Picking the better ad for content network: ctr vs. exposure

seeing something that shouldn't be happening

[bcc1234]

I have two ads where one has higher CTR while the other one is getting higher percentage served.

Wouldn't the system pick the better performing ad (better ad rank) and serve it more often? And isn't the CTR considered the most important factor of ad rank?

I'm seeing it both with "rotate" and "optimize" settings.

I've ran into something similar in the past, but not when I set the campaign to rotate the ads.

Any suggestions on which ad to keep?

On one hand, keeping the one that gets served more frequently would be better. It brings in more clicks (even with lower ctr). For whatever reason, G's algo seems to think it's better.

On the other hand, maybe the algo is whack and I should just pause/delete the lower ctr ad and hope that the ad with higher ctr will take over and not miss out on any impressions that were previously only available to the lower ctr ad?

Any suggestions?

[eWhisper]

Just to make sure...

The percentage of time an ad is served is coordinated to the time frame you are currently viewing.

If one ad has been there for a month, and the other ad has been there for a week, and you are looking at at month of data - then the ad that's been there for a month probably has a higher percentage of displays just due to time.

It's best to make sure that when you look at percentage served, you do so in a timeframe that only includes when both ads were approved and running.

[bcc1234]

eWhisper, yep I know that.

I'm looking at the time frames when both ads have been running.

As a matter of fact, I gave both ads 3 weeks to "battle it out." And still, the one with lower CTR is getting a lot more impressions (and overall clicks).

[eWhisper]

One more check - are these ads search only or also on content?

[bcc1234]

Content only.

This is really weird.

Now, with "even rotation" one ad is getting 98% served. And it's the one with low ctr.


I understand about "participating in the auction" is not the same as "getting an impression". But this is beyond strange.

[eWhisper]

The fact this is a content only campaign might be the answer.

Google looks at the CTR of of an ad on a specific site or similar types of sites when determining which ad to show. They do not just look at the overall CTR of an ad (they also do things like how closely an ad/keyword is to the actual site in determining content QS).

My guess is (and I could be wrong) that the ad with the lower performance is doing better on some high traffic sites so its being shown more often on a handful of placements while the other ad is doing better on the rest of the sties but in aggregate they have less impressions.

The best way to analyze this is to run an ad report that includes placements/keywords as a data point and shows only content data.

That should give you insight into which sites are showing your ads and the CTR on each site for each ad.

[RhinoFish]

Imagine there are only two content websites that your ad gets triggered for...

site #1:
ad #1 = 200 imps, 3 clicks, 1.5% ctr
ad #2 = 300 imps, 7 clicks, 2.3% ctr
G shows ad #2 more...

site #2:
ad #1 = 1000 imps, 9 clicks, 0.9% ctr
ad #2 = 1400 imps, 8 clicks, 0.6% ctr
G shows ad #1 more...

aggregate stats reported to you:
ad #1 = 1200 imps, 12 clicks, 1.0% ctr
ad #2 = 1700 imps, 15 clicks, 0.9% ctr

ad #1 is served 1200/(1200+1700) = 41%
ad #2 is served 1700/(1200+1700) = 59%

the ad with the lower ctr is being served more...

and all is well with the world.

:-)

[RhinoFish]

For content campaigns, load up on choices G can serve and let it do its magic automatically.

[eWhisper]

Thanks for doing the math.

I started but there wasn't enough coffee in me to finish :)

[bcc1234]

The best way to analyze this is to run an ad report that includes placements/keywords as a data point and shows only content data.

That should give you insight into which sites are showing your ads and the CTR on each site for each ad.

Hmm. Is there a way to get a report that includes both ads and placements? I was looking for something like that for a while, but it doesn't seem like any of the available reports actually offer that. It's either placements or ads.

[bcc1234]

For content campaigns, load up on choices G can serve and let it do its magic automatically.

I would love to. But I want to continue testing and don't know which ad to make my new control.

[bcc1234]

Google looks at the CTR of of an ad on a specific site or similar types of sites when determining which ad to show.

The ads are identical except for one detail. One of them has title enclosed in quote marks. So as far as relevance, theme, etc, they should be exactly the same.

[bcc1234]

site #2:
ad #1 = 1000 imps, 9 clicks, 0.9% ctr
ad #2 = 1400 imps, 8 clicks, 0.6% ctr
G shows ad #1 more...

Did you mean to say "#2 more"?

If that's the case, then the question is why would G do that.
If ad#1 has higher ctr on that particular site, why would they keep showing ad#2 more often?