1432669 - Prefetching experiment

Status: RESOLVED INVALID

(Firefox :: New Tab Page, enhancement, P2)

Description

[Ed Lee :Mardak]

On hover, let's prefetch Top Site domains. Let's compare engagement A/B test.

Be careful for big files

Comment 1

[Florian Quèze [:florian]]

If we are not doing speculative connections there yet, we should. I'm not sure how much actually pre-fetching the page would help. I guess it depends how long the user hovers the tile before actually clicking it.

Comment 2

[Ed Lee :Mardak]

What are we trying to measure with the experiment? Introduce a pref to trigger prefetch on hover instead of on mousedown and use existing engagement metrics?

Comment 3

[Tim Spurway [:tspurway]]

Ultimately, we would hope to see more engaged users, though I think it is difficult to draw any conclusions about the success of the experiment in the likely scenario that there is no significant difference in engagement.  

Perhaps this doesn't have to be a Shield experiment.  We could do A/B video screengrab comparison of the two cases and evaluate the success criteria on the basis of a perceived performance improvement.

Comment 4

[Ed Lee :Mardak]

dmose, you've mentioned some prefetching in the past. Was that <link rel=prefetch> ? If I open the Browser Toolbox, I don't see any network activity when adding those <link>. I do see ocsp requests when I mouse down on various <a> tags though.

Comment 5

[Tim Spurway [:tspurway]]

we decided not to do this