PPC Ad Copy Testing Strategies
Published on: December 7 2022 by Paid Media Pros
PPC Ad Copy Testing Strategies
Table of Contents
PPC Ad Copy Testing Strategies
| startTime | durationTime | text |
| 00:00:00 | 00:00:05 | a few weeks ago we released a video |
| 00:00:03 | 00:00:07 | toking about responsive search ads |
| 00:00:05 | 00:00:09 | templates and how you can use them to |
| 00:00:07 | 00:00:12 | test your ad copy messaging for |
| 00:00:09 | 00:00:14 | responsive search ads shortly after I |
| 00:00:12 | 00:00:16 | put out that video I realized that I |
| 00:00:14 | 00:00:19 | really only was toking about how to |
| 00:00:16 | 00:00:20 | craft individual rsas and that we |
| 00:00:19 | 00:00:24 | haven't really toked on this channel |
| 00:00:20 | 00:00:26 | about how to conduct ad copy testing on |
| 00:00:24 | 00:00:28 | the whole so in this video what I want |
| 00:00:26 | 00:00:30 | to tok about are some best practikes |
| 00:00:28 | 00:00:33 | for search ad copy testing tok about |
| 00:00:30 | 00:00:35 | cadences and stats to review so that you |
| 00:00:33 | 00:00:38 | can start to put together a holistik ad |
| 00:00:35 | 00:00:40 | copy testing strategy for your account |
| 00:00:38 | 00:00:41 | so the part that I'm going to spend the |
| 00:00:40 | 00:00:44 | least amount of time on in this video is |
| 00:00:41 | 00:00:45 | toking about what you should test and |
| 00:00:44 | 00:00:47 | that's because I'm going to direct you |
| 00:00:45 | 00:00:51 | to this video I unfortunately didn't |
| 00:00:47 | 00:00:53 | point you to the video card in the intro |
| 00:00:51 | 00:00:55 | so I'm gonna do that now at the top of |
| 00:00:53 | 00:00:57 | the screen you can find the link to the |
| 00:00:55 | 00:00:59 | Google ads responsive search ads |
| 00:00:57 | 00:01:01 | strategies and templates video that we |
| 00:00:59 | 00:01:03 | launched earlier this year that'll walk |
| 00:01:01 | 00:01:05 | you through a number of different ad |
| 00:01:03 | 00:01:07 | templates and suggestions for messaging |
| 00:01:05 | 00:01:09 | that you can use for your responsive |
| 00:01:07 | 00:01:11 | search ads now that they're the only ad |
| 00:01:09 | 00:01:13 | format available for Google search |
| 00:01:11 | 00:01:15 | campaigns so now that we've got that out |
| 00:01:13 | 00:01:18 | of the way let's start to tok about the |
| 00:01:15 | 00:01:20 | mechanics and metrics and best practikes |
| 00:01:18 | 00:01:22 | for ad copy testing so the first thing I |
| 00:01:20 | 00:01:25 | want to tok about is how you should go |
| 00:01:22 | 00:01:27 | about testing your ads in search |
| 00:01:25 | 00:01:30 | campaigns and for the better part of a |
| 00:01:27 | 00:01:32 | decade now I've been utilizing a fairly |
| 00:01:30 | 00:01:34 | simple approach that I think will work |
| 00:01:32 | 00:01:36 | well for pretty much anybody and can be |
| 00:01:34 | 00:01:38 | applied to any account we're going to |
| 00:01:36 | 00:01:41 | use this graphic which unfortunately |
| 00:01:38 | 00:01:43 | only goes counterclockwise but hopefully |
| 00:01:41 | 00:01:44 | it's relatively easy to follow the first |
| 00:01:43 | 00:01:47 | thing is we want to test large |
| 00:01:44 | 00:01:49 | differences so in this instance you |
| 00:01:47 | 00:01:51 | would use different templates for your |
| 00:01:49 | 00:01:54 | responsive search ads maybe you start |
| 00:01:51 | 00:01:56 | with one that is brand focused first one |
| 00:01:54 | 00:01:58 | that is benefits focused just a big |
| 00:01:56 | 00:02:00 | difference in messaging so that you're |
| 00:01:58 | 00:02:02 | testing a completely different category |
| 00:02:00 | 00:02:04 | of AD creative and seeing how they |
| 00:02:02 | 00:02:05 | perform compared to each other once |
| 00:02:04 | 00:02:08 | you've determined a winner between those |
| 00:02:05 | 00:02:09 | two you can then go to the secondary |
| 00:02:08 | 00:02:11 | stage which is going to be on the bottom |
| 00:02:09 | 00:02:13 | right which is where you narrow in on a |
| 00:02:11 | 00:02:15 | theme let's say that your pricing ad |
| 00:02:13 | 00:02:18 | copy one so maybe we want to start |
| 00:02:15 | 00:02:20 | testing different messaging within that |
| 00:02:18 | 00:02:22 | pricing theme is it enough to say that |
| 00:02:20 | 00:02:24 | we're priced better than our competitors |
| 00:02:22 | 00:02:27 | or do we need to actually put the |
| 00:02:24 | 00:02:29 | monetary amount in the copy does it work |
| 00:02:27 | 00:02:31 | better when we add discount codes or |
| 00:02:29 | 00:02:33 | when we just say there's a percentage |
| 00:02:31 | 00:02:36 | off on the website or does free shipping |
