Add use counters for persistent notification interaction.

Add use counters for persistent notification interaction.

BUG=

Committed: https://crrev.com/f52d070a3cae53f27cf9490428770b64e73c16d0
Cr-Commit-Position: refs/heads/master@{#352025}

[Peter Beverloo, 2015-09-01 14:45:07]

peter@chromium.org changed reviewers:
+ isherman@chromium.org, miguelg@chromium.org

[Peter Beverloo, 2015-09-01 14:45:07]

+miguelg, isherman, PTAL

[Miguel Garcia, 2015-09-01 15:18:20]

lgtm

[Ilya Sherman, 2015-09-01 23:02:53]

Have you considered using a histogram to track this, rather than user actions? 
It's ok either way; but for most people, using histograms makes analysis easier.

https://codereview.chromium.org/1315193011/diff/1/tools/metrics/actions/actio...
File tools/metrics/actions/actions.xml (right):

https://codereview.chromium.org/1315193011/diff/1/tools/metrics/actions/actio...
tools/metrics/actions/actions.xml:8597: does not close the notification on
click.
This sentence confuses me.  If a website does close the notification on click,
do both actions get recorded, or just one of them?

[Peter Beverloo, 2015-09-02 12:48:06]

Thank you for the look.

There are three UMA histograms which indirectly measure these cases as well:
Notifications.Database.WriteResult (show),
Notifications.PersistentWebNotificationClickResult (click) and
Notifications.PersistentNotificationDataDeleted (close), but they're not always
a 1:1 mapping, are confusing to read and we suspect that we're not getting all
the data due to either service worker errors, or being killed by the system
before they're recorded.

The goal here is to get a ballpoint mark on the CTR of notifications. Current
measurements vary between 3.6% and 24%, with ~50% of notifications seemingly
going to a black hole.

https://codereview.chromium.org/1315193011/diff/1/tools/metrics/actions/actio...
File tools/metrics/actions/actions.xml (right):

https://codereview.chromium.org/1315193011/diff/1/tools/metrics/actions/actio...
tools/metrics/actions/actions.xml:8597: does not close the notification on
click.
On 2015/09/01 23:02:53, Ilya Sherman wrote:
> This sentence confuses me.  If a website does close the notification on click,
> do both actions get recorded, or just one of them?

This is about user-initiated responses, so then only one would be recorded. I
have clarified the message.

[Ilya Sherman, 2015-09-03 00:50:43]

On 2015/09/02 12:48:06, Peter Beverloo wrote:
> Thank you for the look.
> 
> There are three UMA histograms which indirectly measure these cases as well:
> Notifications.Database.WriteResult (show),
> Notifications.PersistentWebNotificationClickResult (click) and
> Notifications.PersistentNotificationDataDeleted (close), but they're not
always
> a 1:1 mapping, are confusing to read and we suspect that we're not getting all
> the data due to either service worker errors, or being killed by the system
> before they're recorded.
> 
> The goal here is to get a ballpoint mark on the CTR of notifications. Current
> measurements vary between 3.6% and 24%, with ~50% of notifications seemingly
> going to a black hole.

To be more precise, my suggestion is to create a single new enumerated histogram
that has one bucket per user action that you're currently recording.  For a
similar example, see https://goto.google.com/autofill-histogram-example

That said, LGTM if you prefer actions to histograms.  Thanks for clarifying the
comment.

[Peter Beverloo, 2015-10-02 13:28:06]

On 2015/09/03 00:50:43, Ilya Sherman wrote:
> On 2015/09/02 12:48:06, Peter Beverloo wrote:
> > Thank you for the look.
> > 
> > There are three UMA histograms which indirectly measure these cases as well:
> > Notifications.Database.WriteResult (show),
> > Notifications.PersistentWebNotificationClickResult (click) and
> > Notifications.PersistentNotificationDataDeleted (close), but they're not
> always
> > a 1:1 mapping, are confusing to read and we suspect that we're not getting
all
> > the data due to either service worker errors, or being killed by the system
> > before they're recorded.
> > 
> > The goal here is to get a ballpoint mark on the CTR of notifications.
Current
> > measurements vary between 3.6% and 24%, with ~50% of notifications seemingly
> > going to a black hole.
> 
> To be more precise, my suggestion is to create a single new enumerated
histogram
> that has one bucket per user action that you're currently recording.  For a
> similar example, see https://goto.google.com/autofill-histogram-example
> 
> That said, LGTM if you prefer actions to histograms.  Thanks for clarifying
the
> comment.

I honestly cannot decide, so I'm going to stick with this. We'll spend time next
quarter to significantly improve the data we're gathering around (push)
notifications, since it's lacking in more areas.

[Peter Beverloo, 2015-10-02 13:28:14]

The CQ bit was checked by peter@chromium.org

[Peter Beverloo, 2015-10-02 13:28:14]

The patchset sent to the CQ was uploaded after l-g-t-m from
miguelg@chromium.org, isherman@chromium.org
Link to the patchset: https://codereview.chromium.org/1315193011/#ps60001
(title: "rebase again")

[commit-bot: I haz the power, 2015-10-02 13:29:29]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1315193011/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1315193011/60001

[commit-bot: I haz the power, 2015-10-02 14:13:40]

Committed patchset #4 (id:60001)

[commit-bot: I haz the power, 2015-10-02 14:14:28]

Patchset 4 (id:??) landed as
https://crrev.com/f52d070a3cae53f27cf9490428770b64e73c16d0
Cr-Commit-Position: refs/heads/master@{#352025}