AB testing

1. Does a new supplement help people sleep better?

For this kind of testing, a placebo group would be appropriate. 
The control group will be placebo group and the group taking actual 
sleeping pill will be test group. 

Sample must be randomized and as similar as possible. Example, a group of middle aged male and females equal in number who 
suffer have a clinical diagnosis of insomnia.

Our hypothesis is that the supplement will help people sleep better. 

We'd expect changes in sleep cycle, longer duration of sleep, and reaching REM level of sleep. 
We can measure it in a sleep lab or give patients a pre and post survey. There will be bias for both types of measurement,
we just have to decide which will cause less bias. 

Other viarables we can measure is how long did it take for patients to fall asleep(in a sleep lab setting)
Quality of sleep. Duration of sleep. 
______________________________________________________________________________-
2. Will new uniforms help a gym's business?

Get stats/info from old uniform days (maybe at least a month) and compare it to the new uniform days (same amount of time).

Sample group will be gym members. 

We should determine how the uniforms "help" a gym's busines. It should be a measurable outcome such as more gym membership 
or maybe more use of their trainers, which require a member to pay an extra fee. In this instance, let's just say that our
hypothesis will be that new uniforms increases gym membership.

We would like to measure our outcome of interest by determining whether there was increase in gym membership. 

Other outcomes could be that maybe current members would use gym more frequently, not necessarily new members. 
__________________________________________________________________________________
3. Will a new homepage improve my online exotic pet rental business?

We can look at clicks on the new homepage. We can also look at sales/profit after homepage was improved and compare from previous.

Sample groups are customers clicking at the website. 

Hypothesis can  be: My new homepage will result in an increase amount of sales/profit. 

Outcome of interest : We can expect increase in profit due to improvement in website. We can measure by amount of sales.

Other outcome: Other than sales, we should measure the amount of exposure the new website brings. We can do this by looking 
at clicks. 


________________________________________________________________________________________

4.If I put 'please read' in the email subject will more people read my emails?

We can send emails at random with A:regular email and B:"please" read email subject line.

This one is tricky because it will be hard to see who opens the email, unless you get a response from the person. A way we can 
use to give incentives to reply, like a $10 amazon gift card, if they reply to your email.

Hypothesis: More people will read my email if I put "please read" in the subject line. 

Outcome of interest: Our key metric would be the amount of replies. 

Another outcome we can measure is how fast people will reply.