Advanced Segmentation in Customer.io: The Complete Guide
Advanced Segmentation in Customer.io: The Complete Actionable Guide
In August 1854, a cholera outbreak tore through the Soho district of London, killing more than 500 people in ten days. The city's authorities blamed "bad air"—a one-size-fits-all theory that sent them down the wrong path entirely.
Dr. John Snow did something different. He plotted every death on a map, then cross-referenced who died with what water source they used. Workers at a nearby brewery showed zero deaths—they drank beer, not water. A widow living far from Soho but who had water delivered from the Broad Street pump? She died. Snow segmented by behaviour, not just location.
His conclusion: cholera spreads through contaminated water. He had the pump handle removed. The outbreak collapsed.
One insight. The right data. The right grouping. Hundreds of lives saved.
That's segmentation in its purest form. And it's exactly what you're leaving on the table when you treat all your customers the same.
This guide covers everything you need to build advanced segments in Customer.io: behavioural segments, RFM analysis, AI-powered predictive segments, and custom objects for relationship-based grouping. You'll also get ready-to-use templates for the four segments that matter most: active users, at-risk customers, power users, and dormant accounts.
Why Does Segmentation Matter So Much?
Segmented email campaigns generate a 760% increase in revenue compared to batch-and-blast sends. That number sounds wild, but it makes sense when you think about it: you're sending the right message to the right person at the right moment.
Behaviourally segmented campaigns achieve 94% higher click-through rates than non-segmented campaigns. Personalised emails get opened 82% more often and achieve six times higher transaction rates.
The cost of getting it wrong is equally stark. According to Recurly's 2023 churn benchmark study across 1,200+ subscription businesses, a 5% monthly churn rate means you lose nearly 46% of your customers every year. Most of that churn is preventable—but only if you can spot who's at risk before they leave.
Segmentation is how you spot them.
What Are the Four Types of Segmentation in Customer.io?
Customer.io supports four core segmentation approaches, and the magic happens when you combine them.
1. Manual segments — you define static lists or fixed rule-sets. Simple, but they don't update dynamically.
2. Behavioural segments — built from event data. Who logged in, what they clicked, which features they used, when they last took an action.
3. Data/attribute segments — based on profile fields: plan type, country, signup date, company size.
4. AI-powered segments — Customer.io's AI Segment Builder lets you describe your audience in plain language and generates the conditions automatically, using your workspace's real data structure (never your customers' actual PII).
The best segments combine all four. An at-risk segment, for example, might combine attribute data (plan type = paid), behavioural data (no login event in 21 days), and predictive signals (declining feature usage trend).
How Do You Build Behavioural Segments in Customer.io?
Behavioural segments are built from the events you track. Every time a user does something—logs in, completes a task, clicks a feature, views a pricing page—Customer.io can record that as an event and use it in segment conditions.
Setting Up Event-Based Conditions
In Customer.io's segment builder, you can filter by:
- Event performed — has or has not performed an event
- Event frequency — performed an event more than / less than / exactly N times
- Event recency — performed an event in the last N days/weeks
- Event properties — performed an event where a property equals a specific value
You can nest these conditions with AND/OR logic to build precise audience definitions.
Behavioural Segment Example: Active Users (Last 14 Days)
Here's a straightforward active-user segment:
Conditions:
AND
├── Has performed event: "session_start" at least 1 time in the last 14 days
├── Has performed event: "feature_used" at least 3 times in the last 14 days
└── Attribute: status = "active"
This catches users who have logged in and actually used a feature—not just people who opened the app. That distinction matters when you're building engagement scoring.
Combining Behavioural and Profile Data
Behavioural segments get powerful when you layer in profile attributes. A "churning paid user" segment looks very different from a "churning free user"—and should trigger very different journeys.
For more on how these segments feed into automated workflows, check out The Ultimate Guide to Customer.io Journeys.
What Is RFM Analysis and How Do You Apply It in Customer.io?
RFM stands for Recency, Frequency, Monetary Value. It's one of the most battle-tested customer segmentation frameworks in marketing, used in retail and subscription businesses for decades.
- Recency — How recently did this customer take a meaningful action (purchase, login, engagement)?
- Frequency — How often do they take that action?
- Monetary — How much revenue have they generated?
Each dimension gets scored on a scale (typically 1–5), and you combine the scores to categorise customers into meaningful groups.
