Advanced Pay Per Click (PPC) Program

Certification Training
4987 Learners
View Course Now!
40 Chapters +

Advanced Testing Techniques Tutorial

2.2 Introduction

Hi, I'm Brad Geddes the author of Advanced Google AdWords, the founder of Certified Knowledge and the PPC faculty chair from Market Motive. In this lesson we're going to look at some advanced testing techniques and other ways to increase your testing effectiveness.

2.3 Controlling Risk

Now with a lot of testing from the ad copy or landing page stand points, we said just put a second ad in the ad group and send half your traffic to one landing page versus another one. Or just take an ad copy, make a new ad copy, use the same landing page. And then, see which one does better over time. Now that's great if your ad groups are such as the one such as three, four and five in this list. They have some conversions, but it's not selling a huge amount of them as opposed to some of your top ad groups. Now with ad group one here, it makes up almost as many conversions as all the other ad groups in the list. It makes $88,000 in profit a month. So if you just make a new ad copy, essentially, you're putting $44,000 at risk. If your second ad bombs, that's a whole lot of money you could lose, as well as make if it were to succeed. So, there is a way, in AdWords, to use what's known as ACE, or AdWords Campaign Experiments, to help control risk.

2.4 AdWords Campaign Experiments

So, AdWords campaign experiments is defined at the campaign level. So it's in your campaign settings. With the experiment, you can define the percentage of times that you want your experiment to run as opposed to your control. So in the example where you made $88,000 in an ad group, you might look and say, we're willing to put 20% of our revenue at risk. Let's make 80% of our ads, our keywords, our landing pages, whatever we want to test. Let's make 80% of them our control data and only 20% our experimental data. So when you want to control risk, ACE, or AdWords Campaign Experiments, is a good way of doing it. So first, you go in and define the experiment. You put your name in, whatever you want to name the experiment. The split between control and experiment and you can put in start and end dates if you wish to, or you can just manually start it up. Then you'll determine what you want to test. It could be a different ad group. It could be different keywords. It could be ad copies. So you can go in and create a second ad group or second ad copy. In your test one, call it an experiment only. In your control one, call it a control only. So that way what happens is, when an auction occurs, Google looks and sees you have ACE running 30% of the time for an experiment. So then 70%, it will show your control. 30% will show the experiment. Now, for everything else you are not testing, make sure it's listed as both control and experiment. They'll be displayed as normal. So then once you set these, you'll see at your ad group or keyword level all the different icons based upon the status of that particular ad group, keyword, ad copy, so forth. So it's really easy to do this at the ad group level. Just duplicate your ad group, write test ads in the new ad group. If you want, you can also do this at the ad copy level etc., setting. So once you've then defined, here's our control, here's our experiment, and then this is everything we're not testing that's both control and experiment. Then go back to your campaign settings and launch the experiment. So you can launch all the changes, or if you want you can delete the changes as well, and then wait. Make sure you have enough information to make decisions based upon the data you have. Then once you have enough information, it's time to examine the data. So in your reports, when you go to the Segmentation section there's something called Experiment. When you segment your data based upon experiment, you can see your typical information versus your experimental information. And then just within the previous sections we looked at ad copies or landing pages. Look at the data and make your decisions based upon your winners or losers. So the analysis of the data is exactly the same as we've looked through previously. What ACE lets you do is mitigate risk when you have very profitable ad groups, keywords, ad copy, landing pages, and so forth and you don't want to just put half of your revenue at risk. Use ACE in cases like that. The other time to use ACE is when you want to do ad copy testing and you're using Conversion Optimizer. because if you don't use ACE this is what happens. You put two ad copies in an ad group. And let's say one has a really good conversion rate and one has a poor conversion rate. Conversion optimizer does not serve ads, conversion optimizer just sets bids. So what happens is ad copy one gets displayed, it converts. So conversion optimizer raises your bid for the next time that keyword's displayed. And then ad copy two is displayed and it's got a much lower conversion rate, and it doesn't convert. So conversion optimizer lowers your bid. Then when even when you have rotate turned on in Google, it doesn't serve everything completely evenly. So that ad copy two gets displayed again, your poor converting one, and it still doesn't convert. So conversion optimizer sees it doesn't convert and lowers your bid even more. So then what happens is your ad copy one with a good conversion rate then eventually gets displayed. But its bid is so low, it's at the bottom of page one, it doesn't get clicked on, so you potentially lost a conversion. So conversion optimizer sets bids, the other setting, based upon your ad display. For rotate or optimize for either clicks or conversions, displays the ad copy itself. So if you are using Conversion Optimizer, use ACE for your ad testing. So the best way to do this is to duplicate your ad group completely. Same keywords, and then in your new ad group make your test ad. And that way Conversion Optimizer will see them as completely different ad copy queued relationships for the bidding purposes. So use ACE for Conversion Optimizer ad copy testing. Same for landing pages, same process. You're testing landing pages and use Conversion Optimizer. Make a new ad group, duplicate the keywords and ad copy. Just send your experimental data to your test page. Leave your control where it is. So those are the two