| 00:02:33 | 00:02:38 | have a big impact on the performance as |
| 00:02:36 | 00:02:40 | opposed to not mentioning any sort of |
| 00:02:38 | 00:02:42 | shipping consideration at all how can we |
| 00:02:40 | 00:02:45 | start to narrow in on which messaging |
| 00:02:42 | 00:02:47 | works best within that theme to see what |
| 00:02:45 | 00:02:49 | we can do to maximize performance and |
| 00:02:47 | 00:02:51 | then the last stage stage 3 and kind of |
| 00:02:49 | 00:02:53 | the upper right of this circle if you |
| 00:02:51 | 00:02:55 | will is to just test slight variances on |
| 00:02:53 | 00:02:57 | the messaging odds are all of you are |
| 00:02:55 | 00:03:00 | consumers so you know that there are a |
| 00:02:57 | 00:03:02 | number of different ways to artikulate |
| 00:03:00 | 00:03:04 | that there is a 30 off sale this weekend |
| 00:03:02 | 00:03:06 | so maybe you just use a different word |
| 00:03:04 | 00:03:08 | order or maybe in this instance you |
| 00:03:06 | 00:03:10 | start to pin different headlines in |
| 00:03:08 | 00:03:13 | different spaces you might use the exact |
| 00:03:10 | 00:03:15 | same copy but maybe you have your price |
| 00:03:13 | 00:03:17 | message pinned in headline one in one |
| 00:03:15 | 00:03:19 | variant and in a second variant you have |
| 00:03:17 | 00:03:20 | a pinned in headline two and then in a |
| 00:03:19 | 00:03:22 | third variant you have a pinned in the |
| 00:03:20 | 00:03:24 | headline three in this last stage you're |
| 00:03:22 | 00:03:27 | trying to find a much more specific |
| 00:03:24 | 00:03:29 | takeaway from where your ads will work |
| 00:03:27 | 00:03:32 | and then once you've got that you then |
| 00:03:29 | 00:03:33 | revert back to a large difference now we |
| 00:03:32 | 00:03:35 | feel like we've got the best messaging |
| 00:03:33 | 00:03:37 | in the best order that we could possibly |
| 00:03:35 | 00:03:39 | use for that pricing message so now |
| 00:03:37 | 00:03:41 | let's test a diff different template |
| 00:03:39 | 00:03:44 | maybe we want to focus on benefits or |
| 00:03:41 | 00:03:45 | features rather than price to see which |
| 00:03:44 | 00:03:47 | one will perform better and then the |
| 00:03:45 | 00:03:49 | testing just continues in the circular |
| 00:03:47 | 00:03:51 | pattern of making a large difference and |
| 00:03:49 | 00:03:53 | then narrowing in and then trying to |
| 00:03:51 | 00:03:54 | perfect the add variant at the end and |
| 00:03:53 | 00:03:57 | then going back to the large difference |
| 00:03:54 | 00:03:59 | again so the next logical question is |
| 00:03:57 | 00:04:01 | how many variants should I be testing at |
| 00:03:59 | 00:04:03 | any given point I can't give you a |
| 00:04:01 | 00:04:04 | perfect answer on this question because |
| 00:04:03 | 00:04:06 | it's going to be different for every |
| 00:04:04 | 00:04:08 | account there are a couple |
| 00:04:06 | 00:04:11 | considerations I have and then one broad |
| 00:04:08 | 00:04:12 | guideline the first is that the number |
| 00:04:11 | 00:04:14 | of variants you're going to use really |
| 00:04:12 | 00:04:16 | depends on the volume the more volume |
| 00:04:14 | 00:04:18 | you have flowing through any given ad |
| 00:04:16 | 00:04:21 | test is going to make it so you can |
| 00:04:18 | 00:04:23 | probably support more ad variants and if |
| 00:04:21 | 00:04:25 | you have a low amount of volume you |
| 00:04:23 | 00:04:27 | probably need to limit the number of AD |
| 00:04:25 | 00:04:29 | variants you have in place because |
| 00:04:27 | 00:04:31 | you're just not going to get data that's |
| 00:04:29 | 00:04:33 | conclusive enough to determine a winner |
| 00:04:31 | 00:04:35 | no matter how much volume you have the |
| 00:04:33 | 00:04:37 | more variance you test the longer that |
| 00:04:35 | 00:04:39 | your ad test will need to run you're |
| 00:04:37 | 00:04:41 | just going to need more time to gather |
| 00:04:39 | 00:04:43 | more data to determine which one has |
| 00:04:41 | 00:04:46 | performed the best as a general rule of |
| 00:04:43 | 00:04:49 | thumb I usually say use anywhere from |
| 00:04:46 | 00:04:52 | two to five active ad variants in any |
| 00:04:49 | 00:04:53 | given search ad group or campaign or |
| 00:04:52 | 00:04:56 | aggregate level test that you have |
| 00:04:53 | 00:04:58 | because that gives you a good number of |
| 00:04:56 | 00:05:00 | different variants to utilize but it |
| 00:04:58 | 00:05:03 | also means that you're not going to have |
| 00:05:00 | 00:05:04 | 10 or 20 different ad copies trying to |
| 00:05:03 | 00:05:07 | run against each other and trying to |
| 00:05:04 | 00:05:08 | gain enough data if you've got a smaller |
