PPC Ad Copy Testing Strategies

Key Takeaways

a few weeks ago we released a video talking about responsive search ads templates and how you can use them to test your ad copy messaging for responsive search ads shortly after I put out that video I realized that I really only was talking about how to craft individual rsas and that we haven't really talked on this channel about how to conduct ad copy testing on the whole so in this video what I want to talk about are some best practices for search ad copy testing talk about cadences and stats to review so that you can start to put together a holistic ad copy testing strategy for your account so the part that I'm going to spend the least amount of time on in this video is talking about what you should test and that's because I'm going to direct you to this video I unfortunately didn't point you to the video card in the intro so I'm gonna do that now at the top of the screen you can find the link to the Google ads responsive search ads strategies and templates video that we launched earlier this year that'll walk you through a number of different ad templates and suggestions for messaging that you can use for your responsive search ads now that they're the only ad format available for Google search campaigns so now that we've got that out of the way let's start to talk about the mechanics and metrics and best practices for ad copy testing so the first thing I want to talk about is how you should go about testing your ads in search campaigns and for the better part of a decade now I've been utilizing a fairly simple approach that I think will work well for pretty much anybody and can be applied to any account we're going to use this graphic which unfortunately only goes counterclockwise but hopefully it's relatively easy to follow the first thing is we want to test large differences so in this instance you would use different templates for your responsive search ads maybe you start with one that is brand focused first one that is benefits focused just a big difference in messaging so that you're testing a completely different category of AD creative and seeing how they perform compared to each other once you've determined a winner between those two you can then go to the secondary stage which is going to be on the bottom right which is where you narrow in on a theme let's say that your pricing ad copy one so maybe we want to start testing different messaging within that pricing theme is it enough to say that we're priced better than our competitors or do we need to actually put the monetary amount in the copy does it work better when we add discount codes or when we just say there's a percentage off on the website or does free shipping have a big impact on the performance as opposed to not mentioning any sort of shipping consideration at all how can we start to narrow in on which messaging works best within that theme to see what we can do to maximize performance and then the last stage stage 3 and kind of the upper right of this circle if you will is to just test slight variances on the messaging odds are all of you are consumers so you know that there are a number of different ways to articulate that there is a 30 off sale this weekend so maybe you just use a different word order or maybe in this instance you start to pin different headlines in different spaces you might use the exact same copy but maybe you have your price message pinned in headline one in one variant and in a second variant you have a pinned in headline two and then in a third variant you have a pinned in the headline three in this last stage you're trying to find a much more specific takeaway from where your ads will work and then once you've got that you then revert back to a large difference now we feel like we've got the best messaging in the best order that we could possibly use for that pricing message so now let's test a diff different template maybe we want to focus on benefits or features rather than price to see which one will perform better and then the testing just continues in the circular pattern of making a large difference and then narrowing in and then trying to perfect the add variant at the end and then going back to the large difference again so the next logical question is how many variants should I be testing at any given point I can't give you a perfect answer on this question because it's going to be different for every account there are a couple considerations I have and then one broad guideline the first is that the number of variants you're going to use really depends on the volume the more volume you have flowing through any given ad test is going to make it so you can probably support more ad variants and if you have a low amount of volume you probably need to limit the number of AD variants you have in place because you're just not going to get data that's conclusive enough to determine a winner no matter how much volume you have the more variance you test the longer that your ad test will need to run you're just going to need more time to gather more data to determine which one has performed the best as a general rule of thumb I usually say use anywhere from two to five active ad variants in any given search ad group or campaign or aggregate level test that you have because that gives you a good number of different variants to utilize but it also means that you're not going to have 10 or 20 different ad copies trying to run against each other and trying to gain enough data if you've got a smaller account or you want something run on a shorter timeline probably stick to two if you've got a larger account or you're comfortable letting your ad test run for quite a while you can lean closer to that four or five end of the scale but no matter how many different ad variants you have running you really need to have a normal length of time for your ad copy test to run so what is that time range on the whole I say that a minimum of two weeks is necessary for any AD copy test I really don't care how many ad variants you want I'm not comfortable making a call any short of two weeks mostly this is because any week compared to a following week could be very different depending on factors that you know about for seasonality or any number of external effects that you have no idea about that's going on additionally any given day from day to day is going to be quite a bit different Mondays are different than Tuesdays are different than Fridays are different than Sundays everybody knows that that's just the way the world works so utilizing only a couple of days isn't going to be enough time because you're not taking into account the overall flow of the week and at least having two weeks back to back gives you two instances of those Mondays Wednesdays Thursdays to gather enough data to hopefully balance things out a little bit I know some companies that spend millions of dollars might test things for a couple of days I'm personally not on board with that I would suggest you never test anything shorter than two weeks overall as I mentioned the less data you have the longer your test needs to run even if you want something to run for two weeks if you haven't gathered enough data it doesn't make sense for you to pause that test on the long end you can let tests run for two weeks four weeks a quarter six months a year two years if you want to obviously letting a test run for two years is not going