Skip to main content

Customer Quality Scoring

Customer Quality Scoring assigns every tracked visitor a score from 0 to 100 and places them into actionable segments. These segments are sent alongside conversion events to ad platforms, enabling smarter bidding, better lookalike audiences, and reduced wasted spend.

How the Algorithm Works​

The scoring algorithm evaluates five behavioral signals, each with a configurable weight:

SignalDefault WeightHow It's Measured
Purchase count30%Total number of completed purchase events
Total revenue25%Sum of all purchase event values in the customer's currency
Visit frequency20%Distinct sessions (unique days with at least one event) within the scoring window
Recency15%Inverse of days since last activity -- more recent activity scores higher
Cart-to-purchase rate10%Ratio of purchase events to add_to_cart events

Each signal is normalized to a 0--100 scale relative to all customers in the workspace, then combined using the weights above to produce the final score.

Scoring Window

By default, the algorithm considers the last 90 days of activity. Older events are excluded to keep scores reflective of current behavior.

Score Recalculation​

Scores are recalculated automatically:

  • On event ingestion -- When a new event arrives for a customer, their score is queued for recalculation.
  • Scheduled batch job -- Every 6 hours, a background job recalculates scores for all active customers to account for recency decay.

Segment Definitions​

Based on their score and behavioral patterns, customers are assigned to one of six segments:

SegmentScore RangeAdditional CriteriaDescription
high_value> 80--Top-tier customers with strong purchase history and engagement
repeat60--802+ purchasesReturning buyers with consistent activity
promising40--60Active in last 30 daysEngaged visitors who have not yet become repeat buyers
at_risk20--40Declining activity trendPreviously active customers showing reduced engagement
newAnyFirst seen < 30 days agoRecently acquired visitors, regardless of score
churned< 20No activity in 90+ daysDormant customers with no recent signals
Segment Priority

A customer can only belong to one segment. The new segment takes priority -- any customer first seen within the last 30 days is classified as new regardless of score. After 30 days, the score-based rules apply.

Using Segments for Ad Optimization​

Meta (Facebook/Instagram)​

  • Lookalike Audiences -- Export your high_value segment and create a 1% lookalike audience in Meta Ads Manager. This tells Meta to find new users who resemble your best customers.
  • Custom Audiences -- Use at_risk or churned segments for win-back campaigns with tailored messaging.
  • Value Optimization -- When the purchase event includes the customer score as a custom property, Meta can optimize delivery toward higher-value conversions.
  • Value-Based Bidding -- The score is sent as a conversion value adjustment. Google's Smart Bidding uses this to prioritize clicks from users likely to become high-value customers.
  • Customer Match -- Upload hashed email lists from specific segments for targeted remarketing.

TikTok​

  • Custom Audiences -- Build segment-based audiences for retargeting.
  • Value Optimization -- Pass the score with CompletePayment events to enable TikTok's value-based optimization.

Viewing Customer Scores​

Individual Customer Profile​

Navigate to Analytics > Tracking > Customers and click any customer to view:

  • Current score and segment assignment
  • Score history chart showing changes over time
  • Event timeline listing all tracked events
  • Attribution data including first-touch and last-touch sources

Segment Overview​

The Segments tab shows aggregate statistics:

MetricDescription
Customer countNumber of customers in each segment
Average scoreMean score within the segment
Average revenueMean lifetime revenue per customer
TrendWeek-over-week change in segment size

GDPR Compliance​

Customer scoring respects consent and privacy regulations:

  • Consent required -- Scores are only calculated for customers who have granted tracking consent via omni('consent', true).
  • Data export -- Customer score data is included in GDPR data export requests.
  • Data erasure -- When a customer requests deletion, all scoring data, event history, and segment assignments are permanently removed.
  • No PII in scores -- The scoring algorithm uses behavioral signals only. Personal identifiers (email, phone) are hashed before storage and never used in score computation.
warning

If a customer revokes consent, their score is frozen and no new events are processed. The score is retained until an explicit erasure request is made.