RFM Score Mapping
| RFM Score | Segment Label | Description |
|---|---|---|
| 555 | Champions | Bought recently, buy often, spend the most |
| 554–544 | Loyal Customers | Regular buyers, high value |
| 512–511 | Recent Customers | Bought recently, but not yet frequent |
| 155–144 | At-Risk Customers | Used to be Champions, haven't returned |
| 111–211 | Lost/Dormant | Low recency, frequency, and value |
| 455–444 | Potential Loyalists | Recent buyers with average frequency |
Building RFM Segments in Customer.io
Customer.io doesn't have a native "RFM score" field, but you can implement RFM in two ways:
Option 1: Compute and sync scores via your data warehouse
Run RFM calculations in BigQuery, Snowflake, or Redshift. Push the resulting rfm_score or rfm_segment attribute back to Customer.io via API or reverse ETL. Then segment on rfm_segment = "at_risk". This approach is covered in detail in our guide on integrating data sources with Customer.io.
Option 2: Approximate RFM using native event/attribute conditions Build proxy segments directly in Customer.io:
"Champions" Proxy Segment:
AND
├── Has performed event: "purchase" in the last 30 days
├── Has performed event: "purchase" 5+ times (lifetime)
└── Attribute: lifetime_value >= 500
"At-Risk" Proxy Segment:
AND
├── Attribute: lifetime_value >= 200
├── Has NOT performed event: "purchase" in the last 90 days
└── Has performed event: "purchase" at least 2 times (lifetime)
Why RFM Still Works (And Where It Falls Short)
RFM is fast to implement and highly actionable. Businesses using data-driven retention techniques report an average 16x ROI on churn management.
But RFM has limits. It's backward-looking. It tells you what someone did, not what they're about to do. That's where predictive segmentation comes in.
What Are Predictive Segments in Customer.io?
Predictive segments use AI to identify patterns in your customer data that humans would never spot manually. Instead of asking "who has churned?", they ask "who will churn in the next 30 days?"
According to Customer.io's 2025 State of Lifecycle Marketing survey, 85% of lifecycle marketing teams increased their AI usage in 2025, with 45% calling the increase "huge." The top use case? Smarter segmentation.
Customer.io's AI Segment Builder
Customer.io launched its AI Segment Builder as part of its broader AI feature suite. Here's how it works:
-
Describe your audience in plain language. You type something like: "Users on a paid plan who haven't used the core feature in 14 days and haven't opened an email in a month."
-
AI generates the conditions. The AI translates your description into precise segment rules using your workspace's actual attribute names and event schema.
-
Privacy-first by design. The AI only processes field names, never actual customer values. Zero PII goes through the model.
-
Iterate conversationally. Refine the segment by continuing the conversation with the AI assistant until the conditions match your intent.
-
Preview before saving. See live membership counts and engagement metrics before committing the segment.
Real Example: AI-Generated Churn Risk Segment
Here's what the AI segment builder might generate for "users showing early churn signals":
Conditions:
AND
├── Has NOT performed event: "core_feature_used" in the last 21 days
├── Attribute: plan_type = "paid"
├── Attribute: account_age_days >= 30
├── Has NOT opened an email in the last 28 days
└── Has performed event: "core_feature_used" at least 5 times (lifetime)
[^ confirms they *were* an engaged user, not a never-activated account]
This segment catches the dangerous middle: users who were genuinely engaged, then went quiet. That's very different from users who never activated—and the message they need is completely different.
For the broader AI picture in email marketing, see our deep-dive on what's real, what's hype, and what actually works with AI.
Beyond RFM: Multi-Signal Predictive Scoring
The most sophisticated predictive segments don't just look at one or two signals. They combine:
- Product usage patterns (feature adoption depth, session frequency, workflow completion)
- Email engagement (open recency, click patterns, unsubscribe signals)
- Support interactions (ticket volume, sentiment, response time)
- Billing behaviour (failed payments, downgrade events, pause requests)
A 2024 study in Results in Engineering (Elsevier) found that multi-signal ML approaches significantly outperform single-dimension models for predicting customer behaviour. The practical implication: don't just track one thing—connect the dots across your entire customer data ecosystem.
Micro-segmentation based on product usage data yields 25–40% engagement improvements and accelerates lifecycle milestones when applied to triggered messaging.
What Are Custom Objects in Customer.io and How Do They Help Segmentation?
Custom objects let you model relationships and data that don't fit neatly onto a user profile. Think of them as structured records attached to users—things like orders, accounts, appointments, job postings, or subscription plans.
Customer.io's Custom Objects feature defines these as "any object a customer creates, buys, schedules, or interacts with."
