When testing ad variations, how many impressions you guys typically like to see before disabling lesser performing ads?
Off Topic: I searched google for this topic, and this thread was already listed #1 in the results, literally 3 minutes after I posted - nuts.
Wow...1st result and there's nobody using this keyword advertising in my country...lol Ok...For me i look at number of clicks instead of number of impressions for two ad variations in an adgroup. Generally after 100 clicks then i compared the CTR of each ad variation and make a decision.
I give it around 30 clicks and then make a decision. I first look for ROI/Conversions and choose the one that's converting the best. If it's even I then look for CTR. Sometimes Ad A or B has both higher conversions AND a better CTR, so in this instance it's obvious as to which to drop and which to keep. ROI is more important than CTR.
I wait for the difference to be statistically significant. To check, use something like this: http://splittester.com/
I've seen this tool before. What % are you looking for it to be around, in order to be confident the 2 ads have had enough?
for text ad testing I think you need a combination of sufficient impressions, clicks and days running a text ad. Running a text ad for 30 clicks or as much as 200 clicks in a short period of time can result in inaccurate analysis. Suppose you ran an ad on Monday, get 100 clicks by Tuesday and determine which as is best. This is bad, because suppose you continue to run the text ad for an additional day and Wednesday makes the other text ad better. I think running an ad for 2 weeks time period and not during a holiday or special event is the most accurate way to collect enough significant data to make an truly educated decision on which ad performs best. Since variables such as time of day and day of week can be very important, I recommend letting an ad run for 2 weeks time period to establish enough data. Otherwise, you could be just shooting yourself in the foot by deleting the wrong text ad.
For advert testing, I think 90% is plenty. if the chances that A is better than B are 90%, then what are the chances that selecting A will give you a substantially worse CTR (or conversion rate, if that's what you're checking) than B? A is better 90% of the time, B is better 1% of the time, and 9% of the time the two are virtually the same... I'm sure there are others who would disagree...