AdWords Tips

Rotating Ads vs. Optimizing Ads: Which Is Better?

March 15, 2010 Posted In: AdWords Tips Comments: 23

This is a guest post by Alan Mitchell. Alan Mitchell is a Brisbane PPC consultant specializing in highly granular long-tail PPC management. Follow him on Twitter: @alanmitchell.

One question which is regularly asked to any PPC consultant when managing Google AdWords PPC campaigns is whether ads should be set to rotate or optimize.

Google's default setting is "optimize," so if you have multiple ads in one ad group, your better performing ads (generally those with a higher CTR) will be shown more often. This might seem great -- you will automatically receive the maximum number of clicks for your ads. But the more experienced PPC advertisers out there will know that clicks are not generally considered a good measure of success. Instead, conversions -- sales, leads, sign-ups, downloads and other desired outcomes -- are generally considered better measures of PPC performance.

So to aid acute PPC advertisers like Shaun Livengood from PPC Without Pity, who like to optimize ads based on conversions, Google have provided another option in AdWords called "rotate." Rotating ads allows PPC advertisers to override Google's default "optimize" setting and force all ads in an ad group to be shown equally.

Since each ad in an ad group is given the same number of impressions, data-hungry PPC advertisers are provided with a plethora of unbiased comparative ad performance data they can work with. Ad performance analysis becomes considerably easier if all ads in an ad group have the same number of impressions.

adwords-performance-report

I too am extremely data-hungry and a big fan of post-click PPC analysis, so I have always tended to set ads to "rotate" to satisfy my craving of juicy PPC data.

But now I'm not too sure. Rotating ads could be detrimental to profitability.

Why?

Because by rotating ads, you are accepting a lower Quality Score and a higher average CPC in exchange for better analysis data.

Let me explain with an example.

Suppose you had two ads in an ad group (ad 1 and ad 2). Ads were set to rotate, so each were shown an equal 3,000 times over the month. Ad 1 was the best performing ad in terms of CTR (3.0%), delivering 90 clicks over the month. Ad 2, with a CTR of 1.0%, delivered only 45 clicks.

  • Ad 1: 3,000 impressions, 3.0% CTR, 90 clicks
  • Ad 2: 3,000 impressions, 1.5% CTR, 45 clicks

Now let's look at the totals for the ad group. From 6,000 impressions on both ads, 135 clicks were delivered at an average CTR of 2.25%.

Ad Group Total: 6,000 impressions, 2.25% CTR, 135 clicks

Now let's consider what would happen if the ads were set to optimize, where Google chooses which ads to show based on their CTR. We know that Ad 1 has a considerably higher CTR than Ad 2, so to increase their advertising revenue, Google would choose to show Ad 1 more often. Ad 1 gets 5,000 impressions; Ad 2 receives 1,000.

  • Ad 1: 5,000 impressions, 3.0% CTR, 150 clicks
  • Ad 2: 1,000 impressions, 1.5% CTR, 15 clicks

Look at what happens to click volume for the ad group. Since the ad with the higher CTR is being shown more often, more clicks are being delivered. For the 6,000 times both ads were shown, click volume has risen from 135 to 165 -- an increase of 22%.

Ad Group Total: 6,000 impressions, 2.75% CTR, 165 clicks

Average CTR for the ad group has increased from 2.25% to 2.75%. Setting the ads to "optimize" has delivered 22% extra clicks for no extra effort.

But as we pointed out earlier, clicks rarely matter -- it is conversion volume which is more often considered a measure of success. Those extra clicks may just be increasing my costs for no improvement in conversions! Aren't I better off deciding myself how often my ads get shown?

No.

The Importance of CTR

Google loves relevancy. They want their user experience to be as great as possible, and they reward PPC advertisers who provide highly relevant ads. They do so with a Quality Score metric, scored out of 10.

CTR is the most important measure of Google's Quality Score, so the higher your CTR, the more likely searchers (and Google) think your ad is highly relevant. The more relevant your ad, the higher your Quality Score.

