.. you decide that an ad is shit and should be scrapped? or kept. Curious about this mostly regarding search PPC. I've been deciding around 1k impressions but I'm wondering if other people take bigger/smaller samples?
How many impressions before..
- Thread starter TylerL
- Start date
- Status
Depends on search volume and whether you're running on content/search.
You should always be split testing and if one definitely the the loser, then drop it
use http://www.splittestaccelerator.com/freetool.php to see if you're stastcially relevant
Also, if you want a full range of data, then run it for a week and see if you have enough to make a conclusive analysis, ie you can definitely see paterns in different times of the day consistently.
Then start testing different days of the week etc..
You should always be split testing and if one definitely the the loser, then drop it
use http://www.splittestaccelerator.com/freetool.php to see if you're stastcially relevant
Also, if you want a full range of data, then run it for a week and see if you have enough to make a conclusive analysis, ie you can definitely see paterns in different times of the day consistently.
Then start testing different days of the week etc..
Actually 1k impressions might be more than enough. If you know that a certain kw gets 10% ctr on avg, and you have an ad thats gotten 1000 impressions with only 10-20 clicks, then you can definitely say you have a loser.
It's not an absolute dependance on number of impressions. It;s always a comparison that let's you know if you're statistically relevant
This is completely wrong. 10-20 clicks isn't even CLOSE to statistically relevant to ANYTHING, at all.
I always let ads run for at least two days, usually one week to see if one day is better than the rest, and I wait for at least 300 clicks.
You can't base statistical relavance on absolute numbers.
If you have an ad running with a 10% ctr and 100 clicks with 1000 impressions and another one at the same time that also got 1000 impressions, but only 10-20 clicks, then you can absolutely say that it is statistically relavant that they will have long term difference.
Basing statisticaly relavance on number of clicks is somethign I see people do a lot, and that's not right at all.. I can have 2 ads running:
700 clicks ; 2mill impressions
vs 750 clicks; 2 mill impressions and that wont tell me anything
but a
100 clicks; 1000 impressions
vs 20 clicks ; 100 impressions will
If you have an ad running with a 10% ctr and 100 clicks with 1000 impressions and another one at the same time that also got 1000 impressions, but only 10-20 clicks, then you can absolutely say that it is statistically relavant that they will have long term difference.
Basing statisticaly relavance on number of clicks is somethign I see people do a lot, and that's not right at all.. I can have 2 ads running:
700 clicks ; 2mill impressions
vs 750 clicks; 2 mill impressions and that wont tell me anything
but a
100 clicks; 1000 impressions
vs 20 clicks ; 100 impressions will
- Status