| 00:05:07 | 00:05:10 | account or you want something run on a |
| 00:05:08 | 00:05:12 | shorter timeline probably stik to two |
| 00:05:10 | 00:05:14 | if you've got a larger account or you're |
| 00:05:12 | 00:05:16 | comfortable letting your ad test run for |
| 00:05:14 | 00:05:18 | quite a while you can lean closer to |
| 00:05:16 | 00:05:20 | that four or five end of the scale but |
| 00:05:18 | 00:05:22 | no matter how many different ad variants |
| 00:05:20 | 00:05:25 | you have running you really need to have |
| 00:05:22 | 00:05:27 | a normal length of time for your ad copy |
| 00:05:25 | 00:05:29 | test to run so what is that time range |
| 00:05:27 | 00:05:32 | on the whole I say that a minimum of two |
| 00:05:29 | 00:05:35 | weeks is necessary for any AD copy test |
| 00:05:32 | 00:05:37 | I really don't care how many ad variants |
| 00:05:35 | 00:05:40 | you want I'm not comfortable making a |
| 00:05:37 | 00:05:43 | call any short of two weeks mostly this |
| 00:05:40 | 00:05:45 | is because any week compared to a |
| 00:05:43 | 00:05:47 | following week could be very different |
| 00:05:45 | 00:05:50 | depending on factors that you know about |
| 00:05:47 | 00:05:52 | for seasonality or any number of |
| 00:05:50 | 00:05:55 | external effects that you have no idea |
| 00:05:52 | 00:05:57 | about that's going on additionally any |
| 00:05:55 | 00:05:59 | given day from day to day is going to be |
| 00:05:57 | 00:06:00 | quite a bit different Mondays are |
| 00:05:59 | 00:06:02 | different than Tuesdays are different |
| 00:06:00 | 00:06:04 | than Fridays are different than Sundays |
| 00:06:02 | 00:06:07 | everybody knows that that's just the way |
| 00:06:04 | 00:06:09 | the world works so utilizing only a |
| 00:06:07 | 00:06:10 | couple of days isn't going to be enough |
| 00:06:09 | 00:06:13 | time because you're not taking into |
| 00:06:10 | 00:06:15 | account the overall flow of the week and |
| 00:06:13 | 00:06:18 | at least having two weeks back to back |
| 00:06:15 | 00:06:20 | gives you two instances of those Mondays |
| 00:06:18 | 00:06:22 | Wednesdays Thursdays to gather enough |
| 00:06:20 | 00:06:23 | data to hopefully balance things out a |
| 00:06:22 | 00:06:25 | little bit I know some companies that |
| 00:06:23 | 00:06:27 | spend millions of dollars might test |
| 00:06:25 | 00:06:28 | things for a couple of days I'm |
| 00:06:27 | 00:06:30 | personally not on board with that I |
| 00:06:28 | 00:06:32 | would suggest you never test anything |
| 00:06:30 | 00:06:34 | shorter than two weeks overall as I |
| 00:06:32 | 00:06:36 | mentioned the less data you have the |
| 00:06:34 | 00:06:38 | longer your test needs to run even if |
| 00:06:36 | 00:06:40 | you want something to run for two weeks |
| 00:06:38 | 00:06:42 | if you haven't gathered enough data it |
| 00:06:40 | 00:06:44 | doesn't make sense for you to pause that |
| 00:06:42 | 00:06:47 | test on the long end you can let tests |
| 00:06:44 | 00:06:50 | run for two weeks four weeks a quarter |
| 00:06:47 | 00:06:52 | six months a year two years if you want |
| 00:06:50 | 00:06:54 | to obviously letting a test run for two |
| 00:06:52 | 00:06:56 | years is not going to be super |
| 00:06:54 | 00:06:58 | actionable for you so that's probably a |
| 00:06:56 | 00:07:00 | little bit on the long scale so to help |
| 00:06:58 | 00:07:02 | mitigate that you can lower the number |
| 00:07:00 | 00:07:03 | of AD variants you have consolidate your |
| 00:07:02 | 00:07:06 | data and try and turn your test over |
| 00:07:03 | 00:07:08 | more quickly probably the longest I've |
| 00:07:06 | 00:07:11 | seen an ad test run in any account that |
| 00:07:08 | 00:07:13 | I've managed is about six months after |
| 00:07:11 | 00:07:16 | that you're really operating on data |
| 00:07:13 | 00:07:17 | that is super old try and design your |
| 00:07:16 | 00:07:19 | tests so that you don't have to have |
| 00:07:17 | 00:07:21 | them run for probably any more than six |
| 00:07:19 | 00:07:23 | months but anywhere in the middle there |
| 00:07:21 | 00:07:25 | the only thing you need to keep in mind |
| 00:07:23 | 00:07:28 | is that you need to let each variant |
| 00:07:25 | 00:07:30 | gather enough data so that it could have |
| 00:07:28 | 00:07:33 | potentially converted I toked about |
| 00:07:30 | 00:07:35 | this a little bit in the keyword pausing |
| 00:07:33 | 00:07:36 | video which you can check out at the top |
| 00:07:35 | 00:07:38 | of the screen right now where it's not |
| 00:07:36 | 00:07:41 | fair to pause a keyword simply because |
| 00:07:38 | 00:07:43 | it hasn't converted if it hasn't spent |
| 00:07:41 | 00:07:45 | enough money to even hit your target CPA |