All other things equal, a higher Quality Score means a lower average cost-per-click (CPC) price or a higher ad ranking.

So back to our analysis of rotate vs. optimize ad settings.

  • Rotate had a CTR of 2.25%, and delivered 135 clicks
  • Optimize had a CTR of 2.75%, and delivered 165 clicks

Since our optimized ads have a 22% higher CTR than our rotated ads, it is fair to say they will receive a higher Quality Score, perhaps an 8 instead of a 7. If ads were shown in the same average position, it is also fair to say that the optimized ads (with a higher Quality Score) will have lower cost-per-click prices, perhaps by 10%.

This means that by setting ads to rotate, you are accepting a lower CTR, lower Quality Score and higher CPCs for better quality data.

The PPC Manager's Dilemma

Any PPC Consultant faces this dilemna. Now here's the question: Which is better?

  1. A higher CTR, higher Quality Score and lower CPCs, or...
  2. The ability to conduct more reliable ad text analysis?

You can't have both. Setting ads to rotate is inevitably accepting a reduction in CTR, a lower Quality Score and an increase in CPC prices, but allowing for more insightful ad text analysis. Setting ads to optimize is delivering extra visitors, each at a lower cost, but at the expense of meaningful comparative ad text data.

Supporters of the "rotate" setting might suggest that if ads are optimized regularly based on CTR, it is possible to increase Quality Score over time. But however regularly ads are optimized, there will always be a loss in potential CTR and an increase in CPCs during the testing period, where your low CTR ads are shown instead of those with a higher CTR.

Of course, it is generally more important to optimize ads based on conversions, rather than just CTR. But unless you can be sure that one ad is going to convert at a significantly higher rate than another, is it really worth paying a higher price just so you can conduct better ad text analysis?

To rephrase the question as a common scenario which often faces PPC managers: Which is better?

  1. $100,000 in profit, not knowing which ads worked best, or...
  2. $90,000 in profit, but knowing which ads worked best?

I imagine there will be advocates of both.

AdWords Performance Grader




If you enjoyed this post, please consider leaving a comment.

Comments

Monday March 15, 2010

Craig Danuloff (not verified) Said:

Alan - Good post, and interesting question. There is one technical point that does work a little differently than you suggest, and may play into the decision anyone wants to make about which path to follow.

Quality Score is calculated only for keyword-ad pairs; there is not really a 'Quality Score' for the keyword based on the average CTR of all the ads. So the setting of the ads has no impact on QS beyond the fact that when the lower CTR ad is shown, the keyword has a lower QS, and when the higher CTR ad is shown, the keyword has a higher QS.

Even that is a simplification (there's always more complexity, isn't there). One reason that when you set ads to 'rotate' they never actually are shown evenly or equally is that for any given query, if one KW-Ad combo will earn a higher quality score than that same KW matched with a different Ad, the higher QS pair will be shown - in other words Google automatically 'optimizes' based on Quality Score, even when the ad distribution is set to 'rotate'! I know that raises a lot of questions, which space doesn't permit going into, but I just wanted to add these facts to the discussion.

I would add that my own belief is that this is yet another case where the choice is if you want to let Google do something for you, or do it yourself. If you don't have time or energy to monitor ad results and react, then optimize may be a good choice. If you do - if you're running real ad copy tests (and if you're not, why run multiple ads in the first place?) then 'rotate' is better, yes there is cost to the test, but once you act on the results the long term net benefit should be higher.

Great issue, thanks for writing it up.

Monday March 15, 2010

James Ward (not verified) Said:

I'm a fan of 'rotate' on the basis that over the long term you gain more in QS by identifying the best performing ads as quickly as possible and retiring the losers. It is more difficult, and takes longer, to identify winners when ad serving is set to 'optimise'.

