.png)
Landing page optimization with Google Optimize
- sherine mos@
- Dec 8, 2022
- 21 min read
Updated: Dec 9, 2022
If you are involved in marketing, you must have heard of the term optimization or A/B testing. If you haven't heard and came here to understand how to do it better - don't worry, we'll explain everything soon.
In this article, we will see why, although it may be seen as "advanced marketing", landing page optimization is an important and fundamental part of any marketing campaign. More than that, you should make it an integral and permanent part of the marketing strategy for your business, certainly, if it is a digital marketing campaign!.]
Table of Contents
Why do you even need to optimize landing pages and websites?
It's simple: to improve conversion rates and ultimately make the business grow and grow. We cannot predict the growth that a certain change can do for us, until we run a test to check it.
One of the first and most famous tests was performed by a software engineer who worked on Microsoft's Bing search engine. He proposed making a small visual change to how the headlines look in sponsored ads in the search results. This change waited for 6 months, because the project managers defined it as less important, compared For the thousands of changes they had on their waiting list to commit to.Fortunately, since this particular change was small and could be done within a day or two of development, another engineer decided to promote it and run an internal experiment within the company.
Within a few hours of the launch of the experiment, the new titles already produced a huge profit compared to the existing titles (a 12% increase - equivalent to an annual profit of 100 million dollars, just in Bing searches from the US!). These excessive results caused that it was initially thought to be a bug, But after verification, it just turned out to work better. It's important to note that without this experiment, that engineer's small change probably would never have been implemented.
Today, large companies such as Microsoft, Amazon, Booking, Facebook and of course Google run about 10,000 tests a year!
At Microsoft, optimization processes have been implemented in the products, so almost any change that you want to introduce must first go through a series of tests, to make sure that it actually improves performance in an optimal way. From: The Surprising Power of Online Experiments
If the big players think this is a wise move, could it be good and beneficial for us and for you too? Can be. Even with more modest amounts of traffic than Bing or Google gets, still, it would be a worthwhile move.
Ok, we got it, fine. But why, if we have a certain hypothesis about a change on a landing page that can boost our conversion rates, can't we just execute it? Why do we need all the tests, statistics and tools that complicate our lives?
Because speculation is cute, but, by virtue of our role as marketing professionals, and in order not to throw marketing budgets in the trash, it is our duty to base ourselves on objective and measurable data when we create properties and campaigns for businesses.
Another reason why you should run tests regularly is that when you run optimization tests - you are automatically granted a significant advantage. No, this is not an award from the Israeli optimization organization or anything like that, but simply a way to open a significant gap in the industry in which you compete:
It turns out that even in 2020, most of those who do digital marketing and those who already know the concept of optimization, still do not run tests, certainly not continuously or with high frequency. From data in a report produced by a company that produces a builder for landing pages, it appears that only a tiny percentage of all marketers use the A/B testing tool built into their product.
Surprised? The truth is that it surprised me. After all these are digital marketers, right? Why the hell aren't they running tests?!
At first, I thought it might be a specific problem with this builder. Maybe the optimization mechanism for landing pages built into it is unfriendly and difficult to operate. Maybe we should invest more in the "education" of the user audience (through articles and tutorials) to get them to run more tests.
It turned out that this is not a phenomenon unique to a certain builder. After examining the market of builders in-depth, it turns out that the percentage of advertisers who perform optimization tests for landing pages is poor for all of them! The vast majority of website owners and online marketers do not perform optimization tests to improve results at all, and even those who do, mostly run a small number of tests and then... stop touching it.
Did you know, for example, that the most popular builder in Israel today has no built-in landing page optimization mechanism at all? I asked the technical support how I still check and improve the conversion rates on the pages I build. They told me that it would be best for me to use... you guessed it - Google Optimize.
Apparently, this feature is not important enough, because there are not many people who attack it and run tests regularly. the truth? You can understand the crowds. As a marketer, after I've already closed a client and given him an excellent website or landing page with a sponsored marketing campaign that directs quality traffic to him - the preference of most of us is to deal mainly with the ads, and not with the page itself.
That's exactly why it pays to belong to this handful, which is able to rely on tests to raise conversion rates above the average in various industries. This is the way to excel, even in 2020!
If that's not enough to convince you to optimize for landing pages, maybe it will help if I say that it's really easy in the digital world?
It's so easy, if you keep reading, I guarantee you'll feel your palm come closer to your forehead by itself...
For example, unlike digital marketing, offline it is much more difficult to produce such balanced and objective tests. Soon I will share an interesting experiment conducted by a company in England, which explains exactly why it is so difficult to do outside the Internet.
