What is the best way to split test new ads when you have an ad that is an ok performer, but could definitely be improved? Most of what I've read recommends testing 2 ads at the same time, and giving each ad an equal number of impressions. This seems a little risky, since I don't know how the new ad will produce. It seems to me that it would be better to run the successful ad more (maybe like 80%) of the time to avoid losing too much if the new ad is unsuccessful. I realize the split test would take longer this way, but it still seems better. Also, is there a way to have one ad shown 80% of the time and another ad shown 20%? Google only gives the options of rotating ads evenly or using their optimizer. Any feedback is appreciated.
I wouldn't call it risky. Not taking the risk to try new ads that may work much better is worse in the long run than worrying about potentially losing traffic with a poorer ad. You only need a few hundred impressions anyway. You can specify to rotate the ads evenly or let Google decide which performs better (the optimizer). I don't know the percentages but it's probably about 80-20. You cannot set this percentage yourself. I prefer rotate since I'm always testing and like to compare a similar number of impressions. Look at it this way. If your ad gets a CTR of 1% in the next year over 120k impressions. That's 1200 clicks. You know that's what you'll get. But let's say you try new ads, say every 10k impressions or so, about one every month. You should be getting better the more you do. Say you make an ad that gets 2% CTR some day. You might have gotten ads getting 1.5% in the meantime but I'll use the original 1%. That would be 150 clicks that month split over the two ads or 1800 over the course of a year. That's 600 more, surely enough to cover the losers. Sure you'll have losers along the way. Just pause those (don't delete, you might want to try it again later, maybe a seasonal or other reason). But in the long run, your new best performer becomes your control. Even if you decide to stop testing, you've just increased your clicks. Are you willing to risk that? So set your ad serving to rotate. Create more ads and test two or three ads at a time, not more.
I discovered that I can make a new ad that is identical to the ad that is performing well. I made 2 clones of the winner of my first split test. I have all 3 copies of this ad active, and am running it against a new ad I am testing. With my ads set to rotate equally, the proven ad should get 75% of the impressions.
Merc, the reason to test new ads is to see if you can come up with something better. Yes, your original ad is the proven one. It has no choice, it was your ONLY ad. Remove your clones and run your original ad, what you call your proven ad, 50/50 with the new one. Wait until each ad gets 10 to 15 clicks. Then look at the data and remove the one with lower CTR. The one that's left is your winner. But don't stop there. Repeat the process with another ad. And then again. The problem that you and most advertisers is that you cling to the first ad you make and accept whatever click rate it gets thinking it can't get any better. In your case, you are afraid to lose clicks because you see it as the winner. I can guarantee you can do better which means that if you stand still, you ARE losing clicks by not testing or forcing one ad to get more impressions as you are doing.