Monday March 15, 2010

Shawn Livengood (not verified) Said:

Alan,

Thanks for the mention. While I'm on record as being firmly in the pro-rotation camp, I do think you have a valid point. However, there are other things you can do to improve a poor CTR/quality score other than relying on ad optimization. If you have an ad that converts well in a split test but has a poor CTR, increasing the clicks should increase your conversions, as long as traffic quality remains consistent. I would suggest trying higher ad positions (via higher keyword bids), more compelling ad text, or creative use of dynamic keyword insertion before resorting to automatic ad optimization if you're concerned that your low CTR is causing low quality scores.

Monday March 15, 2010

PPC Wizard (not verified) Said:

I agree with your analysis. However, let's look at the same situation with some conversion rates:

Ad 1 Conversion Rate = 3%
Ad 2 Conversion Rate = 5%

Therefore "rotate" works out as follows:

Ad 1: 3000 impressions, 3.0% CTR, 90 clicks, 3 conversions (2.7 to be precise)
Ad 2: 3000 impressions, 1.5% CTR, 45 clicks, 2 conversions (2.25 to be precise)
Total: 6000 impressions, 2.25% CTR, 135 clicks, 5 conversions

And "optimize" works out as follows:

Ad 1: 5000 impressions, 3.0% CTR, 150 clicks, 5 conversions (4.5 to be precise)
Ad 2: 1000 impressions, 1.5% CTR, 15 clicks, 1 conversions (0.75 to be precise)
Total: 6000 impressions, 2.75% CTR, 165 clicks, 6 conversions

If "rotate" has a CPC of $1, that means you spend $135 for 5 conversions, or $27/conversion.
Say "optimize" gets a CPC of $0.90, which means you spend $148.50 for 6 conversions, or $24.75/conversion.

Congratulations, you have a higher QS, lower CPC and lower cost/conversion. However, on "optimize" Google will run the next 6,000 impressions the exact same way they ran the first 6,000. With "rotate" you could eliminate Ad 1 (assuming you're making the decision on conversion rate, not CTR) and the next 6,000 impressions will go off as follows:

Ad 2: 6000 impressions, 1.5% CTR, 90 clicks, 5 conversions (4.5 to be precise)

Spend would have been $90 for 5 conversions, or $18/conversion. At first glance the rebuttal would be that we made fewer conversions. However, with a lower cost/conversion we can afford to increase our bids by up to 35%. That could lead to a higher CTR (better position, even if you bought it) and more sales at a lower cost/conversion than "optimize".

That's the other side of the coin and each company has to make their own decision. "Optimize" will save you time as you won't be testing and optimizing yourself. However, you could be paying more per conversion than you would have otherwise.

Tuesday March 16, 2010

CustardMite (not verified) Said:

Great post Alan,

My perspective on this has always been quite simple. If you believe that your different adverts are likely to have different conversion rates (maybe due to highlighting different USP's, or landing on different pages), then it's essential to monitor the impact of each advert on the conversion rate - as PPC Wizard has highlighted.

If you believe it is unlikely that the conversion rate will be substantially altered by the different adverts (maybe the adverts are testing slightly different wording, but the basic message is the same), then there is no need to wait until you've got statistically significant results on the conversion rate.

If this is the case, then you want to test the clickthrough rate as quickly as possible - the sooner that you get a result, the more tests you can run, and the quicker you can deliver improvements to your clickthrough rate.

Optimised advert rotation concerns me on two fronts. Firstly, by showing one advert less often, you are going to have to wait longer for a statistically significant result. But secondly, if you haven't got a significant result, then what is Google using to display your adverts more/less often?

If you don't have time to test adverts regularly, then Optimised advert rotation will certainly help, as it will eventually show the 'losing' advert so rarely that it almost never appears. In effect, Google stops your advert test for you. If you have time to test your adverts regularly, then I would always advocate even rotation.

Tuesday March 16, 2010

Alan Mitchell (not verified) Said:

Thanks for all your comments. Glad to see it’s an issue that doesn’t just keep me up at night!

@ Craig

Thanks for your clarity on how Google calculates Quality Score based on keyword and ad pairs. You’re right, if there are multiple keywords and multiple ads in an ad group, it will be difficult to ensure the impressions of each ad are equal, even if those ads are set to ‘rotate’. Guess the only way to force a more equal rotation is to give each keyword its own ad group – although this can be impractical.

