Skip to main content
Cross Validated

Return to Question

try to broaden the scope of the question
Link
steffen
steffen
  • 10.5k
  • 2
  • 52
  • 77

How would youto conduct anam A/B test on an opt-infor a feature which cannot be accessed by every visitor?

Attempted to stem the terminal decline of the English language
Source Link
edit approved Apr 29, 2014 at 11:40
M. Berk
edit approved Apr 29, 2014 at 11:40
M. Berk
  • 2.6k
  • 2
  • 16
  • 21

I am prepping for an interview and here is one of the questions I have come across:

How would you conduct an A/B test on an opt-in feature?

I think you cannot perform a "fair" test with an opt-in option. Ppl People viewing the A version of the site will come from the whole distribution of users whereas pplpeople viewing B will only come from the set who have opted in. This seems to violate the conditions for a controlled experiment.

What is your take on this?

I am prepping for an interview and here is one of the questions I have come across:

How would you conduct an A/B test on an opt-in feature?

I think you cannot perform a "fair" test with an opt-in option. Ppl viewing the A version of the site will come from the whole distribution of users whereas ppl viewing B will only come from the set who have opted in. This seems to violate the conditions for a controlled experiment.

What is your take on this?

I am prepping for an interview and here is one of the questions I have come across:

How would you conduct an A/B test on an opt-in feature?

I think you cannot perform a "fair" test with an opt-in option. People viewing the A version of the site will come from the whole distribution of users whereas people viewing B will only come from the set who have opted in. This seems to violate the conditions for a controlled experiment.

What is your take on this?

Source Link
mchangun
mchangun
  • 509
  • 1
  • 4
  • 10

How would you conduct an A/B test on an opt-in feature?

I am prepping for an interview and here is one of the questions I have come across:

How would you conduct an A/B test on an opt-in feature?

I think you cannot perform a "fair" test with an opt-in option. Ppl viewing the A version of the site will come from the whole distribution of users whereas ppl viewing B will only come from the set who have opted in. This seems to violate the conditions for a controlled experiment.

What is your take on this?