The Reports of LCP is smalller than what Google Search Console reports

[Dante-dan]

Is there an existing issue for this?

• I have checked for existing issues https://github.com/getsentry/sentry-javascript/issues
• I have reviewed the documentation https://docs.sentry.io/
• I am using the latest SDK release https://github.com/getsentry/sentry-javascript/releases

How do you use Sentry?

Self-hosted/on-premise

Which SDK are you using? If you use the CDN bundles, please specify the exact bundle (e.g. bundle.tracing.min.js) in your SDK setup.

@sentry/vue

SDK Version

7.8.0

Framework Version

vue.2.6

Link to Sentry event

https://duanjl.sentry.io/dashboard/37879/?statsPeriod=24h

SDK Setup

No response

Steps to Reproduce

1. clone my demo project and deploy it.
2. we should be installed Cloudflare web analysis (The closest to GSC) and Sentry tracing at the same time , and visit it
3. Let's watch the final result.

Expected Result

1. the sentry lcp reports count should be equal with cloudflare web analysis lcp reports count;
2. the P75 LCP should closest.

Actual Result

1. sentry lcp reports count is smaller than cloudflare web analysis
2. lcponly report in pageload transaction. Because the measurements are only available for pageload transactions. This means that only the earlier LCPs are reported with pageload, while the slower LCPs often occur after pageload, resulting in much smaller LCPs being collected than what Google reports.

relate #3019

[lforst]

Thanks for writing in! You're pretty much spot on with your analysis that long LCPs occur after we finish pageload transactions and we can't send Web Vitals out of band. This causes the average to be skewed. It is a known problem.

We usually close issues like these because we know it's not gonna be fixed any time soon due to architectural limitations but we're going to leave this one up as a sort of aggregation issue since it explains the problem quite well. Maybe we will get to fixing this at some point but it will definitely be a larger effort that will span multiple teams.

[Dante-dan]

I may want to know why LCP has to be reported with pageload when it is an objective metric on its own.

[lforst]

We report metrics alongside transactions. The pageload transaction only lasts a maximum number of seconds and if LCP occurs after that we cannot report it unless we would send it with another transaction.

[Dante-dan]

Or, we could consider reporting LCP as a separate transaction, because from the perspective of browser rendering, page-load and LCP are not the same thing in the first place.

Alternatively, when reporting page-load, we could generate a hash on the client side, and then when reporting LCP, if there is a page-load-hash, include it. On the server side, we could merge these two transactions, thereby ensuring that the rendering chain is still available.

[lforst]

Yeah, there are quite a few obvious solutions to this, however, there are a lot of performance concerns when ingesting additional transactions and merging them. Keep in mind Sentry processes billions of events per week.

[AbhiPrasad]

We've made improvements to this in our latest release which should hopefully make the numbers more accurate. Please give it a try: https://github.com/getsentry/sentry-javascript/releases/tag/7.42.0