When do you decide that an add is underperforming?

Trying to decide when to nuke adds...

[CernyM]

I have four different ads all running on the same ad group, and I'm trying to decide how much data is enough to determine when an ad is underperforming the rest.

The ad group I have generates only a modest number of impressions a day (2000 or so). How many impressions do I need to see before I can get a realistic opinion of how well the ads I've written are performing?

[olderscot]

It's a long time since I've done the maths but the number that you need to show will depend on what the CTR you're trying to validate is and what level of accuracy you want.

I would probably just make a practical decision and say that after a week you may have averaged out most of the variances and can make a reasonable decision (though if you've only got 10 clicks or so, then realise that the results are far from conclusive and it would be better to have something like 100).

Mike

[eWhisper]

I generally check ads once a week for the first month it's running. If an ad is underperforming, but I think it should do better, I'll let it run for another week (I've got many ads that one week does well, next week is x% lower, the week after best preforming, etc).
Sometimes, by leaving 4 different ads in a group, one will always do well, 2 average, 1 below average, but overally, your CTR will remain steady as those ads rotate in effectiveness.

If the ad continues to underpreform for the 2nd week, then I axe it, and create a new one to take it's place. The first month of a new ad group, I keep 4-8 ads running, until I can week it down to 2-5 best preforming ads, or until I get the CTR rate I'm looking for.

I then check ads after they have been running for a month, about every 2-4 weeks (depending on their KWs impressions/week). At the beginning of each month, I add at least one ad to a group, just to test it out. I ignore them for the first week until I'm sure they are being shown on the partner sites. Then let them run and see how they preform.

When you review ads, I also think it's imporntant to not just look at how the ad has done in the past week/this month, but also how it's preformed in earlier months. If the ad has a history of doing well, but is having an off month, I'm a little slower to completely delete it until it's had two consecutive bad months. Since I have other ads covering this ads lower CTR rate, my overall CTR doesn't change that much for one underperforming ad.

I think it's imporntant that an ad group either has:
1. A single ad that does very well, so well, that you just can't imagine a higher CTR, and leave it alone.
2. 2-4 ads that overall keep your CTR consistant as the ads ebb and flow over time, and develop new ads to help your CTR slowly raise until you hit your target CTR (or max ctr you think that kind of KW can produce).

And again, post sizes get away from me as I try to finish the first pot of coffee...hope this helps..

[widget]

One thing that's fun to do is throw up two identical ads and then watch the numbers for a while. Can be a real eye opener. And can make you less likely in general to axe too soon.