Personalization studio

Enabling 1:1 Personalization on your site dynamically ranks products differently for each shopper, in an order that is most relevant to their preferences. This personalized product ranking makes finding products easier and can nudge the shopper toward making a purchase.

The shopper's affinities for products with certain attributes are determined by the shopper's interactions. These interactions can include product or category views, query searches, add-to-carts, purchases, and more.

For example, Bloomreach observes how a site visitor (new or returning) interacts with different products. If they tend to view and buy shirts from certain brands over others, then our algorithm will boost the products of the preferred brands in the search results of similar queries like “formal wear” for this particular shopper.

Enable 1:1 Personalization

⚙️

Note

The 1:1 Personalization feature can be accessed if a user has the “Personalization studio editor” role assigned to them.

Run personalization setup

1. To set up 1:1 Personalization for your site, go to AI Studio > Personalization studio.

2. Under the Setup tab, click the Set up personalization button.

   

3. In the pop-up panel that appears, select the catalog for which you want to enable personalization and click Start setup.

These steps will trigger the pipelines required to capture the user profiles of your site visitors for personalization. It takes 24 - 48 hours for the initial setup to complete, until which the catalog’s status will be indicated as setup in progress.

Once the setup is complete, you can enable the personalization experience for your shoppers across different levels of your site. This is explained in the following section.

Create personalization rules

Personalization can be enabled at the Global level, Query level (site search), or Category level by creating personalization rules.

1. Navigate to the tab of the level you want to create the rule for (Global rules, Site search rule, Category rules).

   

2. Click the + New rule button:

   

3. Select your catalog from the dropdown.

   

4. Turn the Enable Personalization toggle off to save a different variant of the rule with Personalization disabled. This helps with A/B testing the rule.

   

5. Here are the details of the further options for each type of rule:

   1. You can create a Global rule to enable personalization for All queries / All categories or both at once.

      

   2. You can create multiple Site search rules to enable personalization for specific search queries. This means that the search results for these queries will follow personalized ranking.

      Type and hit the return key after each search query in the Select queries field.

      

   3. You can create multiple Category rules to enable personalization for specific categories. This means that the products displayed in these categories will follow personalized ranking.

      Select the categories from the dropdown.

      

6. Preview and save your rule.

   

Preview and debug personalization

To better understand the impact and experience of personalization for visitors on your site, preview your rules before enabling them.

The Preview panel can be accessed by the button on the top right corner while creating or editing any personalization rule.

Follow the steps below to use the preview panel:

1. Enter the search query / select category in the search box for which you want to test.

2. Enter the User ID or the cookie2 value of a site user in the field next to the search box (highlighted in the image above). For testing purposes, you can fetch this value using the Bloomreach Pixel validator Chrome extension.

   

3. Review the search results in the Preview panel. On the left side, the top attribute affinities associated with the user that govern the personalized product ranking are displayed (and the number of interactions they made with the attribute).

4. You can view the exact impact on the product ranking in the following ways:

   1. Toggle the Show Changes checkbox at the top right. This will highlight the changes in position of the adjusted products.

      

   2. Switch to the Side-by-Side Preview tab to compare the search results with and without 1:1 Personalization applied.

      

Debug personalization

Ranking diagnostics is an important tool for debugging product ranking. It can be used to identify the overall impact of enabled personalization rules on different queries or categories.

1. Go to Search & Merchandising > Ranking diagnostics in your Bloomreach dashboard.

   

2. Click the pencil icon to set up an Additional scope definition.

   

3. Enter the User ID or cookie2 value of a site user under the User ID field. Click Update to apply the configuration.

   

4. Enter a search query (or category) for which personalization has been enabled in the search field and hit Enter.

5. This will show the search results for the specified user. The Personalization indicator (On or Off) and badge on top of the products can be used to validate 1:1 personalized ranking.

   

A/B test Personalization rules

A/B tests can be created at any of the following site levels:

• Global - for all searches and categories
• Site search - for select queries or groups of queries
• Category - for select categories or group of categories

📘

Note

For optimizing the benefit of Personalization, we recommend you to A/B test at the Global level (for all queries and categories).

To create an A/B test for a Personalization rule:

1. Save a new test variant of your rule with/without Personalization enabled - While editing or creating a Personalization rule, click on the arrow next to Save, and select Save new test variant from the dropdown.

   

2. Review the rule variants -

   

   The Variants field in the list of Personalization rules shows the number of variants available for a rule. Click on it to view all the Rule variants.