to be super actionable for you so that's probably a little bit on the long scale so to help mitigate that you can lower the number of AD variants you have consolidate your data and try and turn your test over more quickly probably the longest I've seen an ad test run in any account that I've managed is about six months after that you're really operating on data that is super old try and design your tests so that you don't have to have them run for probably any more than six months but anywhere in the middle there the only thing you need to keep in mind is that you need to let each variant gather enough data so that it could have potentially converted I talked about this a little bit in the keyword pausing video which you can check out at the top of the screen right now where it's not fair to pause a keyword simply because it hasn't converted if it hasn't spent enough money to even hit your target CPA the same is true for ad copy if a certain variant has only spent 25 dollars and your target cost per conversion is 100 doesn't make sense to pause that ad because it could easily convert on that next incremental click and then it has a very low cost per conversion at maybe 26 dollars compared to your target CPL at a hundred so no matter how many variants you have running in a test make sure each one of them individually has had enough spend to potentially convert upwards of two or three times before you decide to pause now aside from just the target CPA that you have in place there are a number of other metrics you can use to analyze your ad copy tests I'll preface this section by saying that for the most part the companies that I work with are very conversion and bottom line focused we are trying to get conversions we are performance marketers I don't run nearly as many campaigns on branding or engagement so most of the metrics that I'm going to talk about are very conversion for focused the first ones we usually start off with are CPA and row as because the cost per lead or the return on ad spend is going to be not only how many sales we've gotten but how profitable those sales are in business profitability is the name of the game the more profit that you have the better off you are so following that logic the ads that are the most profitable are likely the ones that we want to keep around to determine which messaging works best and what we want to test and iterate on for that next ad test in the cycle moving further away from profitability we also need to look at conversion rate how often do people actually convert when they hit the landing page this will also lean into your CPA and row as numbers but this could be an easy place for you to start to decide which ads are at least generating higher likelihood of a conversion even if the profitability isn't as good and then you can try and reverse engineer to impact the overall margins of those conversions I also look to review click-through rate and Page position because I think it's important to know which of your ads are the most attractive and are driving the highest clicks and what type of impact you're having by showing up in either the absolute top position the top section of the page or just on the first page utilizing those impression share metrics are really important to see where you're showing up and understanding how competitive you are and those will impact your click-through rate which in turn will decide how many people are actually coming to your page and giving you an opportunity to convert them all of these are metrics I regularly review and I look to see what types of patterns we can find in there but if I'm looking at just ad copy testing and I'm trying to narrow down to the ad variant that is the most effective I will almost always make a custom column and that's going to be conversions divided by Impressions this is where I'm trying to see on average given any number of Impressions How likely is it that I'm going to generate a conversion from that impression in Google ads you can create custom columns which if you don't know how to do that we have a video that you can check out at the top of the screen right now but this is the formula you'll need to use you can see just underneath the name that we've got there we've got the plus column and plus function right below that you can see conversions divided by Impressions it's a very easy formula the only thing I want to call out is that you should shift this to a percent data format because it's just a lot easier to read as a percent rather than a number so if I create this stat and apply it to an ad copy test that I have running an account right now you can see that I have my four ad variants and then I have a conversion divided by Impressions column that gives me a percentage that makes it really easy to review how things are performing add variant number one is across the board the winner in terms of pretty much all stats and the impression to conversion rate is 1.71 compared to the other AD variants which are only 1.12 at the second highest for add four and then adds two and three are a bit lower than that but still above one percent so in this instance I wouldn't need to notice that the click-through rate for add one the conversion rate and the cost per conversion are all better than the other AD variants I would only need to look at that conversion to impression number and notice that it has a almost double rate than the other ones so it's clearly the winner now this example is a lead generation account that does not have any value tied to the leads but if you have values associated with your conversions or you're tracking any sort of Revenue with your conversions you could also create a custom column that is revenue divided by impression to understand how much on average you're making per impression From Any Given ad variant the number will look somewhat similar in terms of having one specific stat that you can refer to but since you would be using conversion value or Revenue you need to make sure that that number was formatted as currency rather than a percentage now with some of the numbers that were in that previous slide there were a couple things I wanted to call out as ad testing challenges that simply are part of AD testing you're not going to be able to get around them hopefully they don't impact you every time but they are things that we'll need to deal with and decisions that we'll have to make the first is that nothing rotates evenly anymore although we have four ad variants each of them has relatively similar messaging and there's no big difference in the quality score of the keywords anything like that you'll notice that we have a pretty big discrepancy in the number of Impressions that first ad has 7 800 Impressions the second has 6200 and adds three and four are below the 3000 impression Line This changed quite a while ago in the Google ads algorithm and it has to do with the fact that the decision on which ad to show is made before the impression is one as opposed to after the impression has already been won so unfortunately even if you go to your settings in your ad account and choose rotate ads evenly you'll still get an impression mix that looks something like this it's just the nature of the game and it's another reason why it's important to make sure that every individual variant within your ad test has enough data to have converted a couple or three times before you decide that