If a variation’s page has a unique URL then you can analyze experiments per page basis and sample pollution shouldn’t be a problem. What do you think?
How sample pollution could be a problem when you use a client-side redirect with 2 separate page URLs in an A/B test?
@ux.simply Good question. Let me tag @whoiseddie here for an answer.
So you’re saying the B variation is a new unique URL vs the control? So essentially a redirect test correct?
If so, I would also make sure that the new page is not findable through search, and is only able to be entered through the experiment. Once you confirmed that, then yes, sample pollution should not be a problem. The only way I can see some kind of sample pollution is with ITP, or cross device visits (where a user will send the link to their phones and contribute to “direct visits”).
@whoiseddie Could you resolve this by filtering B variation’s results by source (redirect page only)? Or would that add extra complications?
Yes, that is correct! The traffic comes to the original page from ads only. Direct and Organic accounts only for 1-2% of conversions. So sample pollution shouldn’t be an issue in this case. Right?
I am not sure how will it work, since source is transferred when there is a redirect.
You mean referrer? Believe so.
You’ll just need to make sure you pass the query along or select to pass the queries along to the new redirect page so you can attribute correctly. Queries do get dropped in a redirect that doesn’t specify to pass the queries.
Yes! That’s what I meant. Apologies – it was a little late at night when I asked the question.