Why is Google Optimize the ideal tool for marketers who want to squeeze more conversions from the landing page?
There are a variety of tools that aim to help us perform optimization tests for landing pages. In this guide, I have chosen to focus on the free tool from Google called Google Optimize, or optimize for short.
Google Optimize is not necessarily the best tool on the market today, but it certainly has several advantages over other tools. In the end, when we come to perform experiments - we draw the conclusions and take the steps reflected in the results. Therefore, the direct and easy interfacing of Optimize with Google Tag Manager and with Google Analytics is among its most prominent and powerful advantages for marketers.
In addition, Optimize is a tool that allows you to perform optimization tests for landing pages and websites while running. The intention is that you can set up any experiment you want and start running it on a page that is already live, without interfering, or interfering with its base code. This is a feature that has mainly advantages, but also one weakness that is important for us to recognize. We will talk about her later.
What is the optimization methodology for landing pages that suits my page?
In most cases, the tests in Optimize are conducted using the most common and relatively simple method, called A/B testing. In order not to deviate too much from the focus of this article, we will not go into comprehensive explanations of what A/B testing or other optimization methods are. We will explain in detail:
In the A/B testing method, the traffic that enters the website or page is taken and a percentage of it (usually 50% to begin with) is shown the original version and the remaining percentage is shown a different version with some change. That change should be due to some idea that we think may lead to an improvement The conversion on the page (directly or indirectly).
It is important to note that the additional version we will examine will contain only one and only one change. no more! Otherwise, if there is an increase or decrease, it will be difficult for us to know what exactly caused it, and then... we actually did nothing and the test will go straight to the trash.
How and when do you know who won?
How long should we run our experiment to be sure we got a winning version?
The issue of time can be very flexible because it is not the determining factor in tests.
What does determine? The amount of traffic or surfers. It is important that enough surfers are exposed to your experiment, in order for the results to be considered significant enough for us to consider them.
How do you know how many surfers are needed for our test to be correct?
You can use this efficient calculator. Of course, if the amount of traffic to the page where you run the experiment is small, it means that the experiment you run will take longer to reach the finish line. Flowing a lot of traffic on a daily basis? You will probably finish it in a relatively short period of time.
*You may find that you are close but do not quite reach the minimum amount of traffic required (usually this is a function of the size of the advertising budget you have). In such cases, you can run experiments with less than the recommended amount of traffic, you just have to take into account that the statistical margin of error increases.
After the period you set in advance for the test to run (the default in Google Optimize is 90 days), check which version converted better. If she met the required conditions - she will be declared the winner. Then you can make it the version that is shown to all visitors to the page -> 100%.
There are additional and slightly more complex methodologies, which we will not go into in this guide:
Sequential A/B testing - this is A/B testing in several steps derived from one another in sequence. That is: after we have received the winner of the two options - we take it and test its performance against another change that we want to test, and so on again and again... until finally, we decide that we have received the optimal version that will generate the highest number of conversions on the page.
A/B/C/D testing - with it you can try to shorten the steps in the process. With this method, we create a larger number of versions (variants) and present them to our audience in an equal distribution. In this way, instead of dividing the traffic into 50% and 50% between 2 versions, it may for example be divided between 4 variants with 25% traffic for each.
To work in this way, we must have relatively large traffic, otherwise, it will be difficult for us to achieve statistically significant results with so many versions.
It is worth mentioning that both the basic A/B testing methodology and the two more complex methods have one main limitation: they will not help us understand the relationship between changes we make to several elements on the same page. In order to understand the real effect of changing components on the site and the relationship between them - you should use an even more complex methodology in terms of statistical analysis, which is called:
Multivariate Testing - the method is based on the principle of A/B testing, only it compares more changes of elements on the page, at the same time. For example, combinations between changes in titles, images, colors, buttons, etc. The traffic to our page will be divided between different versions of it, and the goal is to test the effect of each combination of changes, and understand which one works best in terms of conversions.
For example, suppose there are 3 changes to different elements on the page that we want to test which one will work best:
Combination of 1 title + 1 image + 1 button
Combination of title 2 + image 1 + button 1
Combination of title 1 + image 2 + button 1
Combination of title 1 + image 1 + button 2
Combination of title 2 + image 2 + button 2
and so'…
At the end of that test, we will know not only which combination of changes works best, but also which elements have the most positive or negative effect on the panel (Funnel) where we lead the visitors.
