Split-Testing FB Ads Question

jetto

jetto

STRICTLY BUSINESS
Feb 20, 2010
703
5
0
Do you guys think giving 1 ad 50,000 impressions will produce a more or less reliable indication of its CTR?
 


If you spread the impressions across the day, it'll give you a better representation because CTRs can vary quite a bit throughout the day.
 
I don't think this question is so simple. It seems that whenever I throw another 10k imps at an ad, CTR can drastically change.
 
To get an idea of how the CTR will perform in the long run I would say almost run it a day, but usually if I'm just split testing ads to see which is a winner I do ~10K impressions depending on the CTRs.
 
50k should more than be enough to gauge your CTR.

Anything over 10k impressions should be sufficient (though could be skewed due to certain offers performing better at different times during the day, ie dating campaigns not performing well in the AM hours).
 
DanGlobalizer said:
50k should more than be enough to gauge your CTR.

Anything over 10k impressions should be sufficient (though could be skewed due to certain offers performing better at different times during the day, ie dating campaigns not performing well in the AM hours).
Click to expand...

What are you guys all talking about? If you have 10k imps on 2 ads, one with a 0.03% CTR and one with a 0.06% CTR you have barely an 85% confidence interval. Same confidence would be for a 0.10% CTR vs. a 0.15% CTR.

This is dumb advice. Learn how to calculate statistical significance instead of using weird rules of thumb.
 
over9000small.jpg
 
You must log in or register to reply here.
Top Bottom