Excellent point too about considering the long-term net benefit of ad rotation. There will always be a cost of testing ads (through temporarily higher CPCs, lower CTRs, and the time and effort to analyse the results of the ads), so perhaps a better question might be:

Which is better?

1. $100,000 in profit, not knowing which ads worked best, or...
2. $90,000 in profit, knowing which ads worked best, and having the data and insight to potentially make $140,000 in profit in period 2?

Guess it depends how confident you are you can recoup your temporary losses from ad testing in the long-term. And also how long is the long term. A statistically significant ad test result could take months, by which time you could have raked up a huge premium on your click costs because of rotation.

@ James

Again, you’re right that long-term is a key issue here. Rotating ads is definitely more efficient for deciding which ads to cull, but since there is a temporary cost in rotating ads, for it to be worthwhile, these temporary losses will need to be made up in terms of significant improvements.

@ Shaun

No problem – you’ve written some excellent posts recently.

I agree that ad testing is only one tool in a PPC manager’s toolkit to improve the performance of a PPC campaign – ad testing should only really be used in combination with other optimisation methods as you point out.

@ PPC Wizard

You’re right - it does become interesting when conversion rates are taken into consideration. If one ad has a significantly higher conversion rate than another, as in your example, it makes sense to use the ‘rotate’ setting and manually cull the poorly-converting ad to increase the overall conversion rate and lower the overall cost per conversion.

But that’s assuming conversion rates for each ad are very different. I imagine there will be a large number of ad groups whose ads have extremely similar conversion rates (as CustardMite goes on to point out). If you use more similar conversion rates, say 4.0% and 4.5%, setting ads to ‘optimize’ still works out better (6.7 conversions) than setting ads to ‘rotate’ (5.6 conversions).

It kind of goes back to an earlier point I made - unless you can be sure that one ad is going to convert at a significantly higher rate than another, is it really worth paying a higher price just so you can conduct better ad text analysis?

@ CustardMite

I think you’re spot on. Expectation in conversion rates must be the critical factor. As you mention, ads are often so similar (different wordings but same overall message) that it is often unrealistic to assume that similarly-worded ads will perform very differently in the long-term.

The more I think about it the more I believe the best scenario would be something like this:

1. Immediately set ads to ‘optimize’ once a new campaign goes live – CTR will be higher, click volume higher and CPCs lower
2. Wait for conversion data to be collected for each of your ads
3. Look for noticeable differences in conversion rate between ads
4. If, and only if, ads show significant differences in conversion rates, set those ads to ‘rotate’ in order to conduct further analysis (perhaps in a separate ‘optimize’ campaign)

I think this method provides the needed justification for rotating ads, while all the time minimising the significant costs of unnecessary rotation until conversion data suggests it's worthwhile.

If done correctly, perhaps it’s possible to get the best of both worlds...

Friday March 19, 2010

Josh Summerhays (not verified) Said:

@ Craig and Alan

The way I had it explained to me by a Google rep, when you set ads to Rotate, each is entered into query auctions on an equal basis. However, because one ad may not have as high a quality score benefit as another, it will inevitably "flunk out" in a given auction and not be shown. Depending on how tightly themed the keywords are in a given ad group, the disparity in ad impressions may be large (possible opportunity to break down ad group) or small (tightly themed), but that's why you see different impression numbers for most text ad split tests set to "Rotate".

Monday March 22, 2010

Alan Mitchell (not verified) Said:

Thanks for the clarity Josh. This disceprancy in impressions due to low Quality Score makes perfect sense, and impression differences from ratating ads could potentially be used as a basic indicator to assess how different ads impact Quality Score.

What's more, I imagine if ads were set to 'optimize', Google will naturally choose ads which tend to result in a higher Quality Score, so in a way the 'rotate' setting does inherit an element of the 'optimize' setting, albeit only a weak one.