When to Use Custom Objects
Use custom objects when your segmentation logic depends on data that belongs to a relationship rather than a person. Examples:
- SaaS (B2B): Segment account admins by the number of active seats in their organisation
- EdTech: Segment students by the course they're enrolled in and its completion percentage
- eCommerce: Segment customers by their most recent order's product category
- Healthcare: Segment patients by their next scheduled appointment type
- Marketplace: Segment sellers by the number of active job listings in their account
Creating and Using Custom Objects
You can create custom objects via:
- The Customer.io platform UI
- REST API
- Reverse ETL (syncing from your data warehouse)
- Segment group calls
Each object type has its own attributes, and you can use those attributes in segment conditions and Liquid personalisation in your message templates.
Custom Object Segment Example: B2B SaaS Account Health
Segment: "Account Admin — Low Team Adoption"
AND
├── Object relationship: has_object "account" where:
│ ├── account.seat_count >= 5
│ ├── account.active_users_last_30_days < (account.seat_count * 0.5)
│ └── account.plan_type = "team"
└── Attribute: role = "admin"
This fires a proactive outreach campaign to account admins whose teams are using fewer than half their purchased seats—a classic pre-churn signal in B2B SaaS.
To get the most from custom objects, you need clean, well-structured data pipelines. Our guide on integrating data sources with Customer.io covers how to pipe this data in reliably.
The Four Segments Every Customer.io Account Needs
These four segments cover the full customer lifecycle. Build them first, then layer in more nuanced variations as your data matures.
Segment 1: Active Users
Who they are: Customers who regularly engage with your product or service. They've logged in, used core features, and opened emails recently.
Why they matter: Active users are your baseline. They're also your best candidates for upsell campaigns, referral programmes, and case study outreach.
Template:
Active Users Segment:
AND
├── Has performed event: "session_start" at least 2 times in the last 14 days
├── Has performed event: "core_feature_used" at least 1 time in the last 14 days
├── Has opened an email in the last 30 days
└── Attribute: status = "active"
What to send them:
- Feature announcements and product updates
- Upsell sequences based on their usage patterns
- Referral programme invitations
- Community or loyalty programme invites
Key metric to watch: If this segment shrinks without a corresponding growth in power users, you have an engagement problem brewing.
Segment 2: At-Risk Customers
Who they are: Previously engaged customers whose activity has dropped significantly. They haven't churned yet, but they're on the path.
Why they matter: This is your highest-value intervention opportunity. 54.5% of subscription businesses reduced overall churn in 2023—largely through early intervention campaigns like the ones this segment enables.
Template:
At-Risk Customers Segment:
AND
├── Has NOT performed event: "session_start" in the last 21 days
├── Has performed event: "session_start" at least 5 times (lifetime)
[^ confirms prior engagement — not a never-activated user]
├── Attribute: plan_type = "paid"
├── Has NOT opened an email in the last 21 days
└── Attribute: account_age_days >= 60
What to send them:
- Personal check-in email from a named CSM or founder
- "What's changed?" survey (short, 1–2 questions)
- Win-back offer or loyalty incentive
- Product education on features they haven't tried
Exclude from: Promotional campaigns and upsell sequences. They need care, not a sales pitch.
See our complete SaaS retention playbook for the full at-risk journey architecture.
Segment 3: Power Users
Who they are: Your most engaged customers. They use the product heavily, explore advanced features, and often become advocates.
Why they matter: Power users are disproportionately valuable. The top 20% of customers typically generate 80% of revenue—the classic Pareto distribution that RFM analysis consistently validates. Power users are also your richest source of product feedback and case study material.
Template:
Power Users Segment:
AND
├── Has performed event: "session_start" 15+ times in the last 30 days
├── Has performed event: "advanced_feature_used" at least 3 times in the last 30 days
├── Attribute: account_age_days >= 90
├── Attribute: lifetime_value >= [your 80th percentile LTV]
└── Has opened an email in the last 60 days
What to send them:
- Beta feature invitations
- Personalised case study or testimonial requests
- VIP support or dedicated CSM access
- Advanced tutorials and certifications
- Loyalty rewards and exclusive perks
Key insight: Power users who also open emails regularly are gold for any referral programme. Their combination of deep product knowledge and communication engagement makes them ideal advocates.
Segment 4: Dormant Accounts
Who they are: Users who signed up (and possibly activated) but have had zero meaningful engagement for an extended period.
Why they matter: Dormant accounts aren't just lost revenue—they're a deliverability risk. Sending to large volumes of unengaged addresses damages your sender reputation. You need to either re-engage them or sunset them cleanly.
For everything you need to know about protecting deliverability while managing dormant lists, read our complete email deliverability guide.
