1471350 - Add "Top Frecent Sites" to AS Router targeting

Status: RESOLVED FIXED

(Firefox :: Messaging System, enhancement, P2)

Description

[Kate Hudson :k88hudson]

For our CFR Milestone 1, we will need a targeting parameter for checking if a certain site (or probably hostname) is in the user's set of top sites.

Perhaps this could be implemented with in, e.g. `"amazon.com" in topFrecentSites` where topFrecentSites is just an array of host names, or perhaps a more advanced matching pattern is needed (this depends on the exact specification of the URL matching rules).

See ASRouterTargeting.jsm for existing code and https://github.com/mozilla/mozjexl for documentation

Comment 1

[Ed Lee :Mardak]

One thing to be careful about is "top sites" could have defaults and pinned sites whereas "top frecent sites" should be those that the user most visited/recently. But then another thing to be careful about is the targeting intent is for sites that the user visits a lot, but a brand new profile could have a single visit make it become a "top frecent site," so an explicit frecency threshold / check would probably be required.

Comment 2

[Kate Hudson :k88hudson]

Notes from our meeting with Kamyar:

URLS
- For m1, the URL match pattern should look at hostname/TLD. We should be able to target multiple TLDs at once (e.g. amazon.com, amazon.ca, amazon.co.uk etc.). It sounds very similar to how we do tippy-top / top sites, so we might be able to use the same logic
- In the future we would like to look at multiple subdomains as well as the first part of the path (e.g. google.com/maps), however, this is not necessary for m1. We should be mindful of the privacy implications of this.
- Exact URL recipes will be provided by Shell/Scott; 

Thresholds
- Kamyar has some recommendations for thresholds as well as a meaning for "semi-fresh", which he will provide exact definitions of in the future; we should provide support for a specific minimum number of visits and a minimum recency, preferably in a way that can be exposed to/ changed in the targeting expression. We might be able to just use the frecency number from the NewTabUtils query for this.

Comment 3

[Adam Hillier :ahillier]

From discussion with Kate:

- If we're targeting multiple domains, we can't just add the frecency for the domains as then that frecency value may account for more than 10 visits. We also can't just average them, because the average will be lower than the maximum for a single domain, which doesn't make sense. In the Amazon case of locale-specific sites (amazon.com/.ca/.co.uk/etc.) it's probably a reasonable approximation for most users to test if any single domain crosses the frecency threshold, as almost all users will be predominently using just one of the geodomains, so we'll do that for milestone one.

- Should we target a domain if a user has blocked that domain from their topsites? What about in the case where they've blocked one but not all of the multiple domains we're targeting? For milestone one we will ignore blocked sites for domain targeting, and in the case of multiple domains just disregard the ones that have been blocked. However, it's something to reconsider in the future.

Comment 4

[Ed Lee :Mardak]

Frecency is also a per-page value (not per domain). Indeed, most users who have multiple related domains probably primarily visit one, so just checking the max frecency of the most-frecent page of each related domain is probably good enough. Adding the frecency of each matching top domain could work, but that's probably adding unnecessary complexity.

If we're using the data from TopSitesFeed frecentCache / getTopSites, the blocked pages get filtered out already.

Comment 5

[GitHub Bugzilla PR Linker]

Attachment: Link to GitHub pull-request: https://github.com/mozilla/activity-stream/pull/4243

Comment 6

[Adam Hillier :ahillier]

Attachment: Bug 1471350: Add Jexl transforms for ASRouter topFrecentSites targeting

Review commit: https://reviewboard.mozilla.org/r/256980/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/256980/

Comment 7

[Michael Cooper [:mythmon]]

Comment on attachment 8992073 [details]
Bug 1471350: Add Jexl transforms for ASRouter topFrecentSites targeting

https://reviewboard.mozilla.org/r/256980/#review263832

Comment 8

[Kate Hudson :k88hudson]

Comment on attachment 8992073 [details]
Bug 1471350: Add Jexl transforms for ASRouter topFrecentSites targeting

https://reviewboard.mozilla.org/r/256980/#review264430

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/721a89562597
Add Jexl transforms for ASRouter topFrecentSites targeting r=k88hudson,mythmon

Comment 10

[Eliza Balazs [:ebalazs_]]

https://hg.mozilla.org/mozilla-central/rev/721a89562597

Comment 11

[[github robot]]

Commit pushed to master at https://github.com/mozilla/activity-stream

https://github.com/mozilla/activity-stream/commit/69d99c90f0dab5389c36ea6a6b59887cf2feb0b7
Fix Bug 1471350: add topFrecentSites to ASRouterTargeting (#4243)

Comment 12

[Ed Lee :Mardak]

https://hg.mozilla.org/mozilla-central/rev/ea88051a288d