Auto-Segmentation
Auto-Segmentation uses BQML K-Means clustering to discover natural customer groups in your data. Instead of defining segment rules by hand, you let the algorithm find patterns, review the results, and convert any interesting cluster into a rule-based segment that can be synced to ad platforms and messaging tools.
You can find Auto-Segmentation in Customer 360 (under Data → Customer 360), on the Discover tab. The audiences you create from discovered clusters live under Data → Audiences.
Overview
The workflow has four steps:
- Choose a goal or select features manually.
- Configure the number of clusters.
- Review the discovered clusters.
- Convert any cluster into a segment for activation.
Behind the scenes, Vendo creates a BQML K-Means model, trains it in your BigQuery warehouse, and returns the cluster results. No data leaves your project.
Getting Started
- Navigate to Data → Customer 360 and open the Discover tab.
- Click Discover Segments.
- Select a goal preset or choose Custom to pick features yourself.
Goal Presets
Presets pre-select a curated set of calculated properties from Customer 360:
| Preset | Features Selected | Best For |
|---|---|---|
| Find High-Value Customers | Revenue, order count, AOV, purchase frequency, recency | E-commerce teams looking to identify top spenders and loyal buyers |
| Identify At-Risk Customers | Days since last order, engagement score, session frequency, lifecycle stage | Retention teams looking to catch churning users early |
| Custom | You choose from all available calculated properties | Any use case not covered by the presets |
Choosing Features
Whether you use a preset or go custom, the property picker shows all available calculated properties (21 built-in behavioral properties covering revenue, engagement, lifecycle, and more).
Tips for feature selection:
- More features is not always better. Start with 4—8 features that are relevant to the question you are trying to answer.
- Avoid redundant features. If two features measure the same thing (e.g.,
total_revenueandtotal_spent), pick one. - Mix categories. Combining revenue features with engagement features often produces more useful clusters than using one category alone.
Number of Clusters
Choose how many groups the algorithm should create, from 2 to 10. If you are unsure, start with 3—5. You can always re-run with a different number.
Understanding Results
After training completes, the results page shows each discovered cluster with:
| Detail | Description |
|---|---|
| Cluster name | Auto-generated label (Cluster 1, Cluster 2, etc.) |
| Cluster size | Number of customers in the cluster |
| Top defining features | The features where this cluster differs most from the overall population |
| Radar chart | Visual comparison of normalized feature values across all clusters |
Reading the Radar Chart
Each axis of the radar chart represents one feature, normalized to a 0—1 scale. Clusters that extend further on a given axis have higher values for that feature. This makes it easy to spot, for example, a “high revenue, low engagement” cluster versus a “low revenue, high engagement” cluster.
Interpreting Clusters
Look for clusters with clear marketing characteristics:
- A cluster with high order count and high recency is your loyal, active segment.
- A cluster with high historical revenue but low recency may be at-risk churners.
- A cluster with low revenue across the board may be window shoppers or one-time buyers.
Converting to Segments
Once you identify a cluster worth targeting:
- Click Convert to Segment on the cluster card.
- Vendo creates a rule-based segment with filters that approximate the cluster boundaries.
- The new segment appears as an audience under Data → Audiences and can be edited like any other segment.
- From the segment detail page, click Sync to Destination to push the user list to:
- Google Ads (Customer Match)
- Meta Ads (Custom Audiences)
- TikTok Ads (Custom Audiences)
- Snap Ads (Customer List)
- Microsoft Ads (Customer Match)
- Klaviyo (Lists)
- OneSignal (Segments)
- Customer.io (Segments)
This one-click conversion bridges the gap between unsupervised discovery and marketing activation: the algorithm finds the groups, and you decide which ones to activate.
Related
- Segments — The rule-based segments that auto-segmentation creates
- BQML Models — The underlying K-Means model powering auto-segmentation
- Customer 360 — calculated properties providing input features