This article is a textual and pictorial version of the presentation I made for the Google digital workshops at the eCom fair in Geneva. It presents the key concepts of website testing, the presentation of Google Optimize and its use from A to Z …

Key concepts

All site owners have one problem in common: ” how to improve the results generated by my site?

google loves me workshop banner

Indeed, in 2020, there is no longer any doubt for anyone that a website must be a weapon of massive business generation 🙂

1. The base: bringing traffic

The first step for any site owner is therefore to bring traffic (qualified).

It usually starts with natural, social (free) traffic and a little referral.

For the more daring, this may also include some paid traffic (Adwords and / or for example Facebook).

2. Analyze and understand the segments

The second step is to analyze traffic who comes to the site.

This usually goes through the Search console, to understand what searches our site is on, on which pages.

But also (and mainly) by Google Analytics in order to be able to analyze the best performing segments and understand which resources on our site generate the most engagement and at the end of the day the most results.

(I assume that objectives have been defined in Google Analytics)

Why ? Just because relevance is key.

From my experience, the relevance between the research intention (and / or the characteristics of the audience) and what is offered on the landing page is one of the major performance criteria. (credibility / trust is another one by the way :-))

So, being able to identify specific segments (rather than looking at average traffic), to study these segments in order to better understand them (we are investigators!) will allow us to be more precise and more relevant.

Which brings us to the next point.

3. Improve your acquisition strategy

This is usually the next step.

Although improving the business performance of a site is an equation with acquisition on one side and landing pages on the other, the majority of people prioritize (if not only) traffic optimization.

This is how we will start to segment the traffic and send it to different more specific pages (or even create new ones).

It makes sense (off course!). But we find quickly limits.

  • How to go further ?
  • How can we improve performance on a given page for a given segment?
  • How to be more relevant with the context of visitors? Especially since the context is quickly hundreds of parameters …

We are not going to create hundreds of pages…

This is where Google Optimize comes in 🙂

4. Test the elements of the pages

With Google Optimize, you can start testing for a given segment the different elements and / or messages of your page dedicated to this segment:

Instead of creating 3, 4, 5 pages, you create as many variants in 30 seconds.

5. Personalize according to the context of the visitor

You can also personalize your messages / elements according toa multitude of elements of the visitor’s context.

If for example the user is:

  • On iPhone vs Android smartphone;
  • On a mobile or computer;
  • In Paris or Geneva.

And these are not that the basic options.

Thanks to integration with Google Tag Manager and access to data layer variables, we can do incredible things (the subject of a more advanced article project).

Introducing Google Optimize

Let’s be clear. Google Optimize is a relatively young product compared to other players that have been around for years.

On some features, it may be a little less complete or lower (but not that much).

However, the enormous strength of the tool is being fully integrated into the Google ecosystem. A huge asset. (and it’s free too :-))

Thanks to this, we respond to 3 major challenges related to optimization:

Data consistency

In Optimize, you use the metrics and objectives that you have already created in Google Analytics to measure the success of your experiments (you can also create them in the tool if necessary).


Personalize the experiences of your visitors according to the device, geography, provenance, behavior on the site, audiences * …


Integrations with BigQuery, Google Analytics, Adwords, Google Tag Manager allowing advanced segmentation of test results, advanced targeting for experiments and end-to-end optimization for Adwords campaigns.

How to set up Google Optimize on my site?

1. Create a Google Optimize account

Go to

Generally, one account is created per company.

You can create multiple Optimize accounts with the same Google Account. And several containers per Optimize account.

2. Create a Google Optimize container

Generally it is about creating one container per website since you connect Google Analytics and Optimize at the container.

3. Associate your container with a Google Analytics property

Choose the Google Analytics property of the site you want to test from the list.

Note: to make this possible, the Google account used for Optimize must also have Analytics access.

4. Obtain the installation code

After associating the Analytics property with the Optimize account, Google suggests the code to install.

5. Implement the Google Optimize code on your site

Recommended method: via Google Analytics

Google recommends installing Optimize through Google Analytics. You must add a line to the tracking code (analytics.js or gtag.js)

Install the Hiding Snippet (Google Analytics method only)

Additional code to be inserted before the Analytics tracking code to avoid the flickering effect.

For your information (or reminder) the flickering effect is materialized by the fact of seeing the original variant for one second before seeing the modified variant.

src=”▷-Google-Optimize-Full-tutorial-2018-Test-and-customize.png” sizes=”(max-width: 150px) 100vw, 150px” srcset=”▷-Google-Optimize-Full-tutorial-2018-Test-and-customize.png 380w,’écran-2018-04-25-à-15.49.43-300×248.png 300w” alt=”” width=”150″ height=”124″ />Via Google Tag Manager

You can also install Google Optimize via Google Tag Manager.

This requires a good command of the tool to implement without flickering.

The problem is that all the tags inserted by GTM are inserted asynchronously. It is therefore difficult to manage the trigger order.

And the Optimize tag must be triggered first, at the earliest and before the Google Analytics tag.

Personally I almost always implement with GTM. (in principle :-)). I give you my secret to make it work.

I assure myself :

  • That the GTM container is placed immediately after the opening tag;
  • That the Optimize tag is the one with the highest priority (I put 100 in general).

And there is no problem. No need for hipping snippet, no flicker 🙂

Step-by-step video implementation

History of not reinventing the wheel, I link you the video of Measureschool.