The main disadvantage of this landing page optimization methodology is the massive amount of traffic that is required to complete the test. Most of the pages do not have enough traffic for it, therefore, they usually settle for simpler A/B testing.
Necessary conditions for A/B testing with reliable results
Even if you go for a basic A/B test, you have to remember that there are certain conditions that must be met, otherwise, the test may show results that we are not sure we should rely on.
1. The amount of traffic
Even with the simple A/B testing method, you need a minimum volume of traffic per page, to be able to determine decisively that our hypothesis is true, false, or just doesn't change anything. If during the test a minimal amount of traffic did not flow to the page, it would not be advisable to make changes based on the behavior of too small a number of visitors.
2. Traffic diversity
It is worth making sure that we perform our test on a diverse and random audience. Why? story time:
The sales agents of a company that performs insulation for attics in Great Britain distributed a flyer in the mailboxes of houses with a sale: for only 800 pounds you can insulate the attic. A saving of £400!
Despite the serious discount, only a small number of families purchased insulation for their attics.
They sat in the marketing department and thought about how to make the offer more attractive, without cutting the price further. A proposal was made to conduct an experiment in another neighborhood, and check how it affects sales. The sales agents distributed a new flyer:
"For just £800 we'll take all your stuff down from the attic, so you can sort through it and throw away what you don't need. While you're sorting, we'll isolate the attic for you. When you're done sorting, we'll even put back into the attic whatever things you decide to keep!"
Seemingly, a much more invested and profitable service for the customer, and at the same discounted price, but in fact, there is no additional effort here. The attic insulators would perform the exact same process before: remove all the objects from the attic and bring them back up when the job is done. They basically sold people an opportunity to insulate the attic and on the way make orders there...
The experiment worked well, and the same company did close more installations in the same neighborhood. The problem is that even though the experiment seems to have lifted sales, we have no real ability to know what actually happened here!
As mentioned, the testing of the new proposal was done only in one particular neighborhood in the UK. We have no information about the residents of the neighborhood who received the "more attractive" offer for their mailboxes. Is it possible that the demographic composition or the socio-economic situation in this neighborhood biased the results? Maybe this is a neighborhood with many new houses that still have no insulation at all? Could it be that this is a neighborhood that would have bought in high percentages anyway, regardless of the new offer?
As digital marketers, we enjoy a great advantage in this respect. The fact that we conduct experiments online solves this problem through randomization of the target audience. In the A/B testing methodology, we apply our experiment to the entire population equally and randomly: 50% of everyone exposed to our experiment will see the new version and 50% will receive the regular version. In this way it is very unlikely that the results will be swayed by external factors among the audience.
3. statistical significance
Conversion rates that are similar or too close cannot provide us with a good enough reason to make a permanent change to the landing page. In order to know what is difference is statistically significant for us, we will have to take into account the amount of traffic we get to the site.
For example, would you agree that a test that showed an improvement or decrease of 0.5% in the conversion percentage on a site with a million surfers a day is different from a 0.5% change on a site with 10 surfers a day? It would not be correct to rely on the behavior of too few surfers as a conclusion that will dictate how we will build our landing page.
How do you set up tests in Google Optimize?
So how do you set up tests to test ideas for changes that will increase conversions on the page?
Go to Google Optimize and log in to your account. In Google Optimize, the account is divided into sub-accounts, with each account supposed to represent a business or project. In each account, you can create several Containers, as needed. Each such container can contain several tests that we want to run in the same project.
Under this account I created a container that I called Landing Pages. I will put all the different tests that we will run on the landing page under it, but, if we want to run additional tests on his website for example, I would create a separate container (which will be called Website) under the same account.
After I created the container it gets a unique ID. For now we'll put this ID aside, but soon we'll use it to connect the container to the tag manager account of that customer, so that the experiment can run and be measured. For now, let's choose a name for our experiment. It is recommended to give a name as transparent as possible, so that we can understand what it is about without going inside. It will be very effective when we accumulate dozens of experiments here...
For the purpose of the demonstration, Neitzer Test on the landing page of one of the clients of the advertising agency Oasis Media - an old catering company that is active in the entire center and south of the country.
What I am interested in testing in the test is the effect of a subheading that is next to the top form for leaving leads on the page on the conversion percentage (we will not go into the theory itself to maintain business confidentiality for the client). I will test this by creating a variant of the page that does not include the same title and compare the results. Accordingly, I will call the experiment: Contact Form Legend locations.
Then we enter the address (URL) of the page we want to test so that Optimize knows which page to load to the test editor.
