Personalization Decisions

A decision defines what personalized content a specific audience receives when a personalization point is requested. Each personalization point can have up to 25 decisions, evaluated by priority to determine the winning response for each individual.

Anatomy of a Decision

Every decision has four key components:

1. Priority

Each decision is assigned a priority value that determines evaluation order. Key facts:

• New decisions are automatically assigned the lowest priority (evaluated last)
• Priority can be adjusted from the related object list view on the personalization point
• When an individual qualifies for multiple decisions, the highest priority qualifying decision wins
• Personalization returns only one decision response per personalization point — it does not return responses for multiple decisions

2. Targeting Rules

Targeting rules are optional and determine who is eligible to receive a specific decision. If no rules are defined, any individual is eligible.

Rules can be written against:

3. Personalization Attributes

These are the string entry text fields defined by the response template. For Manual Content decisions, this is where business users type in the content values (header text, CTA text, image URLs, etc.). The values entered here are returned in the decision response for qualifying individuals.

4. Recommender

For decisions on personalization points of type Recommendations, you select a recommender that determines which items are returned. Available recommenders must:

• Be configured with the same profile DG as the personalization point
• Return items from the same DMO that the response template is configured against

Creating a Decision — Step by Step

1. Navigate to the Personalization app → Personalization Points
2. Open the personalization point where you want to add a decision
3. In the Decisions section, click New Decision
4. Optionally configure targeting rules to restrict who qualifies
5. Fill in the personalization attributes (for Manual Content) or select a recommender (for Recommendations)
6. Save the decision
7. Adjust priority if needed from the personalization point's decision list view

Multiple Decisions — Priority Evaluation

When a personalization point has multiple decisions, the evaluation follows a strict priority order:

CopyIndividual requests personalization point "home_hero"
│
├── Decision 1 (Priority: Highest) — Rules: "VIP Segment"
│   └── Does the individual qualify? → YES → Return this decision ✓
│                                    → NO  → Continue ↓
│
├── Decision 2 (Priority: Medium) — Rules: "New Customer Segment"
│   └── Does the individual qualify? → YES → Return this decision ✓
│                                    → NO  → Continue ↓
│
└── Decision 3 (Priority: Lowest) — Rules: None (default/fallback)
    └── All individuals qualify → Return this decision ✓

Run-Time Decision Flow

When a personalization request reaches the decisioning pipeline, it passes through five stages:

1. Personalization Request

The request contains:

• An individual ID (anonymous or known)
• One or more personalization point IDs
• Optional context data (anchor item ID, page URL, category, etc.)

The Web SDK automatically sends context data required for any contextual rules configured on decisions or experiments.

2. Augmenting Phase

Personalization uses the individual ID to request the corresponding profile data graph from the Data Cloud profile API. If no profile exists, the individual is treated as a first-time anonymous visitor.

3. Qualifying Phase

For each personalization point in the request, the pipeline evaluates:

• Experiments first — Experiments are always the highest-priority item on a point
• Then decisions by priority — The highest-priority qualifying decision wins

4. Personalizing Phase

Once the qualifying decision (or experiment cohort) is determined:

• Manual Content — The text values from the decision are assembled into the response
• Recommendations — The profile DG is passed to the recommendations service alongside the recommender ID. The service generates personalized 1:1 recommendations based on the real-time profile.

5. Decision Response & Logging

The decision response is returned to the requesting application. Simultaneously, the decision is logged in Data Cloud in the personalization log. This logging is critical for analytics and attribution tracking.

Targeting Rule Examples

Segment-Based Targeting

Show a premium product banner only to members of the "High Value Customers" segment.

Attribute-Based Targeting

Show a localized promotion to users in a specific region.

CI-Based Targeting

Show a re-engagement offer to users who haven't purchased recently.

Contextual Targeting

Show a category-specific promotion only on category pages.

Best Practices

1. Start with 2–3 decisions per point — A targeted decision, a secondary audience, and a fallback
2. Use contextual rules sparingly — They add runtime evaluation overhead
3. Test before going live — Use the WPM preview to validate targeting against specific individuals
4. Monitor blank responses — If a point frequently returns empty responses, add a fallback decision with no targeting rules
5. Coordinate with recommender filters — Decision targeting rules determine who qualifies; recommender filters determine which items are returned. Make sure both are aligned.

Next: Experiments — Learn how to A/B test different decisioning strategies on a personalization point.