graphical user interface

Remember the first time you ran an A/B test? You had probably heard of this powerful technology that would improve your site and make you more money, but you weren’t exactly sure how to get started. So you read up on the basics and eventually, after some trial and error, you got that first test running. Since then each test got easier. Today you’re running tests on the regular.

Except now all of those feelings are flooding back to you with the buzz surrounding the newest optimization technique — personalization. You keep hearing you should do it, but how? Where do you start? What does it even mean?

This article will cover the basics so that you can start personalizing your site today. And with a little practice, personalization will become just as second nature to you as A/B testing.

What is personalization?

Let’s start at the beginning: personalization is showing targeted content to specific segments of your visitors. Personalization comes in two flavors: algorithm-based, and rules-based.

Algorithm-based personalization uses a machine learning algorithm to automatically decide what content to show each visitor. It makes this decision based on a person’s demographic information and past behavior, among other signals. Netflix’s recommended movies are an example of this type of personalization.

With rules-based personalization, you create targeting rules to segment visitors, and decide what content to show each segment. For example, you can show return visitors a message welcoming them back. The experience can also be dynamic, changing based on each person’s behavior. This is the type of personalization that Optimizely makes really easy, and what I’ll focus on for the remainder of this article.

The Building Blocks of Personalization

To personalize your site, you run campaigns to show targeted content to your visitors. There are 3 pieces of information that every campaign needs:

  • Who: which visitor segments (also known as audiences) are you targeting?
  • What: what content are you showing each audience?
  • Where: which page(s) will you show personalized content on?


The shoe audience likes shoes, and the hat audience likes hats, so we run a campaign to give each of them what they want.

To decide the who, what, and where of your campaign, you should create a hypothesis for why a personalized experience will increase conversions. Just like with A/B testing, a good hypothesis is the key to creating successful campaigns that are good for both your customers and your business.

In order to form a good hypothesis, and to measure the success of your campaign, you need a primary metric to optimize. You can (and should) measure multiple metrics to gauge the health of your campaign, but you should have a single metric that is the ultimate measure of the campaign’s success.

To measure the success of your campaign, Optimizely Personalization uses a holdback. A holdback is a small percentage of visitors (5% by default) that see your site’s default experience instead of the personalized one. We then calculate the improvement by comparing the conversion rate of the personalized experience to the default experience. By doing so, you always know the impact of your campaign, and can modify or stop any campaigns that are underperforming.

Once you’ve got your hypothesis and primary metric, you can bring everything together with this hypothesis template:

By showing [what] on [where] to [who], [primary metric] will increase because [rationale].

To learn more about coming up with campaign ideas and good hypotheses, this article on A List Apart is a great resource.

Campaign Setup

With a good hypothesis to guide you, it’s time to translate your ideas into an actual campaign.

Who: Audiences

The first step of setting up a campaign is to choose a group of audiences to target. You should aim for a group of audiences that match the following criteria:

  • High reach: Your campaign should reach a large percentage of your site’s visitors to maximize its impact. A campaign with a 10% lift sounds great until you realize it only reaches 5% of your traffic, which translates to a 0.5% overall increase in your site’s conversion rate.
  • Mutually exclusive: A good group of audiences overlap with each other as little as possible. Visitors can technically belong to multiple audiences at once, which can make it difficult to control the content a person will see. It can also lower your campaign’s reach. If audiences do overlap, however, our product will ensure people only see one experience per campaign. One technique for making mutually exclusive audiences is to choose a visitor attribute, such as location, and create an audience for each value (e.g. west coast, midwest, and east coast).
  • Not too many, nor too few: The more audiences you want to target, the more complex your campaign will be. You’ll have to create rules to define each audience, and produce content for each one. But too few audiences may lack impact because the content isn’t specific enough.

Defining Audiences

To create an audience, you need to define targeting criteria so that the product can determine if a visitor belongs in that audience or not (this is known as bucketing). The easiest audiences to create are ones that use the tool’s built-in targeting criteria, such as browser and location. These criteria are basic, but they don’t require any configuration to use. View the full list of Optimizely’s built-in targeting conditions in the knowledge base.

A more powerful technique is behavioral targeting, which enables you to define audiences based on behavior. A person’s behavior is often an indication of their intent, which unlocks a lot of potential to define really creative audiences. You can take a fuzzy audience, such as “Fancy bag shoppers,” and define it as people who view products with a category of “Bags” worth more than $150 at least 3 times in the past month.

In Optimizely Personalization, you can set up these audiences without writing any custom code. Our visual editor allows you to create Events and Tags with a point-and-click interface, which you can then use in our drag-n-drop audience builder.

graphical user interface, table

”Fancy Bag Shoppers” audience targeting criteria

Finally, you can use integrations to create audiences based on data from third party providers, such as Salesforce, Marketo, BlueKai, Demandbase, or your own data warehouse. You can also upload custom list of visitors to target. These options require some setup and technical knowledge to use, but they let you use data that you can’t get elsewhere. Read about all of our integrations in the knowledge base.

What: Content

Creating content that engages your audiences is vital for a successful campaign. Content that enriches the visitor’s experience, or that helps them accomplish a task, are great options. Once again, the article on A List Apart that I linked to above is a great resource.

Once you know what you want to show people, setting it up is a breeze. Just like with A/B testing, you can make a ton of changes using our visual editor, such as text, links, and images. And if you’re more technically-minded, you can write HTML, CSS, and Javascript to implement just about any change you can imagine.

Finally, a word of warning. A lot of personalization strategies recommend running price campaigns, which offer discounts or coupons to audiences as a way of enticing people to complete a purchase. These campaigns can be highly effective, but they’re complicated to set up. You can’t just change the text of the price element – you also need to tie that change into your backend billing system. Unless you have experience running pricing A/B tests already, I don’t recommend starting with a price campaign.

Where: Pages

With audiences and content ready to go, all that’s left is to set up your campaign’s pages. The best pages are ones with high traffic and a clear call-to-action. High traffic is important so that your campaign will have a measurable impact, just like with A/B testing. And a call-to-action is important because it makes that impact easy to measure. Even if your ultimate goal occurs down funnel, you should measure the intermediate steps to get there so that you have a complete picture of your campaign’s effectiveness.

If you aren’t sure which pages to personalize, your homepage, landing pages, and checkout pages are always great choices. You should also consult your analytics tool to see which pages are your most popular.

Setting up Pages in Optimizely

Pages are a new concept in Optimizely Personalization. They define the places on your site that you can run campaigns on, and they track visitor behavior.

You can target pages in Optimizely by using URLs, or by creating a URL pattern. Read more about URL targeting in our knowledge base.

When creating Pages, you also create Events and Tags to track in Optimizely’s visual editor. Events track a visitor’s behavior on your site, such as “clicks add to cart” or “newsletter sign ups.” Tags track metadata about each Event, such as the product category. You can see this data in your campaign results, and use it to define behavioral audiences. Once they’re set up, they’re tracked all the time, regardless of whether a Page is part of a campaign or not. Learn more about Pages and Events in our knowledge base.


The “Item” and “Price” are tags that will be recorded with the “Purchase” event.

Once you’ve defined the who, what, and where, your campaign is ready to run. After it’s published you can modify it at any time to add or remove audiences, change the content, or stop it altogether.

Now that you’ve read about the basics of personalization, you’ve overcome the initial barrier of not knowing how to begin. That fog of uncertainty that was preventing you from starting should be lifting. Now build on this momentum and personalize the web.

Personalization Resources

To learn more about personalization, the resources below are great starting points.