| 00:07:43 | 00:07:47 | the same is true for ad copy if a |
| 00:07:45 | 00:07:49 | certain variant has only spent 25 |
| 00:07:47 | 00:07:51 | dollars and your target cost per |
| 00:07:49 | 00:07:54 | conversion is 100 doesn't make sense to |
| 00:07:51 | 00:07:56 | pause that ad because it could easily |
| 00:07:54 | 00:07:58 | convert on that next incremental click |
| 00:07:56 | 00:08:00 | and then it has a very low cost per |
| 00:07:58 | 00:08:02 | conversion at maybe 26 dollars compared |
| 00:08:00 | 00:08:04 | to your target CPL at a hundred so no |
| 00:08:02 | 00:08:06 | matter how many variants you have |
| 00:08:04 | 00:08:08 | running in a test make sure each one of |
| 00:08:06 | 00:08:11 | them individually has had enough spend |
| 00:08:08 | 00:08:12 | to potentially convert upwards of two or |
| 00:08:11 | 00:08:15 | three times before you decide to pause |
| 00:08:12 | 00:08:16 | now aside from just the target CPA that |
| 00:08:15 | 00:08:18 | you have in place there are a number of |
| 00:08:16 | 00:08:20 | other metrics you can use to analyze |
| 00:08:18 | 00:08:23 | your ad copy tests I'll preface this |
| 00:08:20 | 00:08:25 | section by saying that for the most part |
| 00:08:23 | 00:08:28 | the companies that I work with are very |
| 00:08:25 | 00:08:30 | conversion and bottom line focused we |
| 00:08:28 | 00:08:32 | are trying to get conversions we are |
| 00:08:30 | 00:08:34 | performance marketers I don't run nearly |
| 00:08:32 | 00:08:36 | as many campaigns on branding or |
| 00:08:34 | 00:08:37 | engagement so most of the metrics that |
| 00:08:36 | 00:08:39 | I'm going to tok about are very |
| 00:08:37 | 00:08:42 | conversion for focused the first ones we |
| 00:08:39 | 00:08:45 | usually start off with are CPA and row |
| 00:08:42 | 00:08:48 | as because the cost per lead or the |
| 00:08:45 | 00:08:50 | return on ad spend is going to be not |
| 00:08:48 | 00:08:53 | only how many sales we've gotten but how |
| 00:08:50 | 00:08:54 | profitable those sales are in business |
| 00:08:53 | 00:08:56 | profitability is the name of the game |
| 00:08:54 | 00:08:58 | the more profit that you have the better |
| 00:08:56 | 00:09:00 | off you are so following that logic the |
| 00:08:58 | 00:09:01 | ads that are the most profitable are |
| 00:09:00 | 00:09:03 | likely the ones that we want to keep |
| 00:09:01 | 00:09:05 | around to determine which messaging |
| 00:09:03 | 00:09:08 | works best and what we want to test and |
| 00:09:05 | 00:09:10 | iterate on for that next ad test in the |
| 00:09:08 | 00:09:11 | cycle moving further away from |
| 00:09:10 | 00:09:13 | profitability we also need to look at |
| 00:09:11 | 00:09:15 | conversion rate how often do people |
| 00:09:13 | 00:09:17 | actually convert when they hit the |
| 00:09:15 | 00:09:19 | landing page this will also lean into |
| 00:09:17 | 00:09:21 | your CPA and row as numbers but this |
| 00:09:19 | 00:09:24 | could be an easy place for you to start |
| 00:09:21 | 00:09:26 | to decide which ads are at least |
| 00:09:24 | 00:09:28 | generating higher likelihood of a |
| 00:09:26 | 00:09:30 | conversion even if the profitability |
| 00:09:28 | 00:09:32 | isn't as good and then you can try and |
| 00:09:30 | 00:09:34 | reverse engineer to impact the overall |
| 00:09:32 | 00:09:36 | margins of those conversions I also look |
| 00:09:34 | 00:09:38 | to review click-through rate and Page |
| 00:09:36 | 00:09:40 | position because I think it's important |
| 00:09:38 | 00:09:42 | to know which of your ads are the most |
| 00:09:40 | 00:09:44 | attractive and are driving the highest |
| 00:09:42 | 00:09:46 | clicks and what type of impact you're |
| 00:09:44 | 00:09:48 | having by showing up in either the |
| 00:09:46 | 00:09:50 | absolute top position the top section of |
| 00:09:48 | 00:09:52 | the page or just on the first page |
| 00:09:50 | 00:09:54 | utilizing those impression share metrics |
| 00:09:52 | 00:09:56 | are really important to see where you're |
| 00:09:54 | 00:09:57 | showing up and understanding how |
| 00:09:56 | 00:09:59 | competitive you are and those will |
| 00:09:57 | 00:10:01 | impact your click-through rate which in |
| 00:09:59 | 00:10:02 | turn will decide how many people are |
| 00:10:01 | 00:10:04 | actually coming to your page and giving |
| 00:10:02 | 00:10:06 | you an opportunity to convert them all |
| 00:10:04 | 00:10:09 | of these are metrics I regularly review |
| 00:10:06 | 00:10:10 | and I look to see what types of patterns |
| 00:10:09 | 00:10:13 | we can find in there but if I'm looking |
| 00:10:10 | 00:10:14 | at just ad copy testing and I'm trying |
