Hi - We’ve just started trialling a prerender on our landing pages, to get the JS and CSS for our product availability page before the customer does a search.
In WPT the cost of the Prerender is being added to our homepage. Our JS has doubled, Start Render delayed by 2s, overall page weight has doubled. This leads to a false negative. We put this out in a rush last week, and I’m glad we did because I’d not have done so if I’d looked at WPT.
When I look at Chrome dev tools it is not, it acts according to the spec, as if being opened in another window. We use mPulse for measuring real users and have seen a real benefit from this on the availability page, in our custom first render marker and other timings, and no negative effect on the homepage.
Why would the prerender be affecting Start Render time, adding to the overall page weight when it is (supposed to be) executed after the initial page has loaded and in a separate window?
We are separately working on making these files smaller, but that is a lot more difficult.
Many thanks
Ben
Ben, can you post a link to an example?
Hi - yes of course, bit of an omission.
So Request 32 to v2.holidayextras.co.uk is where the prerender kicks off. I don’t see these requests in dev tools, and I wouldn’t expect them to contribute to page load.
WPT measures all browser network activity while Dev Tools only reports activity for the tab being debugged. The prerender basically loads the page in a hidden tab and WPT is showing how that activity competes with your main page load activity. It doesn’t directly impact the page load completing (i.e. onload doesn’t wait for any of the resources) but there is contention for CPU, bandwidth and the socket connection pools to the various servers (6 requests per fqdn across all tabs).
It may be worth trying to experiment with the timing of the prerender (maybe inject the link tag after onload or whenever you consider the page loaded) so you still get the benefits but without impacting the current page.
Thanks. The separate tab is how I thought it worked, so that’s put my mind at rest. We can’t download all that stuff for free, absolutely. I’m very keen to make sure we don’t adversely affect customers on the landing page but not sure how to measure it.
I’ll try your suggestion with the link tag, thanks.