Strategy essentials: Personalization

  • Updated
This topic desscribes how to:
  • Discover actionable Audiences to target for personalization
  • Plan and launch Campaigns that impact your key business metrics
  • Interpret and take action based on your results

Personalization is a powerful way to communicate empathically with your customers and tailor your business to their particular needs. A personalization strategy lets you identify segments of visitors with distinct preferences or needs, and then create targeted experiences for them.

This article provides a high level overview of the strategic decisions you will make when using Optimizely Personalization.

Just getting started on your personalization strategy? Here are a couple of questions you may want to consider first:

  • What key metrics do I want to influence?
  • What is audience segmentation and why is it important to leverage the data that I gather on my customers? Read this blog post to learn more about turning your analytics into a personalization strategy.

Research shows that U.S. customers expect a personalized retail experience. Read more about their expectations in this study.

Discover audiences

Audiences are groups of customers you want to deliver a targeted experience to. Conceptually, an audience is simply a group of visitors to your site who have something in common. The first step in developing an effective personalization strategy is to discover actionable audiences.

What is an "actionable" audience?

Audiences are only "good" if they are actionable. Two characteristics determine how actionable an audience is:

  1. Impact: Is this audience shaped by the qualitative and quantitative intelligence that drives your business strategy? Is there a clear message that you want to tell this group? Is the reach of this audience a substantial percentage of your overall traffic?

  2. Feasibility: Can you target this audience using our available audience conditions? Will you need to use an integration with a different part of your technology stack to access these technical attributes? We will provide examples of technical attributes that you can target below.

To develop audiences that are both impactful and feasible, drive your audience discovery process with the qualitative (user personas) and quantitative (historical data) data that is fundamental to your business.

We will describe both user personas and historical data below.

User personas

A persona is a qualitative description of a visitor to your website. It is a model that describes the needs, concerns, and context of that representative customer. This model directly informs the personalized messages that you will deliver in your campaigns.

Do you already have some personas built? You have a strong foundation for targetable audiences. Your next step is to explore the technical feasibility of delivering personalized campaigns to those personas in Optimizely Personalization.

If you do not have personas yet, consider the following questions about visitors to your site. These questions will help you flesh out actionable audiences, based on a qualitative understanding of your customers:

  • What actions can a visitor take on your site that are valuable to your business?
    This is the foundational question underlying all optimization. Define the business-critical behaviors that personalization will accelerate.

  • What are some of the motivations that drive visitors to complete those actions?
    Why do visitors take those actions? Break out the different rationales that visitors might have for completing key actions. Do these motivations align with different targeting conditions? Can you use these conditions to create an Audience definition?

  • What level of familiarity with a brand, product, or service is required to complete a transaction?
    Is there some barrier of brand awareness, that the product or service is right, or simple comfort that comes from exposure that your audience needs?

  • Do certain visitors exhibit a preference for a different kind of product or service?
    The motivation to complete a transaction and the messages you use to drive that motivation might vary depending on the product, service, or action that your visitor prefers.

  • What signals indicate that a person might be ready to complete a transaction?
    Change your messaging and CTAs based on your customers' basic readiness to take action.

  • How are a visitor's needs, experience, or expectations changed by previous experience with a competitor?
    How would your message change if you knew that the individual has interacted with competing offerings?

Ultimately, personas should enable your creativity when designing campaigns. They clearly demarcate the various needs, wants, and contexts of visitors coming to your page.

In isolation, personas are a powerful tool for shaping strategic audiences. By layering on insights gleaned from customer data, you can lay the foundations for a comprehensive audience strategy.

Analyze customer behavior

Customer data is information about how visitors have behaved on your site in the past. This historical data includes results from exit surveys and user testing studies, transaction records of purchase recency and frequency, and actions that visitors take on your site. Evidence-based customer behavior complements the empathetic human factors that shape your user personas. You can use them separately or together to drive your audience discovery.

The range of data available for audience discovery may be immense and potentially overwhelming. However, many of these data points are not actionable. Here are a few questions to guide your efforts to unearth actionable data about your customers:

  • Are there drop-offs in your conversion funnel that are more severe for particular device types, campaign visitors, or verticals?
    One of the easiest ways to segment your visitors is by cross-referencing key behavioral performance reports, such as a funnel analysis, with basic factors like location, device type, media campaign, new vs. returning, etc. This is an actionable opportunity to personalize your customers' experiences on your site.

