Chapter 9: Refining Rules

The Problem

Your first segment worked, but real targeting is rarely that simple. You need: "High-value customers who purchased electronics OR home goods, but exclude anyone who returned an item in the last 60 days, and also exclude people already in the VIP program."

Multiple conditions. OR logic. Exclusions. Segment references. This chapter shows you how.

The Key Idea

Core concept

Complex segments combine simple rules with AND/OR logic and include/exclude references to other segments.

AI handles the syntax. You describe the logic in plain English, then verify the YAML captures your intent.

AND vs. OR Logic

By default, conditions combine with AND using type: And:

rule:
  type: And
  conditions:
    - type: Value
      attribute: lifetime_value
      operator:
        type: Greater
        value: 1000
    - type: Value
      attribute: country
      operator:
        type: Equal
        value: "USA"

This means: lifetime_value > 1000 AND country = USA.

For OR logic, use type: Or:

> "Target customers from USA OR Canada"

rule:
  type: Or
  conditions:
    - type: Value
      attribute: country
      operator:
        type: Equal
        value: "USA"
    - type: Value
      attribute: country
      operator:
        type: Equal
        value: "Canada"

Complex Combinations

Real targeting often mixes AND and OR. Nest rule groups to combine them:

> "High-value customers from USA or Canada who purchased electronics or home goods"

rule:
  type: And
  conditions:
    - type: Value
      attribute: lifetime_value
      operator:
        type: Greater
        value: 1000
    - type: Or
      conditions:
        - type: Value
          attribute: country
          operator:
            type: Equal
            value: "USA"
        - type: Value
          attribute: country
          operator:
            type: Equal
            value: "Canada"
    - type: Or
      conditions:
        - type: Value
          attribute: last_purchase_category
          operator:
            type: Equal
            value: "Electronics"
        - type: Value
          attribute: last_purchase_category
          operator:
            type: Equal
            value: "Home & Garden"

Reading this: LTV > 1000 AND (USA OR Canada) AND (Electronics OR Home & Garden).

Using the In Operator

For multiple values in the same field, In is cleaner than multiple ORs:

> "Target customers in USA, Canada, or UK"

rule:
  type: Value
  attribute: country
  operator:
    type: In
    value: ["USA", "Canada", "UK"]

This is equivalent to: country = USA OR country = Canada OR country = UK.

Excluding Audiences

To exclude customers, use type: exclude with a segment reference:

> "Exclude customers in the Returns segment"

rule:
  type: And
  conditions:
    - type: Value
      attribute: lifetime_value
      operator:
        type: Greater
        value: 1000
    - type: exclude
      segment: recent-returners

The type: exclude removes anyone in the referenced segment from your audience.

Referencing Other Segments

Instead of duplicating rules, reference existing segments:

> "Target high-value customers but exclude the VIP segment"

rule:
  type: And
  conditions:
    - type: Value
      attribute: lifetime_value
      operator:
        type: Greater
        value: 1000
    - type: exclude
      segment: vip-customers

This includes customers matching your rules, minus anyone already in "vip-customers."

You can also include other segments:

rule:
  type: And
  conditions:
    - type: include
      segment: engaged-email-subscribers
    - type: Value
      attribute: lifetime_value
      operator:
        type: Greater
        value: 500

This targets people in "engaged-email-subscribers" AND with LTV > 500.

Available Operators

Ask AI what operators work for different field types:

Comparison (numbers, dates):

• Equal, NotEqual
• Greater, GreaterEqual
• Less, LessEqual
• Between

Lists:

• In, NotIn

Strings:

• Contain, StartWith, EndWith
• Regexp (pattern matching)

Time/Behaviors:

• TimeWithinPast, TimeWithinNext
• TimeToday, TimeThis, TimeNext
• TimeRange, TimeInBetweenNext

Null checks:

• IsNull (field has no value)

Mental Model: Venn Diagrams

Think of each rule as a circle of customers:

Your segment is the intersection of the circles you want, minus the circles you exclude.

Building Complex Segments Step by Step

For complicated logic, build incrementally:

> "Start with high-value customers"

Review the audience size.

> "Add: must have purchased in last 90 days"

Review again.

> "Add: exclude the Churned segment"

Final review.

This iterative approach helps you understand how each rule affects the audience.

Verifying Complex Logic

After AI generates the YAML, verify by asking:

> "Explain this segment's targeting logic in plain English"

AI reads the YAML and explains:

This segment includes customers where:
1. Lifetime value is greater than $1000
2. AND country is either USA or Canada
3. AND they have NOT made a return in the last 60 days
4. AND they are NOT in the VIP Customers segment

Estimated audience: 2,847 customers

If the explanation doesn't match your intent, refine the rules.

Pitfalls

"The OR logic isn't working as expected."

Check the nesting. OR groups need to be explicitly defined:

> "Show me the rule structure as a logic tree"

"The exclusion excluded too many people."

Exclusions might be broader than intended. Preview the segment:

> "How many customers are being excluded by the return rule?"

"I referenced a segment that doesn't exist."

Validation catches this:

✗ Segment 'VIP Customers' not found
  Available segments: VIP Members, VIP Program, ...

What You've Learned

• Conditions combine with type: And or type: Or
• Nest And/Or groups for complex logic
• In operator handles multiple values cleanly
• type: exclude removes segment members
• type: include adds segment members
• Build complex logic incrementally
• Verify by asking AI to explain the rules

Next Step

You can target precisely. Chapter 10 shows you how to send these audiences to marketing channels—Salesforce, Google Ads, email platforms, and more.

---

You've mastered targeting logic. Next, you'll connect it to the outside world.