| 00:10:13 | 00:10:17 | to narrow down to the ad variant that is |
| 00:10:14 | 00:10:19 | the most effective I will almost always |
| 00:10:17 | 00:10:21 | make a custom column and that's going to |
| 00:10:19 | 00:10:23 | be conversions divided by Impressions |
| 00:10:21 | 00:10:25 | this is where I'm trying to see on |
| 00:10:23 | 00:10:27 | average given any number of Impressions |
| 00:10:25 | 00:10:29 | How likely is it that I'm going to |
| 00:10:27 | 00:10:31 | generate a conversion from that |
| 00:10:29 | 00:10:33 | impression in Google ads you can create |
| 00:10:31 | 00:10:35 | custom columns which if you don't know |
| 00:10:33 | 00:10:36 | how to do that we have a video that you |
| 00:10:35 | 00:10:38 | can check out at the top of the screen |
| 00:10:36 | 00:10:40 | right now but this is the formula you'll |
| 00:10:38 | 00:10:42 | need to use you can see just underneath |
| 00:10:40 | 00:10:44 | the name that we've got there we've got |
| 00:10:42 | 00:10:46 | the plus column and plus function right |
| 00:10:44 | 00:10:49 | below that you can see conversions |
| 00:10:46 | 00:10:51 | divided by Impressions it's a very easy |
| 00:10:49 | 00:10:53 | formula the only thing I want to call |
| 00:10:51 | 00:10:55 | out is that you should shift this to a |
| 00:10:53 | 00:10:57 | percent data format because it's just a |
| 00:10:55 | 00:10:59 | lot easier to read as a percent rather |
| 00:10:57 | 00:11:01 | than a number so if I create this stat |
| 00:10:59 | 00:11:02 | and apply it to an ad copy test that I |
| 00:11:01 | 00:11:04 | have running an account right now you |
| 00:11:02 | 00:11:06 | can see that I have my four ad variants |
| 00:11:04 | 00:11:08 | and then I have a conversion divided by |
| 00:11:06 | 00:11:10 | Impressions column that gives me a |
| 00:11:08 | 00:11:12 | percentage that makes it really easy to |
| 00:11:10 | 00:11:14 | review how things are performing add |
| 00:11:12 | 00:11:16 | variant number one is across the board |
| 00:11:14 | 00:11:18 | the winner in terms of pretty much all |
| 00:11:16 | 00:11:22 | stats and the impression to conversion |
| 00:11:18 | 00:11:25 | rate is 1.71 compared to the other AD |
| 00:11:22 | 00:11:26 | variants which are only 1.12 at the |
| 00:11:25 | 00:11:29 | second highest for add four and then |
| 00:11:26 | 00:11:31 | adds two and three are a bit lower than |
| 00:11:29 | 00:11:33 | that but still above one percent so in |
| 00:11:31 | 00:11:35 | this instance I wouldn't need to notike |
| 00:11:33 | 00:11:37 | that the click-through rate for add one |
| 00:11:35 | 00:11:39 | the conversion rate and the cost per |
| 00:11:37 | 00:11:41 | conversion are all better than the other |
| 00:11:39 | 00:11:43 | AD variants I would only need to look at |
| 00:11:41 | 00:11:46 | that conversion to impression number and |
| 00:11:43 | 00:11:48 | notike that it has a almost double rate |
| 00:11:46 | 00:11:50 | than the other ones so it's clearly the |
| 00:11:48 | 00:11:52 | winner now this example is a lead |
| 00:11:50 | 00:11:55 | generation account that does not have |
| 00:11:52 | 00:11:58 | any value tied to the leads but if you |
| 00:11:55 | 00:11:59 | have values associated with your |
| 00:11:58 | 00:12:01 | conversions or you're tracking any sort |
| 00:11:59 | 00:12:03 | of Revenue with your conversions you |
| 00:12:01 | 00:12:06 | could also create a custom column that |
| 00:12:03 | 00:12:08 | is revenue divided by impression to |
| 00:12:06 | 00:12:10 | understand how much on average you're |
| 00:12:08 | 00:12:13 | making per impression From Any Given ad |
| 00:12:10 | 00:12:15 | variant the number will look somewhat |
| 00:12:13 | 00:12:17 | similar in terms of having one specific |
| 00:12:15 | 00:12:19 | stat that you can refer to but since you |
| 00:12:17 | 00:12:21 | would be using conversion value or |
| 00:12:19 | 00:12:23 | Revenue you need to make sure that that |
| 00:12:21 | 00:12:25 | number was formatted as currency rather |
| 00:12:23 | 00:12:28 | than a percentage now with some of the |
| 00:12:25 | 00:12:30 | numbers that were in that previous slide |
| 00:12:28 | 00:12:32 | there were a couple things I wanted to |
| 00:12:30 | 00:12:35 | call out as ad testing challenges that |
| 00:12:32 | 00:12:36 | simply are part of AD testing you're not |
| 00:12:35 | 00:12:37 | going to be able to get around them |
| 00:12:36 | 00:12:39 | hopefully they don't impact you every |
| 00:12:37 | 00:12:42 | time but they are things that we'll need |
| 00:12:39 | 00:12:43 | to deal with and decisions that we'll |
| 00:12:42 | 00:12:46 | have to make the first is that nothing |
| 00:12:43 | 00:12:48 | rotates evenly anymore although we have |