Monday April 05, 2010

The Pitfalls Of A/B Ad Split Testing, Part 2 (not verified) Said:

[...] Equations makes this point effectively in his response to Alan Mitchell’s recent article, “Rotating Ads vs. Optimizing Ads: Which Is Better?” on the Wordstream [...]

Tuesday April 06, 2010

Matt Van Wagner (not verified) Said:

Nice post and great comments. I've referenced your post in my latest SEL column, Pitfalls of A/B Ad Split Testing (http://searchengineland.com/the-pitfalls-of-ab-ad-split-testing-part-2-3...) where I address some of the peripheral issues related to this Rotate v. Optimize discussion.

I've got a theory that presents an alternative to the oft-levied criticism that the default setting of optimize ads is purely a mechanism for Google to make more money on paid search. There are common situations, I believe, where it is not in Google's best interest to optimize to an individual ad, but rather to optimize ad the ad groups level using an ad set that contains multiple ads. My belief is that optimizing to a single ad can actually cause ad group performance degradation under common conditions.

Unfortunately, I can't find a way to prove my theory that a set of strongly performing ads, an "ad set" can outperform the winning ad from A/B Ad testing, because they can't be tested simultaneously. I'm offering a small bounty for anyone who can help me design an experiment that can develop or debunk this theory. The reward is a lobster dinner for two from lobster.com that can be delivered to you anywhere in the US. See my Search Engine Land article for more details.

Thank you for developing such a high quality blog!
Matt Van Wagner

Tuesday April 06, 2010

Mark Lee (not verified) Said:

All interesting comments. If your real goal is conversion or A/S, keep in mind that "sugar-coated" ads and those with careless keyword insertion will usually pull a higher CTR (and a better QS), but often runs much lower conversion than a more "plain vanilla" ad.
@Matt (I read your full post on SearchEngineLand): as an old catalog/list circulation guy, I'm accustomed to the "regression towards the mean" phenomenon, where you just can't expect the winner of a split test to repeat the same results (any outliers, good or bad, will tend towards the average).

Friday May 14, 2010

Alan Mitchell (not verified) Said:

@ Matt

Thanks for your comment. Liked your article and hypothesis. If I've understood you correctly I believe you're arguing that having two very different ads in an ad group will often result in better results (CTR, conversion rate, sales etc.) than having only one 'killer' ad, which has previously been optimised.

That makes perfect sense. Since broad and phrase-matched keywords can match to a wide of search terms, I imagine Google will choose to choose to show the best ad for each search query. So only having one ad in an ad group (even if it's previously performed better than other ads) will limit the diversity of your ad offerinf relevancy and could be worse than having two 'weaker' ads in an ad group.

An experiement which tests this would be awesome. In terms of executuion, the only way I think it would work is using an exact match keyword (so you can be sure what the user has searched for), putting each test case in a seperate campaign, and pausing and unpausing each campaign simultaneously every 15 minutes. It will never be able to run two identical keywords simultaneously, but 15 minute campaign rotation is about as close to simultaneous as I think it's possible to get.

Cheers,
Alan

Thursday July 08, 2010

Moises (not verified) Said:

Great Blog and Comments. I have learned a lot reading them. I admit my apprenticeship here. While I was reading, my logic insists on the following. Start with "rotate" let's say with an add group of 6 adds, test which ones perform better, then reduce the add group to 3 adds, and set it to "optimize". It is only natural to have a higher CPC at the beginning (Like in any other type of investment, the beginning is more costly),then it will settle. Then, re do it again, create new add groups, set them to "rotate", test them, reduce add amounts within group, set them to "optimze", and so on. Cyclically.

Wednesday July 21, 2010

Alan Mitchell (not verified) Said:

@ Moises

That's an interesting argument - that higher initial CPCs of rotating ads can be seen as a necessary investment for unbiased ad performance data.

But I'm not sure higher CPCs are really that necessary. Surely it's better to test ads more cheaply from the start (using the optimize setting), and then look into rotating ads once you have a better idea of which ads are working?