main uses of ACE, mitigating risk and Conversion Optimizer bidding. Now here's another issue you run in to, is let's say you ran five tests, and you look at the data and you're not sure what should you choose. What is your best combination here? Because if we look at Test 4, our click through rate, fantastic, 10%. Nothing else is above 5.3%, so fantastic CTR compared to the others, but not necessarily a great conversion rate. It's actually one of the lower ones. So when we look at Test 5, well Test 5, fantastic conversion rate. 23%, much higher than any others. It's also the highest cost per conversion because its average CPC is so high because it's deemed not relevant. Well, okay, so we have Test 2. Test 2 is a great cost per conversion, but it's not a fantastic CTR, it's not bringing in as many customers. So which one of these is better for you, overall? It's tough to tell. Well, you could go by profit metrics. But the problem with profit is, Google doesn't serve everything evenly, so your impressions aren't equal. One test has 9000 impressions, one test has 5600 impressions. So you can't even use profit. So in cases like this, and profit's your most important metric. It's not conversion, it's not cost per conversion. Profit's what you really want to optimize for. There's a better metric to use to determine winners and losers. So let's go back for a second to have people search. We looked at this in our very initial lessons. We search because we want to know the answer to a question. So that's why someone searches initially. Now when they do a search, your ad's displayed. You get an impression, someone searched for your keyword, your ad was displayed. So essentially every single time your ad is displayed, you have a chance of a conversion. You chose a keyword, someone searched for the keyword. There's a chance of a conversion. So then at that point in time, someone gets to your landing page. And your landing page's goal is to instruct them how to get the answer and of course, get the conversion. So when we have all these mixed metrics, click through rates, impressions, conversion rate, cost per conversion. Let's boil it down to the most common denominator. The impression, every time your ad's displayed, you have a chance of a conversion. So, there's a very simple way to determine the winner. We first get rid of all the other metrics except for impressions and profits. Now, let's just take profits and divide it by impressions. That's how much money we make every time the ad is actually displayed. So in Test 3 here, we make $0.11 on the display, not the click, the actual display. Over the course of say 10,000 impressions Test 3 is going to make you $1100 in profit. Test 2 is going to make $800 in profit. That's a pretty decent percentage over time. So instead of dealing with mixed metrics where it's hard to choose, we have better CCRs, lower conversion rates, higher conversion rates, lower CTRs. Do a simple formula, profit divided by impressions. So when we look back to our chart here, Test 3 is our winner. It's not our highest CTR, that's Test 4. It's not our highest conversion rate, that's Test 5. It's not our lowest cost for conversion, that's Test 2. It's number two or three in all of these combinations, but together it makes more profit than anything else. So in this case, you also want to look at the outliers. So we wouldn't want it to stop here. You want to look at Test 4, why was it such a good CTR? Could we make another test, but try to borrow elements of Test 4 from a click-through standpoint? Now Test 5 has an unbelievable conversion rate. The problem with Test 5 was the fact that Google deemed it not relevant. Its average CPC was much higher than anything else. In this test, the bids were $5 CPCs. The averages range from $0.35 to $3.99. Really big difference, depending on what's considered relevant in Google's eyes. So we wouldn't want to throw away Test 5. Test 5's information with such good conversion rates, we want to see if we can borrow it. How can we make Test 3 a better conversion rate? So we save our Test 3 information. We make another one with Test 4 data to try to increase CTR. We make another one with Test 5 data to try and increase conversion rates and then continue our testing. So save your best test, learn from the outliers, then delete the losers and restart your test again. Now impression is good from the search standpoint because your ad's displayed. You chose a keyword, it was displayed. It's also good for placements because with a placement across a display network you actually chose a site. So you can look at how much money you make every time your ads displayed in that placement. But for the content network as a whole, when you're choosing keywords, and You're just telling Google, hey, place my ads based on my keywords. Sometimes the ad's below the fold, sometimes just reading the article, looking at an impression basis is really tough. It's actually done test on display where we broke a billion impressions. 1 billion impressions in a month, at sometimes with less than $12 in spend, because the CTRs worked just so well. So you can't really look at display, always on an impression basis. So instead when you're dealing with automatic placements, do it on a click basis instead. Look at your profit. Look at your clicks. Do simple division and look what you make every time you get a click. So to see the difference these range from $0.99 profit per click to $2.94. That's a huge difference in profit, and what you're making on a click basis. In fact, over just 100 clicks, the profit ranges from $99 to $294. That's almost triple the difference. So, looking at display on a click basis, useful. Now if you were using the standard workflow of run your place or performance reports, look to see where you have poor traffic, delete the losers. Look where you have good traffic, make them in a placement campaign. [00:13:08]Again once you're back at a placement campaign look at profit per impression for that exploratory campaign, or for automatic placements, you just want lots of visibility, use profit per click. So going into profit per impression or profit per click, will let you get down to a single metric, which just removes the interpretation of clicked rates and conversion rates. It's a much better way of doing testing when you really care more about profits than anything else.