  • Is there a cohort of visitors whom you want to exclude from promotions based on the number of times they have already completed a transaction or other action (for example, 6 website visits or 3 purchases)?
    Conversely, is there a cohort of relatively inactive visitors whom you want to incentivize more aggressively? In the world of database marketing, Recency + Frequency + Monetary Value (RFM) is a proven axis where businesses can effectively target campaigns to customers. How can you design a site experience that takes advantage of these three data-points?

  • Is there a "magic number" of key behaviors that indicates that a visitor is likely to convert on your site?
    For example, an Optimizely Personalization retail customer found that the number of times a visitor viewed the product category page to be a strong predictor of purchase intent. When a visitor views the product category page twice, she is much more likely to complete a purchase. That signal formed the foundation for a behavioral retargeting campaign that was highly effective at improving the site's conversion rate. In many verticals, returning visitors are much more likely to convert than new visitors, who are seeing the site for the first time.

By connecting data-driven insights (such as the fact that returning visitors are more likely to convert than new visitors) with user personas (new visitors are users who need to achieve familiarity with the site or brand before converting), you can customize messaging on your site to cater to those behaviors and expectations. In this example, we might show more educational content and CTAs for nurture campaigns to new visitors, instead of messaging that prompts the final transaction.


In Personalization, tracking components such as pages, tags, and events are always-on. They constantly gather data about how the visitors who come to your website behave. When starting with Optimizely Personalization, make sure to set up a robust net of pages, events, and tags, so the platform's ability to surface actionable audiences is maximized.

Concrete data about your visitors’ behavior is a perfect complement to the personas that model your most meaningful audiences. Together, personas and historical data provide a thorough picture of who is coming to your site, how they behave, and what they might be looking for.

Your next step is to connect these insights to technically available visitor attributes so you can target them in Personalization.

Target technical attributes

In general, your visitors can be identified through three types of attributes: behavioral, contextual, and demographic. Below, we discuss these attributes and provide some examples of how you might personalize to each.

Behavioral attributes

Behavioral attributes are groupings based on current or past behavior.

  • Search: What search terms tend to correlate with conversion events on your website? Are there important keywords that correspond to certain, important audiences?
    Strategy: Show content or a CTA that aligns well with the keywords you’ve identified.

  • Content: Are there certain types of content, formats, or products that your visitors viewed and abandoned?
    Strategy: Show or reference that content to those visitors when they return your website a second time.

  • Conversion abandonment: Did a certain segment of visitors enter the checkout funnel but abandon it before completing the process?
    Strategy: Customize your site experience for those customers based on near-conversion behavior. You might welcome them back to your site and show them the product they almost purchased.

Contextual attributes

Contextual attributes describe the environment from which your visitors access your site, such as device, time of day, or promotional campaign.

  • Device: Is a customer browsing your website from their phone?
    Strategy: Consider providing a phone number for the customer to contact you or create opportunities for social sharing on mobile.

  • Time of day: Have you noticed that on the weekend, your customers tend to spend more time browsing before making their purchase?
    Strategy: Surface a broader variety of products for your weekend visitors or present them with browsing categories that they have visited in the past.

  • Media campaign: Is your visitor clicking through from a certain promotional campaign?
    Strategy: Align some of the messaging on your landing page to that campaign.

Demographic attributes

Most demographic data is either integrated from your analytics platforms or inferred from your customers’ behaviors on your site. Demographic data can be highly actionable.

  • Location: Are your visitors in a certain metro area, in a rural versus an urban environment, or in a particular state?
    Strategy: Personalize the messaging on your landing page or your product suggestions for audiences in those locations.

  • Age, gender, and income: Do you know from your data management platform or your analysis of visitor behaviors that a visitor is interested in a certain type of product, based on age, gender, or income level?
    Strategy: Carefully adjust the messaging on your website to meet the expectations of those audiences.

By layering on the "qualitative" (persona), the "quantitative" (behavioral data), and the "what’s possible" (technical attributes), you can define highly actionable audiences to which you can deliver personalized campaigns.

Plan experiences

Now that you know for whom you want to personalize experiences, the next step in your personalization strategy is to determine what you want to say and where you want to say it.

A messaging matrix can help you develop personalized messages that resonate with each audience you have created and map them to different areas on your site. Learn more about planning experiences for a single campaign.

To deliver the best single experience to visitors who qualify for multiple audiences, prioritize your audience overlap.

To deliver multiple experiences to visitors who qualify for multiple audiences, use a multi-threaded campaign strategy.