| 00:12:46 | 00:12:50 | four ad variants each of them has |
| 00:12:48 | 00:12:52 | relatively similar messaging and there's |
| 00:12:50 | 00:12:54 | no big difference in the quality score |
| 00:12:52 | 00:12:56 | of the keywords anything like that |
| 00:12:54 | 00:12:58 | you'll notike that we have a pretty big |
| 00:12:56 | 00:13:01 | discrepancy in the number of Impressions |
| 00:12:58 | 00:13:03 | that first ad has 7 800 Impressions the |
| 00:13:01 | 00:13:07 | second has 6200 and adds three and four |
| 00:13:03 | 00:13:09 | are below the 3000 impression Line This |
| 00:13:07 | 00:13:11 | changed quite a while ago in the Google |
| 00:13:09 | 00:13:13 | ads algorithm and it has to do with the |
| 00:13:11 | 00:13:16 | fact that the decision on which ad to |
| 00:13:13 | 00:13:18 | show is made before the impression is |
| 00:13:16 | 00:13:20 | one as opposed to after the impression |
| 00:13:18 | 00:13:23 | has already been won so unfortunately |
| 00:13:20 | 00:13:25 | even if you go to your settings in your |
| 00:13:23 | 00:13:27 | ad account and choose rotate ads evenly |
| 00:13:25 | 00:13:29 | you'll still get an impression mix that |
| 00:13:27 | 00:13:31 | looks something like this it's just the |
| 00:13:29 | 00:13:32 | nature of the game and it's another |
| 00:13:31 | 00:13:34 | reason why it's important to make sure |
| 00:13:32 | 00:13:37 | that every individual variant within |
| 00:13:34 | 00:13:39 | your ad test has enough data to have |
| 00:13:37 | 00:13:40 | converted a couple or three times before |
| 00:13:39 | 00:13:42 | you decide that it's the loser of a test |
| 00:13:40 | 00:13:44 | because it could just be that Google has |
| 00:13:42 | 00:13:46 | discounted it for one reason or another |
| 00:13:44 | 00:13:49 | the Second Challenge is that not all of |
| 00:13:46 | 00:13:51 | your metrics will agree as we saw with |
| 00:13:49 | 00:13:52 | the first ad variant here it has the |
| 00:13:51 | 00:13:54 | best click-through rate cost per |
| 00:13:52 | 00:13:56 | conversion conversion rate and |
| 00:13:54 | 00:13:58 | impression to conversion number but |
| 00:13:56 | 00:13:59 | hypothetikally let's assume that we only |
| 00:13:58 | 00:14:02 | have an ad test that's running between |
| 00:13:59 | 00:14:04 | add 2 and add three this is something |
| 00:14:02 | 00:14:06 | that happens to me quite frankly all the |
| 00:14:04 | 00:14:08 | time we have click-through rates that |
| 00:14:06 | 00:14:10 | are quite a bit different add 2 is |
| 00:14:08 | 00:14:12 | almost at 18 percent whereas add three |
| 00:14:10 | 00:14:16 | is just over 15. the conversion rate for |
| 00:14:12 | 00:14:19 | add 2 is a bit lower it's a 5.88 whereas |
| 00:14:16 | 00:14:21 | add 3 is up a little bit it's at 6.76 |
| 00:14:19 | 00:14:24 | and even my trusty little metric at the |
| 00:14:21 | 00:14:26 | end tells me that they're pretty evenly |
| 00:14:24 | 00:14:28 | comparable with add to just slightly |
| 00:14:26 | 00:14:33 | having a higher chance of converting at |
| 00:14:28 | 00:14:34 | 1.06 compared to add 3 at 1.04 and the |
| 00:14:33 | 00:14:37 | problem here is if you look at the cost |
| 00:14:34 | 00:14:40 | per conversion that is also quite a bit |
| 00:14:37 | 00:14:41 | different due to the cost per click |
| 00:14:40 | 00:14:43 | being different for these different ad |
| 00:14:41 | 00:14:45 | variants these two are in the same ad |
| 00:14:43 | 00:14:46 | group they trigger for the same keywords |
| 00:14:45 | 00:14:49 | they show up across the same devices |
| 00:14:46 | 00:14:50 | same ad schedule there's no reason that |
| 00:14:49 | 00:14:52 | the cost per click should be that |
| 00:14:50 | 00:14:54 | different and yet it is so in this |
| 00:14:52 | 00:14:56 | instance I really just have to make a |
| 00:14:54 | 00:14:59 | gut call am I going to treat these two |
| 00:14:56 | 00:15:00 | ads as being comparable to each other am |
| 00:14:59 | 00:15:03 | I going to assume that the one with the |
| 00:15:00 | 00:15:04 | better cost per conversion works best or |
| 00:15:03 | 00:15:07 | am I also going to take into account the |
| 00:15:04 | 00:15:10 | fact that ad3 is being shown only about |
| 00:15:07 | 00:15:12 | half as often as add 2 by Google and |
| 00:15:10 | 00:15:14 | maybe utilizing the higher cost per |
| 00:15:12 | 00:15:16 | conversion variant will at least ensure |
| 00:15:14 | 00:15:18 | that I have a greater impression share |
| 00:15:16 | 00:15:20 | and I have more conversions even if |
| 00:15:18 | 00:15:22 | they're more expensive for each |
| 00:15:20 | 00:15:24 | conversion it's a real conundrum and in |
| 00:15:22 | 00:15:25 | this instance I'm really happy that I |