A/B testing is selected as the type of test, and we click on Create to proceed to the settings of the test itself. Here we need to add a variant to test against our existing original page that is currently running.
As a reminder, the page we are dealing with right now is live and there are Google ads directing traffic to it, so we have to be careful with what we do, because it will have a direct impact on the money the client invests in the campaign!
I will add a variant that I will call Form Legend 2, which I will run as a test against the original version - Original. In principle, it is better to give our experiments transparent names, and not to call them by a generic name like: Variant 2.
You can see that Google Optimizer automatically assigns a weight of 50% of the traffic to the new variant and the other 50% goes to the original version. If I were to add another variant here, each of them would receive a third - 33% of the traffic, and so on.
Soon we will see how to edit and actually define the difference between the variants, but first of all, let's define who will see the variant:
If it is an entire site, and not just a specific landing page, I have the option to choose which pages my test will run on. Let's say I run a change in the main menu of the site, I can set it to run every time someone surfs to an address that contains the URL of our site, that is, on all pages on the site that are under that address.
In our case it is only one landing page, so it is not relevant to us. That's why I enter the exact address and define - When URL matches - that is, only when the exact address I entered is reached.
Below that I can choose which audience my test is targeting. If you decide to go into the resolutions of filtering the audience in our test, you can decide, for example, that only visitors who arrive through Google will be exposed to our test, because we know that they probably have a higher intent towards our offer, because they performed an active search on Google. I can also decide to duplicate the page and create a unique URL for it, to run the test only on those who reach this address. I can run it only on visitors from mobile, or only on returning visitors, or only on visitors from a certain country, etc. etc.
The possibilities are many, but we need good and specific reasons to filter the audience on which we will apply our test. If I don't change anything, by default, the audience will not be filtered and the test will run on all visitors to our page.
In addition, in the settings at the bottom, under Traffic Allocation, I have the option to choose not to reveal 100% of all traffic to our landing page for the experiment.
For example, if I am marketing a large and popular e-commerce site like Amazon - I would probably prefer to carry out tests and changes as often as possible, that is, I would not run the experiment on all my millions of surfers at once! If I have a million surfers a day to the site I don't want to take a risk that my experiment will cause them to convert less well, and possibly lose a lot of money in sales. Therefore, I can choose to apply the experiment only to a smaller part, say 20% of the total traffic:
Of course, if the results of the experiment are successful in terms of the goals we set - we will update the change accordingly, to all visitors.
How to use the Google Optimize test editor?
The optimization process for landing pages with Google Optimize's test editor is friendly and easy. This is a visual editor that works according to the What You See Is What You Get method - what you change and see in the editor is exactly what you will get on your page. This is the most common technology used by most of the products for building websites and landing pages on the market today, so probably if you have ever come across one of these, it will be easy and intuitive for you to work with the Google Optimize editor.
To edit the new variant I added, I will simply click on the blue Edit button to the right of the new variant:
Google Optimize will take me directly to the new variant edit screen, where I can simply select (click) the element on the page I want to change.
As soon as you select any element on the page, a menu will appear with a variety of editing options for you to choose from: you can of course change dimensions, spacing, font, colors, border, background, images and more. As much as you like.
In our case, we generally need to select the text of the subheading of the form for keeping leads at the top of the page and remove it. Therefore, after selecting the element I wish to edit, clicking on the blue Edit Element button in the edit menu (or right-clicking with the mouse on the element itself) allows me to remove the element completely: Remove.
that's it! In principle, we made the change we wanted, and you can also see that the optimizer editor lists the number of changes we made on the page - change 1. Now you can click on Save to save the change we made.
The common mistake of marketers who use Google Optimize (and how to avoid it)
Here it is important to note that I try to be very careful when I make changes to the page through the Google Optimize editor. Why? Because every change, but literally every change I make, is counted and a script is produced for it with lines of code that are injected into the page when loading. Each such script burdens the loading of the page a little more, and sometimes, if it is about drastic changes, may even create technical problems for us.
What do I mean by "drastic change"? Without going too deep into the technical side of building websites, each page is built with advance thinking and with a certain hierarchy that exists between the elements that make it up. If I start playing and moving too many components I might create conflicts that will break the page, and then we will pay for our loss.
This is one of the big drawbacks I mentioned earlier that Google Optimize has. This is a software that is built in a way that allows us to run experiments on an existing page that is already running in the air without changing anything in the original version, but this may have a price.
