Will Kahn-Greene [:willkg] ET needinfo? me Assignee
	Description • 10 years ago

We want to switch from happy/sad to a 1-5 gradient rating.

First thing we need to do is add a new field to the Response model. This new field should be an integer ranging from 1-5 inclusive.

Further, on save:

   If there's a happy/sad and a 1-5, then we'll save both to the db.

   If there's a happy/sad, but no 1-5, then we'll save happy=2 and sad=4 
   (i.e. not the extremes).

   If there's a 1-5 but no happy/sad, then we'll save 1-2=happy, 3-5=sad.


We should also add tests for the above to make sure it's correct.

One thing I'm not sure about is whether we want to log that we coalesced the values somewhere so we know which values were coalesced. E.g. if the feedback has happy/sad and we generated the gradient from that.

	Will Kahn-Greene [:willkg] ET needinfo? me Assignee
	Comment 1 • 10 years ago

Putting this in 2014q2 for now.

	Will Kahn-Greene [:willkg] ET needinfo? me Assignee
	Comment 2 • 10 years ago

Bumping this to 2014q3.

	Will Kahn-Greene [:willkg] ET needinfo? me Assignee
	Comment 3 • 10 years ago

Making this a P2. This is important to do this quarter.

	Will Kahn-Greene [:willkg] ET needinfo? me Assignee
	Comment 4 • 10 years ago

I'm changing my mind on how this should get implemented. We should do this:

1. add a new integer field to the Response model (and thus the feedback_response table)

2. on save:

   2.1. if there's a gradient, but no happy/sad, generate the happy/sad
        1-2=happy, 3-5=sad.

3. add tests to make sure all of the above works


The difference here is two fold. First, I put it in an itemized lists. Lists ftw. Second, I ditched the "if there's a happy/sad, create the equivalent gradient". Why? Because this first stage is A/B testing where we need to understand the relationship between happy/sad and gradients and in order to do that, we need to know whether the user saw an A or a B and this lets us do that trivially.

We can change how this works later once we're past the A/B phase.

	Will Kahn-Greene [:willkg] ET needinfo? me Assignee
	Comment 5 • 10 years ago

Oops... I somehow inverted the numbers.

   1 ................. 5
very sad           very happy

Thus the gradient -> happy/sad is 1,2,3=sad 4,5=happy.

	Will Kahn-Greene [:willkg] ET needinfo? me Assignee
	Comment 6 • 10 years ago

In a PR: https://github.com/mozilla/fjord/pull/363

	Will Kahn-Greene [:willkg] ET needinfo? me Assignee
	Comment 7 • 10 years ago

Landed in master: https://github.com/mozilla/fjord/commit/a49f1eb914aef3e924b35a5e038f73fd8ebf41c3

	Will Kahn-Greene [:willkg] ET needinfo? me Assignee
	Comment 8 • 10 years ago

Pushed to production just now.