Test your campaigns and experiences

After you map your campaigns in a messaging matrix, there is nothing stopping you from building and launching those campaigns right now. However, the data-driven decision-making that you used in A/B testing is still an important part of your personalization strategy. The main principle behind integrating testing and personalization is this: test multiple variations and launch the winners as personalization Campaigns.

Learn more about integrating testing and personalization processes in this article.

Even if you do not explicitly test your campaign designs before launch, Optimizely Personalization still makes it easy to measure results with statistical rigor. Optimizely Personalization uses a holdback to show you the difference between the original experience and your personalized experience, so you can track your campaign’s progress.

Core personalization tactics

Ultimately, a comprehensive approach to personalization strategy should ensure that your personalization effort is maximizing its potential. Below you will find three core tactics to help you get started. For additional inspiration, check out this article of personalization use cases.

Symmetric messaging

  • What: The landing page of your site should match the expectations that banner ads, paid search campaigns, and email campaigns set for your visitors. In online marketing, this is referred to as “symmetry.” Tightly tailor the look and feel, value propositions, and keywords of your site for each audience that arrives on your site from a marketing campaign.

  • Where: in messages, CTAs, and value propositions

Behavioral retargeting:

  • What: Show visitors products that they have already shown interest in a second time to encourage a purchase. With Personalization, you can display products to shoppers in the same session as they navigate from page to page. If a visitor has selected or searched a destination, price point, or a particular hotel, personalize the ongoing experience with promotions and messages that match this intent. You can also actively adjust the logic that powers your recommendations using this data to ensure every step in their journey is a great fit.

    If your business is comfortable with using discounts and promotions to drive purchases, you can use urgency messages or discounted cross-sells (bundling) to increase cart size and conversion rate. Show an incentive to visitors who have looked at a product multiples, returned to your site without purchasing, or added items to a cart without purchasing.

  • Where: search results, promotions, recommendations, tailored images, checkout funnel

Contextual empathy:

  • What: Adjust the tone, language, imagery, and navigation options that your visitors see to be more relevant and empathetic to their context. This can help to provide a smooth, high-value experience for your customers.

    A visitor is on a mobile device may have less time to browse or a greater likelihood to share on social networks. A visitor in a foreign country may have particularly interested in your store’s shipping policy. A new visitor to your site or app may not be as ready to complete a transaction with you; build trust by showing more informational material and proof points. Conversely, you might display a streamlined conversion funnel for a returning visitor.

  • Where: images, navigation, banners, recommendations

Run campaigns

Once you have discovered audiences and planned personalized experiences, you have built the essential elements of your personalization strategy. You are ready to run campaigns on your site and monitor the results.

Monitor your Campaigns

As a campaign runs, Optimizely Personalization tracks events and pageviews and uses them to calculate how well personalization is working. You can see how well a campaign is performing on its Results page.

The Results page lets you answer four key questions:

  • Overall, how is the Campaign doing? The top-level summary tells you how much your campaign is helping or hurting. It finds the difference in conversion between all the visitors who saw a personalized experience that was delivered by the campaign, compared to visitors who were eligible for an audience but were bucketed into the holdback.

  • How large are my Audiences? The Results page provides a chart that shows how large each audience as a percentage of all visitors eligible for a campaign. You may have a particular experience that performs well but is shown to a very small audience, so the campaign's overall impact is small. In that case, you might add additional audiences to the campaign or expand the definition of the audiences you already have. For example, you could broaden a "Shoe Shoppers" audience from visitors who bought shoes in the last week to also include visitors who browsed shoes in the last month

  • How is each Experience doing? You can break down your overall campaign results by its individual audiences. Sometimes, a campaign appears to neither increase nor decrease lift. But within that campaign, some experiences will perform much better than the baseline and others will perform much worse. If an individual experience is underperforming for an audience, you can change the experience, then publish it for those visitors and see whether lift improves.

  • How can I extend my personalization strategy’s reach and impact? In Recommended Audiences, explore how your different segments behave differently than one another, and discover new audiences to extend the reach and impact of your campaigns.

Take action on results

Unlike in testing, there is no natural breakpoint to take additional action—such as when an experiment reaches statistical significance. Establish a framework for monitoring your running campaigns to help you decide when to pause, stop, or iterate on an idea. Share results with executives and optimization stakeholders to keep them updated on the progress of your campaigns.

Read more about those decisions in this article.