Wednesday December 01, 2010

PPCOptimization (not verified) Said:

I do not rotate my ads, instead when I create a new ad mix, it is tuned to vet out the top performing ad with the highest conversions. This lets Google do its work and get the highest CTR and best conversions.

Typically I will run 3 or 4 new ads, The existing ad, new ad copy to existing landing page, and existing ad copy to new landing page. These variations give me visibility into how the ad copy and landing page work together while also seeing which provide highest CTR and conversions.

Sunday January 16, 2011

The Pitfalls Of A/B Ad Split Testing, Part 2 | GetMeSEO.com (not verified) Said:

[...] Equations makes this point effectively in his response to Alan Mitchell’s recent article, “Rotating Ads vs. Optimizing Ads: Which Is Better?” on the Wordstream [...]

Tuesday May 03, 2011

Stever (not verified) Said:

I work with small businesses at local levels where traffic volumes tend to be relatively low. It can sometimes take weeks or more to get enough clicks and impressions for even simple A/B split testing of ad copy, let alone conversion rates. I find Google's "optimize" settings rarely gives a new ad version enough of a chance, at least at these time frames. So I always opt for "rotate".

Thursday May 05, 2011

Alan Mitchell (not verified) Said:

@ Stever It makes sense to use the 'rotate' setting if you place a high value on collecting reliable testing data. But if this means that CTR is lower (which is likely to be the case as Google tends to 'optimise' ads based on CTR), surely this reliable data is at the expense of extra visitors and cheaper CPCs. In effect, are you not simply paying a premium for your reliable data?

Tuesday June 07, 2011

Randall Gniadecki (not verified) Said:

Great article and great conversation/debate. I fall somewhere in the middle. Each is appropriate at different times. I began 6 years ago with AdWords and have always used "rotate" and have over those years developed dynamic split-testing that begins with 16 ads and quickly optimizes the ads down to 2-3 of the best overall combination of ad elements based on Conversions and ROI. Of course this is easiest with a site that actually sells products online but some manipulation of data in excel can usually quickly reveal the losers to be eliminated. Once about 2/3 of the adgroups in a campaign are optimized I set the ads to optimize. For me it is all about isolating variables. Once I have the ads to a good place I steady them with the "optimize" setting and begin testing one of the other dozens of aspects I split-test. How you use the tools Google provides depends a lot on how you organize your campaigns and how you prefer to have your data separated.

Monday February 20, 2012

Leeds Acupuncture Clinic (not verified) Said:

I'm going to try ads for my business.  My traffic volumes will be very low due to the newness and small size of my business.

If I go for rotating adds how long should I leave them before I get meanigful data?

Or should I go optimised and just let Google do the thinking for me?

 

Monday June 25, 2012

Anonymous (not verified) Said:

Great info thanks for sharing

Tuesday September 11, 2012

Mike (not verified) Said:

Sure, I'd take the $100,000 profit over the $90,000 any day of the week. But since this is a hypothetical example, we don't truly know if "Optimize" or "Rotate" would produce the higher conversion. So in the final analysis, I'd say you've made a great case for testing both.

Wednesday November 06, 2013

C Ned (not verified) Said:

Allright folks, all the data and all the chit chat but not deffinate answers asto what is better? 

I would much rather prefer to make more money than less, and I am not a republican either. 

Put the adds running on optimize for clicks and forget it.

Leave a Comment

Type the characters you see in this picture. (verify using audio)
Type the characters you see in the picture above; if you can't read them, submit the form and a new image will be generated. Not case sensitive.
 
Free Keyword Tool

Get thousands of relevent keyword suggestions - more,
faster, free!

Free Keyword Niche Finder

Discover profitable pockets of keywords for your
business.

Free Negative Keyword Suggestion Tool

Identify wasted spend before it happens and increase
your paid search ROI.

Contact Us | Company | Support | Site Map | Trademarks | Privacy Policy © 2007-2014 WordStream, Inc. All rights reserved.