For comparison, other tools, which are initially designed for building websites or landing pages, with a built-in option for creating optimization experiments (A/B Testing) work a little differently. Usually, such tools will create a duplicate version of the same page, in which you can insert the changes you wish to test in an experiment. This way it will turn out that 50% of your visitors will be served the original copy of the page and the remaining 50% will be served the copy with the changes. In this way, each page loads cleanly and there is less risk that the page will fall apart, or that it will load slowly, just because we are running a test.
Bonus: selecting elements on the page for advanced users
As mentioned, this is a visual editor, so there is no real need to know how to write code in order to make changes and conduct tests. However, in order to get to advanced features, for example like editing elements that for various reasons cannot be seen just like that on the page, you need to know, at least in a basic way, what HTML is and how the hierarchy of elements on web pages works.
For example: not long ago I optimized landing pages in a popup window that pops up on a client's site the first time the surfer brings the mouse cursor closer to the button to close (Exit intent). When I wanted to make the change in the Google Optimize editor, I did not have the option to select the popup window, because it did not appear on my screen... In order to still be able to select the same popup window, I had to locate its specific ID within the site's code, and then Use the option of selecting an element by ID in the Google Optimize editor.
It is important to examine the changes on mobile as well!
Lots of surfers visit our pages through a mobile device, and accordingly, most websites serve them a mobile-friendly version. For this reason it is important that we make sure that the change we made in the variant looks good even on small screens, and that we did not "break" anything in the page view.
To switch to the mobile view, click on the button that currently shows Standard in the bar above and select the type of mobile device you want to test:
Assuming everything is in order - we are fine. If you recognize that the mobile display is "broken" due to the change - you must fix it and then save it. When you finish editing, you can click Finish and exit the editor back to the settings page.
How do you share the experiment?
You can share the changes I made on the page with the client or anyone else before the test goes live. To share I can click on Preview of the variant I created and then on Share Preview.
Just copy the link and send it to the customer for approval. Whoever receives this link does not need to be logged in to the account to view the change, he simply enters the link you sent and views it.
A few final connections and our experiment is ready for launch
Now comes the part where we need to make several connections between accounts, so that we can get all the green marks in the list that Google Optimize requires us to fill out. They only need to be executed once per container, no
Every time a new experiment is created!
Let's start by connecting our Google Optimize container to the tag manager account of that client. This is a necessary condition for our experiment to run, because Google Tag Manager is the one that actually injects the script that Google Optimize generates for our experiment into the page.
Inside Google Tag Manager, we will embed the ID of our container as the built-in Optimize tag:
In Google Tag Manager we are required to define a trigger for each tag, that is: what needs to happen for this specific tag to run? For Google Optimizer experiments we chose to run the tag without a specific trigger. Instead, we'll specify that we want it to run before each time the Google Analytics tracking code runs. That is, before every time the page is loaded in the browser of any surfer:
We will save the changes we made in Google Tag Manager and continue to connect Google Optimize to its corresponding account in Google Analytics.
It's a simple connection: enter the Google Analytics Property of the relevant account, to allow a free flow of information between our container in Google Optimize and Google Analytics. This way we can track the results of our experiments in perfect overlap with the Goals we defined in Google Analytics.
We will not go into the thick of Google Analytics here. I'll just mention that it's important to have defined goals. If this is something you do anyway - great! If not, it is worth defining them, because this is what will allow Google Optimize to examine the success of the test against the Goal we defined.
Assuming we correctly connected Google Analytics and Google Optimize - under Measurement and Objectives click to add a main Goal and then we can choose the most appropriate one for our experiment from the list of goals we have in Analytics.
Now we are really done with all the settings. To really start running the experiment on page visitors, all we have to do is click Start.
That's it... we're in the air!
You can enter our page from two different browsers (or from incognito browsing of the same browser) and see the two versions served alternately.
Part of the idea is to make sure that every surfer always sees the same version of the page (think how embarrassing it would be if you did an experiment that included changes in the prices of your product, and then the same surfer would see different prices every time he entered - not to the point!). For this purpose, Google Optimize uses cookies that are stored in your browser. If Optimize detects that a surfer has already visited the page, it will make sure to serve him the same version that he was already exposed to on his first visit.
Summary: Optimization guide for landing pages with Google Optimize
Google Optimize
He is a great tool! It allows us to perform tests that will help us improve the conversion rates of the landing pages to which we send traffic from marketing campaigns. In this way, we can make much better use of our customers' marketing budgets.
It is worthwhile for you, as marketers, or as business owners, to belong to the segment of digital marketers who run A/B testing regularly and routinely. Even in 2020, this is the way to excel and raise the conversion percentages higher than average!
Comments