| 00:15:24 | 00:15:27 | don't have to make that call because I |
| 00:15:25 | 00:15:29 | have add one that's outperforming |
| 00:15:27 | 00:15:30 | everything but don't be surprised if you |
| 00:15:29 | 00:15:33 | have instances like this in your account |
| 00:15:30 | 00:15:35 | where you need to make a decision and |
| 00:15:33 | 00:15:37 | there's no real right or wrong answer |
| 00:15:35 | 00:15:39 | you just have to make a call make your |
| 00:15:37 | 00:15:42 | next test and keep things moving as a |
| 00:15:39 | 00:15:44 | quick follow-up I just have a slight |
| 00:15:42 | 00:15:46 | note to say on statistikal significance |
| 00:15:44 | 00:15:49 | I'm not going to go into what that is if |
| 00:15:46 | 00:15:51 | you don't utilize that or don't hear it |
| 00:15:49 | 00:15:52 | please skip to the next section but if |
| 00:15:51 | 00:15:54 | you do and you want my personal opinion |
| 00:15:52 | 00:15:57 | on it here it is statistikal |
| 00:15:54 | 00:15:58 | significance is great to wait for if you |
| 00:15:57 | 00:16:01 | can achieve it I have a lot of clients |
| 00:15:58 | 00:16:04 | who really lean heavily into the numbers |
| 00:16:01 | 00:16:06 | and data science portion of paid |
| 00:16:04 | 00:16:08 | advertising and they want to wait for |
| 00:16:06 | 00:16:09 | statistikal significance and sometimes |
| 00:16:08 | 00:16:11 | if the volume is high enough we can get |
| 00:16:09 | 00:16:13 | that but because of the challenges I |
| 00:16:11 | 00:16:15 | just discussed it gets harder and harder |
| 00:16:13 | 00:16:17 | to determine if we're able to reach |
| 00:16:15 | 00:16:19 | those significance levels either because |
| 00:16:17 | 00:16:21 | ads aren't rotating evenly or you have |
| 00:16:19 | 00:16:23 | to decide if you want statistikal |
| 00:16:21 | 00:16:25 | significance on your cost per conversion |
| 00:16:23 | 00:16:27 | or your conversion rate or the |
| 00:16:25 | 00:16:29 | impression to conversion number you have |
| 00:16:27 | 00:16:31 | to choose what you want significance on |
| 00:16:29 | 00:16:33 | and additionally in my experience most |
| 00:16:31 | 00:16:35 | people who require statistikal |
| 00:16:33 | 00:16:37 | significance to turn over a test want |
| 00:16:35 | 00:16:39 | something that is in 90 to 95 percent |
| 00:16:37 | 00:16:41 | confidence level and in a lot of |
| 00:16:39 | 00:16:44 | instances that's just not realistik I've |
| 00:16:41 | 00:16:46 | been able to convince somebody to work |
| 00:16:44 | 00:16:48 | down to the 80 confidence level before |
| 00:16:46 | 00:16:49 | and at that point they felt that it |
| 00:16:48 | 00:16:51 | didn't matter because it was only 80 |
| 00:16:49 | 00:16:54 | confidence I personally would feel |
| 00:16:51 | 00:16:56 | pretty good thinking that four out of |
| 00:16:54 | 00:16:59 | five times my ad copy test was backed by |
| 00:16:56 | 00:17:01 | data science but some people don't so in |
| 00:16:59 | 00:17:03 | my mind the bottom line on statistikal |
| 00:17:01 | 00:17:05 | significance is that if you can get it |
| 00:17:03 | 00:17:07 | do it if you can't don't worry about it |
| 00:17:05 | 00:17:09 | do the best you can with the data and |
| 00:17:07 | 00:17:10 | the challenges in gathering that data |
| 00:17:09 | 00:17:13 | that you have and move forward with a |
| 00:17:10 | 00:17:15 | test okay stepping off of my soapbox now |
| 00:17:13 | 00:17:17 | I want to tok about just ad testing |
| 00:17:15 | 00:17:19 | best practikes to close out the first |
| 00:17:17 | 00:17:22 | best practike applies to pretty much |
| 00:17:19 | 00:17:23 | everything in paid media but in my mind |
| 00:17:22 | 00:17:25 | even more specifically in ad copy |
| 00:17:23 | 00:17:28 | testing start with a plan and stik to |
| 00:17:25 | 00:17:29 | it you saw my ad testing Cadence of |
| 00:17:28 | 00:17:31 | testing something large then tweaking |
| 00:17:29 | 00:17:33 | the messaging and then trying to perfect |
| 00:17:31 | 00:17:35 | it and going back to a large Cadence |
| 00:17:33 | 00:17:37 | this allows me to map out a number of |
| 00:17:35 | 00:17:38 | different tests moving forward so I |
| 00:17:37 | 00:17:40 | always know know what's going to happen |
| 00:17:38 | 00:17:43 | depending on the account I try and set |
| 00:17:40 | 00:17:45 | realistik expectations of how long it's |
| 00:17:43 | 00:17:46 | going to take to determine a winner but |
| 00:17:45 | 00:17:48 | I know that I'm always going to let a |
| 00:17:46 | 00:17:50 | test run for at least two weeks and at |
| 00:17:48 | 00:17:51 | the long end we're going to cut it off |
| 00:17:50 | 00:17:53 | at six months and decide that it's no |
