feat: google analytics for drop-off statistics

[kylerwsm]

Problem

Make use of Google Analytics to measure drop-offs that might occur with the introduction of the transition page.

Closes #95.

Solution

I have made use of event logging in Google Analytics.

I named this event which is focused on TransitionPage as transition-page. There are three possible event actions in this Event:

1. served - Logged by the server when serving a transition page.
2. loaded - Logged by the user after rendering the transition page.
3. proceeded - Logged by the user when the proceed button is clicked.

GA logging on the TransitionPage is done using analytics.js gtag.js.

In this configuration, realtime statistics can be view under Reports > Realtime > Events.

Tested on:

• Local development
• Staging, on links
• Staging, on files

Improvements:

• The googleAnalytics directory is refactored into Services.

[JasonChong96]

Would having the client browser send the event to GA directly cause it to be blocked by ad blockers?

[kylerwsm]

Would having the client browser send the event to GA directly cause it to be blocked by ad blockers?

That is a good question. Through my manual tests, sending events to GA with my adblocker turned on is still working as intended. If you have tried it, is it also working as intended for you?

To slightly alleviate things, we do have an intermediate loaded event, which might give us insights as to whether the transition page is loading properly, or if events are not sending properly from the users side.

The alternative I can think of right now will be for the transition page to ping our server, so that our server can send the analytics instead. But that could double calls to our servers on every redirects.

@liangyuanruo @yong-jie do you have any inputs on this?

Edit: From my quick research, ad-blockers in general do not block GA. The exceptions are ghostery, and Firefox, which blocks all trackers by default.

[yong-jie]

Although the approach of routing our tracking requests through our server before sending to GA could solve this problem, I feel like it might be a bit too much development effort for the benefits it provides. Using this bypass might also be deemed unethical(?).

If we're concerned that firefox users could skew our data, would ignoring redirect events coming from firefox help?

[kylerwsm]

I agree on the unethical part. I believe we can keep my current implementation, to observe the proportion of users blocking GA first, from the difference in served and loaded events. We might only need a solution if the proportion of untracked users is huge. What do you think?

[yong-jie]

sounds good to me.