Skip to main content

📊 Segments & Scoring

Every person in your Audience Graph receives a dynamic engagement score and is automatically placed into segments. These scores and segments drive personalization across the entire platform — from AI Auto-Reply context to Escalation Rules to Client Reports.


Engagement Scoring​

The engagementScore on each AudienceNode is a composite metric calculated from:

SignalWeightDescription
Comment frequencyHighTotal comments across all platforms (totalComments)
DM interactionsHighTotal direct messages (totalDMs)
Purchase historyVery HighNumber of purchases (totalPurchases) and total spent (totalSpent)
Average order valueMediumavgOrderValue — higher AOV increases score
RecencyHighTime since lastInteractionAt — recent activity boosts score significantly
Purchase recencyHighTime since lastPurchaseAt
Interaction diversityMediumEngaging across multiple platforms and interaction types

Scores are updated automatically every time a new interaction is recorded. The algorithm uses a time-decay function so that recent activity is weighted more heavily than historical data.

Behind the Scenes

Each AudienceInteraction logged — whether it's a comment, DM, mention, or purchase — triggers a recalculation of the parent AudienceNode's engagement score. The sentiment and intent fields on interactions also factor in: positive sentiment interactions contribute more than neutral ones.


Auto-Segmentation​

UniPulse AI automatically groups your audience into segments based on engagement score patterns and purchase behavior:

SegmentCriteriaDescription
ChampionsHigh engagement + frequent purchasersYour best customers — engage often and buy regularly
Loyal FansConsistent engagement over timeRegular interactors who may or may not purchase
At RiskDeclining engagement (score dropping)Previously active customers showing disengagement signals
New ContactsfirstSeenAt within last 30 daysRecent first-time interactors
Potential BuyersHigh purchaseIntentScore in conversationsShowing purchase intent signals but haven't converted yet
DormantNo interaction for extended periodNo activity in the lastInteractionAt window

Segments are re-evaluated continuously. A customer can move between segments as their behavior changes.


AudienceSegment Model​

Each segment (both auto and custom) is stored as an AudienceSegment:

FieldDescription
nameSegment name (e.g., "Champions", "Holiday Shoppers")
descriptionHuman-readable description of the segment's purpose
rulesJSON array of rules that define membership (see below)
memberCountCurrent number of audience nodes in this segment
isAutomatictrue for AI-managed segments, false for custom segments

Custom Segments​

Create your own segments in Audience > Segments > New Segment based on any combination of criteria:

CriterionExamples
Engagement scoreScore > 80, Score between 40–70
PlatformOnly Instagram followers, Only TikTok commenters
Interaction typeHas sent a DM, Has commented more than 5 times
Purchase dataTotal spent > $100, Has purchased in last 30 days
TagsHas tag "VIP", Does not have tag "Unsubscribed"
TierCustomer tier is "Premium"
LocationBased on available profile data
SentimentAverage sentiment across interactions is positive
Time-basedFirst seen before a specific date, Last interaction within X days

Rules are defined as JSON and support AND/OR logic for complex combinations.


Using Segments Across UniPulse​

Segments are a powerful building block used throughout the platform:

FeatureHow Segments Are Used
AI Auto-ReplyPersonalize bot responses based on customer segment
Escalation RulesTrigger escalation when a VIP or Champion customer contacts you
Client ReportsGenerate segment-specific performance reports
Content targetingPrioritize content for specific audience segments
Coupon generationAuto-generate personalized coupons for at-risk or high-value segments