2.5 What Makes a Conversion?

When we look at what makes a conversion, you have traffic source, ad copy, and landing page. If you change any one of these, then your overall conversion rates, cost per conversions, will change as well. Your traffic source is paid search in the lessons we're talking about. But you can do ad copy testing and as your ad copies change, they change expectations for landing pages. As your landing pages change, they have different ways then that they're convincing consumers to buy your products, but your ad copies may be setting different expectations than what your new landing page's message is. So often another test you can run, Is looking at the ad copy and landing page relationship. To see, how as ads change or landing pages change, your profit numbers and your conversion numbers change as well.

2.6 Landing Page Testing

This is again, a very simple test you can run at an ad group level. Take your ad copy, send it to page 1, duplicate it, send it to page 2. So that's how we do a standard landing page test. Now, make a new ad copy you want to test. Send Ad Copy 1 to Landing Page 1, duplicate Ad Copy 2, send it to Landing Page 2. So now really what we have is four combinations. We're looking at the landing page and ad copy relationships. Now, you can either analyze the data from a conversion rate, cost per conversion standpoint, or you can just look at your Profit Per Impression metrics as well. So, if you really want to get into looking at how individual ads and landing pages affect conversions that's an easy way of running the test. Just putting two ads in an ad group, run to page one and two, make a new ad, run to page one and two, and then look at the same data. Your interpretation of data and the amount of data you need to make decisions is still the same as we looked at previously. It's just another way to really start thinking through how are ads set expectations and how are landing pages meet those expectations. And how as we change these, we change the consumer perception of our product and what they're going to find. So it's another way of doing testing. Now the problem is I've shown a fair amount of decent size numbers. Sometimes you might have 10,000 clicks looking, at other times a few hundred clicks, but what if you look at your ad groups and you have a hundred clicks and 274 clicks and 400 clicks? Another was 74 clicks and very few conversions. In cases like this it becomes very hard to do testing at the ad group level. Then you might have this because either you don't have a big account you're only spending a thousand or $5000 a month. Or you have a big account you're doing a lot of long tail keywords it doesn't matter why you have, low traffic ad groups. If you have them, you can still do testing. Essentially our testing method changes a little bit. So first think about what you want to test. Is your Call to action, Unique selling proposition, your Customer benefit, different feature messages, a price versus a discount, what do you want to test? You're not going to be able to test everything at once. So prioritize what you really want to learn. If in doubt, start with calls to action. That's a good one always to start with.