it's the loser of a test because it could just be that Google has discounted it for one reason or another the Second Challenge is that not all of your metrics will agree as we saw with the first ad variant here it has the best click-through rate cost per conversion conversion rate and impression to conversion number but hypothetically let's assume that we only have an ad test that's running between add 2 and add three this is something that happens to me quite frankly all the time we have click-through rates that are quite a bit different add 2 is almost at 18 percent whereas add three is just over 15. the conversion rate for add 2 is a bit lower it's a 5.88 whereas add 3 is up a little bit it's at 6.76 and even my trusty little metric at the end tells me that they're pretty evenly comparable with add to just slightly having a higher chance of converting at 1.06 compared to add 3 at 1.04 and the problem here is if you look at the cost per conversion that is also quite a bit different due to the cost per click being different for these different ad variants these two are in the same ad group they trigger for the same keywords they show up across the same devices same ad schedule there's no reason that the cost per click should be that different and yet it is so in this instance I really just have to make a gut call am I going to treat these two ads as being comparable to each other am I going to assume that the one with the better cost per conversion works best or am I also going to take into account the fact that ad3 is being shown only about half as often as add 2 by Google and maybe utilizing the higher cost per conversion variant will at least ensure that I have a greater impression share and I have more conversions even if they're more expensive for each conversion it's a real conundrum and in this instance I'm really happy that I don't have to make that call because I have add one that's outperforming everything but don't be surprised if you have instances like this in your account where you need to make a decision and there's no real right or wrong answer you just have to make a call make your next test and keep things moving as a quick follow-up I just have a slight note to say on statistical significance I'm not going to go into what that is if you don't utilize that or don't hear it please skip to the next section but if you do and you want my personal opinion on it here it is statistical significance is great to wait for if you can achieve it I have a lot of clients who really lean heavily into the numbers and data science portion of paid advertising and they want to wait for statistical significance and sometimes if the volume is high enough we can get that but because of the challenges I just discussed it gets harder and harder to determine if we're able to reach those significance levels either because ads aren't rotating evenly or you have to decide if you want statistical significance on your cost per conversion or your conversion rate or the impression to conversion number you have to choose what you want significance on and additionally in my experience most people who require statistical significance to turn over a test want something that is in 90 to 95 percent confidence level and in a lot of instances that's just not realistic I've been able to convince somebody to work down to the 80 confidence level before and at that point they felt that it didn't matter because it was only 80 confidence I personally would feel pretty good thinking that four out of five times my ad copy test was backed by data science but some people don't so in my mind the bottom line on statistical significance is that if you can get it do it if you can't don't worry about it do the best you can with the data and the challenges in gathering that data that you have and move forward with a test okay stepping off of my soapbox now I want to talk about just ad testing best practices to close out the first best practice applies to pretty much everything in paid media but in my mind even more specifically in ad copy testing start with a plan and stick to it you saw my ad testing Cadence of testing something large then tweaking the messaging and then trying to perfect it and going back to a large Cadence this allows me to map out a number of different tests moving forward so I always know know what's going to happen depending on the account I try and set realistic expectations of how long it's going to take to determine a winner but I know that I'm always going to let a test run for at least two weeks and at the long end we're going to cut it off at six months and decide that it's no longer viable we'll start a new test there's nothing worse than putting in all of the effort to put together an ad copy test and then turning it off too soon or deviating from the plan and not having any takeaways from it next is to use aggregate level testing wherever possible I know that it can be really attractive to write individual ad copy messaging for every single ad group and every single campaign that you have in your account but that leads to very small data sets and makes it really hard to turn over ad copy tests not to mention you would have an ad copy test running for every single ad group so you'd have to monitor the data for every single ad test and determine timing stats winners potentially statistical significance all that stuff I personally would rather use aggregate level testing even if the ad messaging is not exactly identical vehicle across ad groups you can still test a pricing theme in all ad groups compared to a features or benefits theme you can then narrow down which type of messaging works best while still keeping keyword specific ad components in those ad groups overall aggregate level testing consolidates data gives you more insights to work on and allows you to turn over ad tests faster and requires quite frankly just less work overall and lastly always allow enough time to pass and enough data to be collected I can't stress this enough I know I've already talked about it a number of times but turning over a test too quickly or not letting enough data flow through really just wastes everybody's time and it wastes the Insight that you could gain from that data if you just let it gather a little bit more and determine which winner you got out of it ad testing is a fundamental part of search campaign management it's really important because it is the first time that you get your messaging out in front of your potential customers there's a lot you can learn from running ad copy tests as long as you make sure that you have a good strategy going in you're looking at the right metrics to determine what performance is and then you're utilizing that to inform future tests to make sure that you're always moving things forward there are quite a number of components and considerations that go into ad copy testing I think I've covered the main ones that come to my mind in this video but if you have any questions or any follow-ups on any of these pieces or if you have any additional considerations that I didn't include in this video about ad testing I'd love to hear about it in the comments below thanks for watching our video if you thought it was useful give us a thumbs up below we release a new video at least once a week so if you want to get notified of when a new one comes out be sure to subscribe to the paid media Pros Channel foreign