   

   In the Rule variants, “The default behavior” (or v0) for a rule depends on how Personalization is configured globally. For example, if Personalization is enabled at the global level (for all site searches / all categories / both), then the v0 for any local rule of that type (Site search or Category) will also have Personalization ON. The Control label next to a variant depicts your custom Personalization rule (the control group).

3. Select variants and set up test - From the list of Rule variants, include the variants that you want to include in your A/B test and click on Set up new test. There can be two scenarios for selecting variants:

   1. Test a custom variant with the default v0: In this case, be mindful that the v0 is affected by the global Personalization configuration. Updating the global rule will change the behavior of the v0.
   2. Test two custom variants together: You can create two different variants, v1, and v2, with Personalization ON and OFF, respectively. Use these variants and split the traffic between them. Note that since this case doesn't include v0, the test will not be affected by changes in the global configuration.

4. The window to run an A/B test will open, where you can set up your test buckets and allocate traffic. Refer to the How do I run an A/B test? guide for a step-by-step overview of setting up and running the A/B test.

The stats and details for the A/B test are then available in the Testing tab in your Bloomreach dashboard.

📘

Note

To avoid accidentally enabling Personalization on a query/category while testing it, you can toggle Personalization OFF in the rule, and create a new test variant with Personalization ON to run the A/B tests with.

How A/B testing Personalization helps

A/B testing lets you monitor and evaluate the effectiveness of your 1:1 Personalization rules. You can deliver personalized search results to a portion of your traffic, compare them with the default behavior, and analyze their performance before enabling them widely.

The A/B test results can help you evaluate:

• The Personalization configurations and rules that lift key performance metrics like Revenue Per Visitor or Repeat visitors for you.
• The ideal duration for the Personalization to show effective lift.
• The common/strong user affinities on your site, based on user preferences.
• The performance of 1:1 Personalization together with different merchandising rules.

📘

Note

We recommend you to run A/B tests for at least a duration of 30 days to see significant results.

How personalized ranking works

User profile creation

1:1 Personalization uses user data from Bloomreach Pixels to create an anonymous user profile of the site visitor. User journeys are created by monitoring the user’s interactions with different products and pages.

• If it is a returning visitor and has an existing user_id, their long-term profile will be remembered for up to three months.
• If it is a new visitor, the br_uid_2 (cookie2) parameter from the Pixel is used to create their short-term profile for up to one month.

📘

Note

More information on the data collected by the user cookie and its sensitivity can be found in Pixel data protection.

Affinity-based ranking

For each user profile, 1:1 Personalization operates by monitoring the user’s interactions and remembering affinities for different product attributes like brand, color, size, etc. It automatically learns which attributes are relevant for creating and matching affinities.

🚧

Important

Numeric attributes are not used for personalization right now.

The feature uses a multi-attribute personalization approach. Here are the key objectives it achieves:

• It can track the user’s preferences on different pages based on context.
  For example, a ‘Dresses’ page might identify a preference for color or size, while a ‘Perfumes’ page might focus on brand preferences.
• It can transfer relevant affinities from one query or category to another related one.
  For example, if a shopper shows a preference for “size: S” on a ‘Dresses’ page, this preference can also influence related categories like ‘Gowns’ but not unrelated categories like ‘Sandals’.

Personalized ranking precedence rules

1. Bloomreach Merchandising rules will always take precedence over personalized ranking.
2. The latest query intent always takes precedence (for example, a search query “Red shirt” will always override any previous shirt color preferences).
3. In-session affinities take precedence over long-term user profile data.

Understanding session-based personalization

Personalization works over multiple user sessions for both new users (non logged-in) and repeat users (logged-in).

Personalization for non logged-in users

For a new site user with no past interaction data, products are initially ranked using the Bloomreach default algorithm. After the user makes at least 3 interactions on the site, affinities start building up, and personalization starts applying to their subsequent relevant searches.

At this point, the user profile for this visitor would have been generated, and they would get personalized results on later visits for up to a month (unless they log in).

Personalization for logged-in users

For a returning site user with at least 6 past interactions, personalization is applied from the start of a new session (based on their previous affinities). After the user makes at least 3 new interactions on the site, affinities start updating, and real-time personalization starts applying to subsequent relevant searches.

All the interaction data collected over the session is added to the user’s past interaction data, which is stored for up to three months.

Personalization across devices

• 1:1 Personalization will continue to apply to logged-in users with valid user_id, and personalization data will be collected even if they log in from a different device.
• 1:1 Personalization will not apply over multiple devices for non logged-in users (without a user_id).