2.7 Low Volume Ad Copy Writing

So then what we'll do is we'll take our ad group. We'll take our ad copy. And we'll write our ads as we normally would for our headline and description line one. In this case we're testing call to action. Description line two then, we'll take our ads, we'll put an ad copy one and ad group one and call to action one. We'll duplicate the ad. Well change just the Call to Action line 2. We'll go to another ad group. Might have a different headline and description line, because it's a different ad group. But our calls to action are always good to test, so we take the ad from Ad Group 2. And we put our call to action on the same one that we used in our first ad group. We duplicate that ad. We put in our second call to action. So now you're still writing your ad copies as normal based upon the keyword to ad copy relationship. All you're changing is the script's line two and a call to action. Now that's one way of doing a test. Another way is writing your static description line one and description line two and then writing different test headlines. So, essentially, all we're doing, then, is taking things we want to test, and we're getting more information by putting it in multiple ad groups at the same time. So then once we have that information, let it run for awhile

2.8 Combine and Analyze Data in Pivot Tables

and then we run our reports and put it in a pivot table. So the pivot table in Excel then will recombine the data so now you can just look at your cost per conversion. Your number of conversions, your conversion rate, even your click through rate, whatever metrics are important to you. Write your description line ones, or by your headlines. So if you write the exact same call to action across multiple ad groups, then make a different call to action, write it identically across those ad groups. Makes it really easy in a pivot table just to say combine all this data now across all the ad groups, but show it to me at the description two line which is in this case your call to action. So that's why you want to be exactly the same. Now, you could also if you wanted to work from profit numbers, do the same testing and just put in your profit for description line two Across these ad groups and divide it by the impressions. Or your profit forward description line two divide it by clicks if you're doing display networking. So the analysis of the data's the same for low volume accounts. All we're doing is increasing our volume that we have to analyze by using those same items and multiple ad groups. Independent ad copy testing is better. If you have a enough data that you can test ad one versus ad two, that's much better. But if you don't, it's still better to do ad copy testing even with low volume data, whether it's long tail or low budget accounts. So better to do some ad copy testing with this method than not do any at all. So it really doesn't matter how much data you get per month per ad group. All that changes is the way you analyze the data whether it's independent ad copies or whether you're just looking at combined data with a pivot table.

2.9 Recap

So just to recap, when you need testing control because you're using conversion optimizer, or you have high profit ad groups. Use ACE, average campaign experiments, set your test experiment, set your control data to control, and then set your other information to control and experiment. When profits your main concern and you don't really care about clicks, sync conversions and cost per conversion, you want profit. You want to put money in your bank accounts. Use profit per impression for search or placement tests. Use profit per click for automatic placement tests. And if you get into a lot of testing, if you test ad copy landing page combinations, there's a lot you can learn to see how expectations are being set from keyword to ad copy to landing page, because now you're truly testing the entire search process at once. When you have low volume accounts, your test ads or landing pages across multiple ad groups. So just make two unique selling propositions, put them in ads across six ad groups and recombine the data. So testing is essential for accounts. Only by testing can you make your account better. You have an initial account and that's your data, and maybe you've made some guesses at keywords and ad copies and landing pages, but that's just a guess. So first up, you set your goals, what do you really want to accomplish? That's important, next then, start your testing, keywords, ad copies, landing pages. Use the metrics that are most important to you, which is what your goals were for. Whether it's profits, click through rates, conversion rates, cost per conversion, number of conversions. And your goals will tell you which one then is the winner, learn from outlier data. Devise your new task, delete the losers, start again. So it doesn't matter if you have low volume accounts or high volume accounts. You need to do testing, all that changed is the method. But only by testing can you really learn from your paid search accounts, learn from searcher behavior so you can just increase the effectiveness. At all of your marketing over time based upon all this data you're getting that you can only acquire through running tests.

Find our Advanced Pay Per Click (PPC) Program Online Classroom training classes in top cities:

Name Date Place
Advanced Pay Per Click (PPC) Program 5 Jun -20 Jun 2021, Weekend batch Your City View Details
  • Disclaimer
  • PMP, PMI, PMBOK, CAPM, PgMP, PfMP, ACP, PBA, RMP, SP, and OPM3 are registered marks of the Project Management Institute, Inc.

Request more information

For individuals
For business
Phone Number*
Your Message (Optional)
We are looking into your query.
Our consultants will get in touch with you soon.

A Simplilearn representative will get back to you in one business day.

First Name*
Last Name*
Work Email*
Phone Number*
Job Title*