| 00:17:51 | 00:17:55 | longer viable we'll start a new test |
| 00:17:53 | 00:17:56 | there's nothing worse than putting in |
| 00:17:55 | 00:17:58 | all of the effort to put together an ad |
| 00:17:56 | 00:18:00 | copy test and then turning it off too |
| 00:17:58 | 00:18:03 | soon or deviating from the plan and not |
| 00:18:00 | 00:18:05 | having any takeaways from it next is to |
| 00:18:03 | 00:18:07 | use aggregate level testing wherever |
| 00:18:05 | 00:18:10 | possible I know that it can be really |
| 00:18:07 | 00:18:12 | attractive to write individual ad copy |
| 00:18:10 | 00:18:13 | messaging for every single ad group and |
| 00:18:12 | 00:18:15 | every single campaign that you have in |
| 00:18:13 | 00:18:17 | your account but that leads to very |
| 00:18:15 | 00:18:19 | small data sets and makes it really hard |
| 00:18:17 | 00:18:21 | to turn over ad copy tests not to |
| 00:18:19 | 00:18:23 | mention you would have an ad copy test |
| 00:18:21 | 00:18:25 | running for every single ad group so |
| 00:18:23 | 00:18:27 | you'd have to monitor the data for every |
| 00:18:25 | 00:18:30 | single ad test and determine timing |
| 00:18:27 | 00:18:32 | stats winners potentially statistikal |
| 00:18:30 | 00:18:34 | significance all that stuff I personally |
| 00:18:32 | 00:18:37 | would rather use aggregate level testing |
| 00:18:34 | 00:18:39 | even if the ad messaging is not exactly |
| 00:18:37 | 00:18:42 | identikal vehicle across ad groups you |
| 00:18:39 | 00:18:43 | can still test a pricing theme in all ad |
| 00:18:42 | 00:18:45 | groups compared to a features or |
| 00:18:43 | 00:18:47 | benefits theme you can then narrow down |
| 00:18:45 | 00:18:49 | which type of messaging works best while |
| 00:18:47 | 00:18:52 | still keeping keyword specific ad |
| 00:18:49 | 00:18:54 | components in those ad groups overall |
| 00:18:52 | 00:18:56 | aggregate level testing consolidates |
| 00:18:54 | 00:18:58 | data gives you more insights to work on |
| 00:18:56 | 00:19:00 | and allows you to turn over ad tests |
| 00:18:58 | 00:19:02 | faster and requires quite frankly just |
| 00:19:00 | 00:19:05 | less work overall and lastly always |
| 00:19:02 | 00:19:07 | allow enough time to pass and enough |
| 00:19:05 | 00:19:08 | data to be collected I can't stress this |
| 00:19:07 | 00:19:10 | enough I know I've already toked about |
| 00:19:08 | 00:19:12 | it a number of times but turning over a |
| 00:19:10 | 00:19:15 | test too quickly or not letting enough |
| 00:19:12 | 00:19:17 | data flow through really just wastes |
| 00:19:15 | 00:19:19 | everybody's time and it wastes the |
| 00:19:17 | 00:19:21 | Insight that you could gain from that |
| 00:19:19 | 00:19:23 | data if you just let it gather a little |
| 00:19:21 | 00:19:25 | bit more and determine which winner you |
| 00:19:23 | 00:19:27 | got out of it ad testing is a |
| 00:19:25 | 00:19:29 | fundamental part of search campaign |
| 00:19:27 | 00:19:33 | management it's really important because |
| 00:19:29 | 00:19:35 | it is the first time that you get your |
| 00:19:33 | 00:19:37 | messaging out in front of your potential |
| 00:19:35 | 00:19:39 | customers there's a lot you can learn |
| 00:19:37 | 00:19:40 | from running ad copy tests as long as |
| 00:19:39 | 00:19:42 | you make sure that you have a good |
| 00:19:40 | 00:19:43 | strategy going in you're looking at the |
| 00:19:42 | 00:19:45 | right metrics to determine what |
| 00:19:43 | 00:19:47 | performance is and then you're utilizing |
| 00:19:45 | 00:19:49 | that to inform future tests to make sure |
| 00:19:47 | 00:19:51 | that you're always moving things forward |
| 00:19:49 | 00:19:53 | there are quite a number of components |
| 00:19:51 | 00:19:55 | and considerations that go into ad copy |
| 00:19:53 | 00:19:57 | testing I think I've covered the main |
| 00:19:55 | 00:19:59 | ones that come to my mind in this video |
| 00:19:57 | 00:20:01 | but if you have any questions or any |
| 00:19:59 | 00:20:03 | follow-ups on any of these pieces or if |
| 00:20:01 | 00:20:05 | you have any additional considerations |
| 00:20:03 | 00:20:07 | that I didn't include in this video |
| 00:20:05 | 00:20:09 | about ad testing I'd love to hear about |
| 00:20:07 | 00:20:10 | it in the comments below thanks for |
| 00:20:09 | 00:20:12 | watching our video if you thought it was |
| 00:20:10 | 00:20:13 | useful give us a thumbs up below we |
| 00:20:12 | 00:20:15 | release a new video at least once a week |
| 00:20:13 | 00:20:16 | so if you want to get notified of when a |
| 00:20:15 | 00:20:19 | new one comes out be sure to subscribe |
| 00:20:16 | 00:20:22 | to the paid media Pros Channel |
| 00:20:19 | 00:20:22 | foreign |
