gcp-us-east1” regardless of where the request was initiated from.
That seemed unusual to me, but I’m not as familiar with GCP’s infrastructure. I don’t want to introduce any red herrings, but I know we faced a similar issue with our apps early on where the locality of our original CDN was not optimized and that made a big difference in our LCP scores once we corrected that.
Hi @Liam-Shopify - any updates on this issue? Just checking in as it has been 2 weeks since the last update and this is now a really urgent one for us as we are 12 days away from loosing our Built for Shopify status, with no way to rectify it.
Hi folks - we’re still investigating this, and trying to identify what caused the spike that hasn’t readjusted. If you’re experiencing this issue and have gotten an “at risk” email from Shopify please contact the BFS review team or support and reference this forum post.
We see the same thing, and no longer meet our LCP requirement. It seems like LCP just isn’t reported if the number that would be reported is too small (less than 210ms or so). Above 210ms or so, it seems to be pretty reliably reported. So if we’re “too fast,” it seems like we get penalized, and they only look at slow requests. Just an observation.
Last week we filed an issue on the web-vitals package about the problem with small LCP values not being reported. Could be a Chrome bug.
Hi Liam, Do you have any information? Our application is only 30 days away from losing BFS. When BFS spiked, we did not change the code or intervene in anything. But it spiked until now. And even now, in the past week, CLS has started to be unstable again. We do not know what Shopify has changed. Many people have this problem, so I think it is a problem from Shopify. Hope to receive feedback from you soon.
Hello everyone, we’re looking into this and have paused the assessment. If you see a change in trend in the next 1-2 weeks, would love to hear from the group. Thanks a lot for sharing app context, we’re using this info to come up with solutions.
Thanks @MarkT. As another data point, I switched to managed installs a week ago and that made a huge improvement for my LCP score (see below). I wouldn’t think managed installs would make much difference since my app is loading everything else the same, assets, images and fonts etc. Could that be an indication the LCP timing is including other things like the loading of the outer Shopify resources?
Shopify managed installations is where Shopify handles asking for the scopes so there is no OAuth redirect sequence. You may already be using managed installs, which means you’d have to find other efficiencies like a CDN, optimized font/image loading etc.
In my case, switching from OAuth to managed installs helps because the OAuth reloads the entire Shopify admin which impacts the LCP timings for the embedded app iframe. Here’s the docs for managed installs: Enable Shopify-managed installations for your app
Is there any update on changes to LCP analysis? We’re seeing LCP hover at 2.5s range since early July and app bridge continues to report around that range. Our other tools (Newrelic) report lower numbers - difference is at least 400ms. I know the assessment of it towards BFS is paused, but I’d love to know if we should expect anything in terms of next steps.
All: Thanks for checking in on this. We are exploring different recommendations (see How to Improve Admin Performance) but the solution highly depends on how you’ve built the app. Curious to know if you are looking into other tools or drilling down on usage stats?
@sandy-stoq for New Relic, are those numbers rolled up into a single view or did you do any segmentation? Also interested to know how they are detecting LCP inside the iframe. Thanks for sharing this.
Just wanted to chime in — we’re seeing this in our app too. Has anyone heard any updates on it?