How Many PPC Ads Should You Test?

SMS Text

There are many essential ways to improve your AdWords account’s performance, but one of my favorites is ad testing. When I review people’s existing AdWords accounts, one of the most common bad habits I see is running just one ad per Ad Group. This is a missed opportunity. You’re losing revenue and ROI by not testing ads.

Ad Testing has huge benefits for the advertiser:

  • Discover scientifically which offer gets the best response from your prospects
  • Increase CTR and quality score
  • Increase conversion rates
  • Improve your ROI
  • Outshine your competitors

Once you decide to start running more than one ad per Ad Group, you might ask:

  • How many ads should I run?
  • How long should I run them?
  • Should I test offers or copy changes first?

In this post, I’m going to answer the first two questions-

How many ads should you test and for how long should you run them?

It depends on your click volume: how many clicks your ad group gets per month. Each keyword gets a different numbers of searches per month, so each Ad Group gets a different number of searches per month.

Let’s look at an example:


This is one month of clicks per ad group in one of our client’s PPC accounts. Ad Group names are blurred out to protect our client’s competitive advantages. But you can see that clicks per month range from as low as zero, seven, or 22 to as high as almost 13,000.

Calculating The Number of Ads to Run Based on Click Volume

Generally speaking, I like to see 100 clicks per ad before I compare their CTR to decide which to keep. If you want to test your ads monthly, then all you need to do is divide the number of clicks per month by 100, and that tells you the maximum number of ads you should run per month.

# of ads to run = # of clicks per month / 100

For the ad group that gets 300 clicks per month, you can run three ads. For the one that gets 12,700 clicks per month, you can run 120 ads. That’s a lot of ads. I normally test from two to five ads at a time. If you want to run 120 ads, that’s going to take a lot more creativity and planning, but it might be the right move in some cases.

If you want to be more aggressive but less sure of your data, you can judge ads after only 50 or 75 clicks.

You can also calculate how expensive your tests will be:

Avg CPC * 200 = Minumum cost to test two ads

Don’t freak out about spending $500 to test two ads. You would spend the $500 anyway- the question is whether you come out learning something that will improve your results going forward. It’s typical to find ads that produce 2-5 times the ROI of others. That $500 you spent and the time you spent doing ad testing could easily increase your profit by $1500 the next month.

How Not to Screw Up Your Ad Testing

It’s best if other things are NOT optimized at this time. If you add or remove keywords from the ad group in the middle of the month, this will affect your ad’s statistics. Some ads might work better with the set of keywords before you changed them, and some work better with the modified set. This creates a confusion and divides your ad testing into two more tests- do you get enough clicks monthly to do that? Planning what’s going to be tested ahead of time and not changing things willy nilly makes for less headaches later when you try to understand the results.

Same goes for landing pages- if you start testing a variety of landing pages using Website Optimizer, you’re adding complexity to the mix that makes your data less reliable. The goal is to find out which ads work best for prospects searching a set of related keywords, so keep keywords and landing pages stable while testing ads.

Judging Ads for Conversions

In the end, you want the ads that produce the best ROI. You probably want great CTR, great Quality Score, great Conversion Rate, great Average Sale, and great ROAS or ROI. So, judge ads by both CTR and conversion metrics. To me, the best situation is one where you can check ROAS (revenue divided by ad cost) for each ad. To compare ads this way, you’ll have to run a report… ROAS for ads is not shown in the normal interface.

The next problem is: how many conversions per ad do you need to be sure your results are reliable? I’d love to avoid getting into “statistical confidence” here- I’ve found online tools that make it easier to judge this according to clicks for CTR, but I haven’t found a free tool that helps you with statistical confidence for conversion rate or for ROAS.

If your sale amount varies per conversion, you need a number of sales to be sure that you’re seeing a reliable average sale number per ad. I’d wait for 5-10 sales before optimizing at this level. That’s not ideal (in reality we need some advanced statistical calculations for confidence) and I don’t like optimizing this way, but if you wait for 100 sales, that could take from 2,000 to 10,000 clicks and you could easily spend $4,000 to $20,000 per ad test. Unless you have huge budgets and are really uptight about statistical perfection, I wouldn’t advise that.

Brian Carter
Brian is author of The Like Economy: How Businesses Make Money With Facebook and Facebook Marketing: Leveraging Facebook's Features For Your Marketing Campaigns, How to... Read Full Bio
Brian Carter
Get the latest news from Search Engine Journal!
We value your privacy! See our policy here.
  • I appreciate the article – especially about how NOT to screw up your testing. It’s often a challenge to isolate process long enough to get a real take on what works and doesn’t, yet it’s so important that it needed to be emphasized and it’s good to see that on a scale the size of your client’s you apply the same principle.

  • i agree with that I appreciate the article – especially about how NOT to screw up your testing. It’s often a challenge to isolate process long enough to get a real take on what works and doesn’t, yet it’s so important that it needed to be emphasized and it’s good to see that on a scale the size of your client’s you apply the same principle.

  • I agree with testing multiple ads but how do you determine the number to start out with in a campaign you are starting from the ground up? I’ve chosen to do 5 different ads to begin with each with a different type of audience and offer perspective…what would you advise?

  • Good stuff! It’s easy to get caught up in testing, and wind up testing too much. If you can’t get a good enough sampling, you’ll never have a good idea about what’s working and what’s not.

  • Finally I found a great post about ads split testing. Thanks!
    One thing I do differently is I like to wait till I get 300 clicks. I also like to run landing pages test at the same time. So I got 2 different copies and 2 LP’s = 4 text ads.

  • You can use any tool (like that works for CTR with conversion rate as well. Just put the number of clicks where you would normally put the number of impressions and put the number of conversions where you would normally put the number of clicks.

  • That may be true Richard, but CR is not usually our KPI either. How do you do it for ROAS? Or you want to get optimal CTR, CR, Avg Sale, AND CPC? I don’t have the statistics training to figure that out but my math intuition tells me it won’t work with Am I wrong?

  • I love using winneralert tool… it emails you when there is action needed on an adgroup’s ads.

  • I have seen Perry Marshall recommend testing no more than two or three ads at a time and then use what I think he calls “peel and stick” to drive up the CTR. So start with your two or three ads, run them till you have sufficient click volume then pause and replace the lowest performing ad (i.e. the one with the lowest ctr). Keep on doing that with the aim of always driving up the CTR.

    To me, this makes sense since it provides we with a limited number of ads to get my limited brain around!

  • i use Turbo Powered Typo Generator. PPC typo traffic can work wonders – particularly with trademark-blocked keywords on Google. This typo generator offers more control over output than most.