It shows step by step the implementation of the Optimize code for the 3 methods: here.

Create your first test

After installing the code on your site, you are ready to create your first test.

Excited? Great, let’s go 🙂

Start by clicking on the blue button to create a test.

Then, in order:

  • Give a name (explicit) to your test;
  • Enter the URL you want to test.

Important: This is the web page that you will use to create variations.

But based on the changes you’re going to make to a template page, you can then apply that to a set of pages of the same type. History of learning viiiiite 🙂

Example: I take a specific product sheet to make my modified variant. Then, during targeting (which we will see right after), I apply that to all product sheets.

  • Determine the type of test;
  • Click the blue Create button.

Break: understanding the types of tests

A / B / N Testing

Compare 2 or more variations of a page to see which one performs best.

Ideal for optimize a page and improve in a granular way.

Example: I have an already segmented page that receives targeted traffic and I want to “fine tune” its performance by testing different messages and / or visuals and / or deleting or adding elements.

Multivariate tests

Test multiple item variations on a single page for granular optimization.

Same principle as above. We test more things at the same time, but we learn more slowly too.

Depends on your goal (learn quickly or test more things).

Redirect tests

Test big changes (example: a page redesign) or different paths in a tunnel by testing one URL against the other.

Define your goals

  1. Choose a Google Analytics goal (from the associated property) or create one directly in Optimize;
  2. (Optional) Add one or more secondary objectives;
  3. Formulate a clear hypothesis. In a month, when you look at the results of your various tests, you may not remember exactly what you were looking to learn / demonstrate. Very important to inform the hypothesis 🙂

Define your targeting

Define the rules for applying your experiment. The possibilities are many 🙂

It can be as simple as different variants for different devices or different places.

But it is also possible to be ultra advanced with the use of cookies or Datalayer data variables.

The use of cookies can for example bypass one of the main limitations of the free version of Optimize: not being able to target remarketing audiences (functionality currently reserved for Optimize 360).

For more details on each targeting, see this page.

Tip: As mentioned above, URL targeting allows you to apply your changes to a set of pages (of the same type or not, although logically most of the time we do this on the same type of page).

Personally, I missed out at the start. This is why I insist.

Let’s take my example again: I made a simplified variant of the product sheet. And now I want to apply it to all of my product listings.

So I’m going to overcreate a rule and then I click on URL.

There i target all URLs that match my product listings. To do this, I can use different operators:

  • Correspond to ;
  • Equals ;
  • Contains;
  • Starts with ;
  • To end by ;
  • and Matches the regular expression.

To make sure you haven’t made a mistake, you have verification functionality.

Find out more about URL targeting here.

!!!Warning!!! : when making a modification on a set of pages, make sure you only make modifications to elements of the page template.

If you modify an element which is not an element of the template, but an element specific to a product file (example: the title of the product), this change will be applied to all pages and will apply this title to all product files.

Create a page variant

Now that you have defined your objectives and targeting, you can create your variants:

Note: to use the visual editor you must have the Google Optimize extension and Chrome (version 45 or higher).

Visual editor: create experiments without developers

The visual editor allows:

  • Minimum setup time (30 seconds may be sufficient)
  • A simple & intuitive interface: we select the elements and a dialog box allows us to simply modify a lot of things.
  • Edit texts, images, css styles, blocks
  • Edit HTML and javascript (advanced)

Please note, the visual editor is not in no case a page builder à la Unbounce for example. (And that’s a shame)

This is the second major limitation of the tool in my opinion.

This editor allows you to modify the elements of a page relatively simply, but as soon as you want to make slightly advanced modifications, it quickly becomes complicated.

Start your test

When you have defined objective (s), targeting (s) and variant (s), you can start your test:

See your test results

Track your test

In each test you have a reporting tab which allows you to follow your test :

When the test is finished, Optimize will notify you when a winning variant is detected. Or if the test did not yield a winner.

Tip: determine the duration of the test in advance

Never stop your test on an opinion or an impression.

I recommend using a tool to define in advance how many sessions you will need to accumulate per variant to ensure the reliability of your test (statistically).

Optimizely offers one here:

More information on the concept of statistical signifiance here:

Attention: The averages lie

Your test can (must) also be analyzed in Google Analytics.

Why ?

Because typically, the test gives you the average result for all targeted visitors.

You have to go segment test results in Google Analytics by applying segments.

Classic example: segmentation by device:

Warning : when you segment, you reduce (virtually) the size of the observed sample.

Your test may therefore not be statistically significant.

So you will have to accumulate more data.

Simple example: if you must have 2000 visits per variant to be reliable, you will also need 2000 visits per variant per segment.

Don’t waste your cartridges

In the free version of Google Optimize, you are only entitled to 5 experiments at the same time.

And a test takes time to get reliable results.

Do not waste your 5 cartridges to test the color of the buttons.

Test for maximum impact.

Some ideas :

  • Test your value proposition: can you be more clear? More profit oriented?
  • Try to reduce friction: how could you make the actions simpler? faster?
  • Motivation: how could you increase the motivation of your visitors to carry out the expected actions?
  • Relevance: in relation to the source of traffic, or the place of consultation or the connecting device, how could you make the experience even more relevant?

Additional resources

Official Google Optimize Help Center:

Diagnosis of Optimize installation problems:

Some excellent resources to optimize your site (understand the profession of Conversion Rate Optimizer independently of any tool):