Configure contextual personalization

This guide walks you through configuring contextual personalization. You'll add variants, define a business goal, and set up contextual features.

Before you start, make sure you understand how variants, value, and goal work together.

Prerequisites

• A Bloomreach account with the contextual personalization feature enabled.
• At least two variants of a weblayer, scenario node, or email campaign ready to test.

Configure contextual personalization

Step 1: Add variants and assign value

Add at least two variants. By default, all variants have a value of 1 — meaning Loomi AI treats them equally. Enable the Value toggle to assign different values and tell Loomi AI which variants matter most to your business.

Example: You're running a subscription banner with two variants — one offers a 10% discount, and one doesn't. Subscribers without discounts are more valuable long-term. Set Variant A (no discount) to 2 and Variant B (discount) to 1. Loomi AI learns to favor Variant A — but only for customers where it's likely to convert.

When assigning value, keep in mind:

• Loomi AI uses the ratio between values, not the absolute numbers. Setting 2 and 1 has the same effect as 20 and 10.
• If a customer triggers multiple goal events within the attribution window — for example, clicking a link and then completing a purchase — the values are summed together.
• When evaluating results, factor in value ratios. A variant with a lower conversion rate but a higher value can outperform a variant that converts more often but is worth less to your business.

Step 2: Enable comparative A/B test

Enable the Comparative A/B test toggle to run a control group alongside contextual personalization. Set traffic distribution to 80% contextual personalization and 20% comparative A/B test. This gives Loomi AI enough data to learn quickly while keeping a meaningful control group for evaluation.

Step 3: Set a goal

Your goal tells Loomi AI which customer action signals that a variant worked. When a customer completes the goal, Loomi AI registers it as a success for the variant that was served.

Goal tracking works differently depending on the channel:

• Email and scenario channels: Loomi AI automatically tracks opens and clicks. You don't need to configure anything to get started. Optionally, define an additional goal — for example, a purchase — to give Loomi AI a stronger business signal.
• Weblayers: There's no automatic tracking, so you must define a goal. A banner click is a good starting point, as it's easy to track and signals customer interest.

To add a goal, click Add goal and define:

• Event type: The customer action to track, for example purchase_item, add_to_cart, or level_up.
• Filter (optional): Conditions to match specific event variations, for example only purchases where category = 'electronics'.

Examples by industry:

• Ecommerce and retail: add_to_cart, use_a_discount_code, or purchase_item.
• Gaming: level_up, make_an_in_app_purchase, or invite_a_friend.
• Banking and finance: starting_a_loan_application or opening_a_new_account.

👍

Tip

If your goal is a purchase but you don't have enough purchase events yet, use an intermediate goal like add_to_cart or a product page view to help Loomi AI learn faster.

For more goal setup examples, see Contextual personalization use cases

Step 4: Configure contextual features

Contextual features are the customer data points Loomi AI uses to decide which variant to serve. During setup, choose one of two options:

1. Optimized by Loomi AI (recommended)

Loomi AI automatically selects the most relevant data for your project. This option is pre-selected by default. All six feature categories are active, but you can uncheck any you don't want to use — at least one must remain active:

• Browsing behavior: Site activity, product views, and areas of interest.
• Campaign response: How customers respond across channels.
• Purchase history: Transaction frequency and buying patterns.
• Time and preferences: When customers prefer to engage.
• Session activity: Device, traffic source, and current actions.
• Customer profile: Demographics such as location and customer type.

2. Custom

Choose specific data points manually — such as your own properties, aggregates, expressions, segmentations, or predictions. Keep these guidelines in mind:

• Keep features dense. Group customers into a small number of meaningful categories. For example, instead of exact hours of the day, use three groups: morning, afternoon, and evening. Avoid unique values like product titles — too many distinct values make it harder for the model to spot patterns.
• Use fewer, more relevant features. More features mean more contexts, which means the model takes longer to learn. Use your business knowledge to select features that genuinely influence customer behavior.
• Size your audience accordingly. Use this formula to estimate the audience size needed for statistically significant results: 100 × number of contexts × number of variants ÷ estimated conversion rate = required audience size. For example: 100 × 20 contexts × 2 variants ÷ 0.2 conversion rate = 20,000 customers.

Next steps

Continue with the setup guide for your channel:

• Contextual personalization for weblayers
• Contextual personalization for email campaigns
• Contextual personalization for scenarios