Template:
Dormant Accounts Segment:
AND
├── Has NOT performed event: "session_start" in the last 90 days
├── Has NOT opened an email in the last 90 days
└── Attribute: status = "active"
[^ still technically active — not already cancelled]
Optional refinements:
├── Attribute: account_age_days >= 30 [exclude brand-new signups]
└── NOT in segment: "Never Activated" [separate problem, separate flow]
What to send them (a 3-step re-engagement sequence):
- Email 1 (Day 1): "We miss you" — simple, curious, no pressure. Highlight one new feature since they last logged in.
- Email 2 (Day 7): "Here's what you're missing" — social proof, a success story, or a usage insight ("Teams like yours achieved X using [feature].").
- Email 3 (Day 14): Last-chance offer or account status notice. "Your account goes inactive on [date]" performs well here for urgency.
If they don't engage after all three, suppress them from future sends. Clean lists outperform large lists every single time.
How Do Nested Segments Work in Customer.io?
Nested segments let you use an existing saved segment as a condition inside a new segment. This is one of Customer.io's most underused features, and it changes how you manage segmentation at scale.
Why it matters: Instead of rebuilding complex condition sets every time you create a new segment, you reference an existing one. When you update the core segment, every downstream segment updates automatically.
Example:
You build a foundational "Engaged Users" segment:
Engaged Users:
AND
├── Has opened an email in the last 30 days
├── Has performed event: "session_start" in the last 14 days
└── Attribute: status = "active"
Now you build "Upgrade Candidates" by nesting it:
Upgrade Candidates:
AND
├── IN segment: "Engaged Users" ← nested reference
├── Attribute: plan_type = "starter"
├── Has performed event: "premium_feature_viewed" at least 2 times in the last 30 days
└── Attribute: account_age_days >= 45
If you later update your definition of "Engaged Users," every segment that references it inherits the change instantly. This is essential for teams managing large Customer.io accounts.
How Does Segmentation Connect to Behaviour-Triggered Journeys?
Segments are the who. Journeys are the what happens next. The combination is where the real revenue lives.
When a user enters a segment—or leaves one—that transition can trigger a journey automatically. Someone falling into "At-Risk" at 3am should trigger a re-engagement sequence without anyone pressing a button.
We've written an entire guide on moving from time-based drips to behaviour-triggered journeys. The short version: time-based sequences treat all users the same. Segment-triggered journeys treat them as individuals.
This is also how you build a proper lifecycle email marketing strategy — each stage of the lifecycle has its own segment definition, its own journey, and its own success metric.
Tips for Keeping Your Segments Accurate and Performant
Keep event names consistent. If your engineering team logs login in one place and session_start in another, your segments will miss users. Enforce a naming convention and document it in Customer.io's Data Index.
Use the Data Index to document your events. Customer.io's AI segment builder produces better results when events have clear descriptions. Write them. Your future self will thank you.
Set a segment review calendar. Check your key segments quarterly. Business rules change. Products evolve. A segment built six months ago may no longer reflect what you actually want.
Monitor segment sizes. Sudden drops or spikes in a segment's membership are often your first signal that something upstream changed—an event stopped firing, a field name changed, or a data pipeline broke.
Separate activation failures from re-engagement problems. Never-activated users (signed up, never used the product) are a completely different problem from dormant users (used the product, then stopped). Mix them up and you'll send the wrong messages to both groups.
Frequently Asked Questions
What is the difference between a manual segment and a behavioural segment in Customer.io?
A manual segment is a static list you create by uploading contacts or using fixed rules that don't update based on user actions. A behavioural segment is dynamic—it automatically adds and removes users as their behaviour changes. If a user in your "Active Users" segment goes 21 days without logging in, they leave the segment (and potentially enter your "At-Risk" segment) without any manual work. Behavioural segments are almost always preferable for lifecycle automation because they stay current without maintenance.
How many conditions can you add to a Customer.io segment?
Customer.io doesn't publish a hard limit on conditions, but performance and maintainability are practical constraints. For complex logic, use nested segments to keep individual segment definitions readable and modular. A segment with 15 nested conditions is harder to debug and maintain than three simpler segments combined into one.
Can you use custom object data in segment conditions?
Yes. Customer.io lets you build segment conditions based on custom object attributes and relationships. For example: "users who have an 'account' object where account.seat_count is greater than 10 and account.active_users_last_30_days is less than 5." This makes custom objects essential for B2B SaaS segmentation where account-level health drives user-level messaging.
How does Customer.io's AI Segment Builder work?
You describe your target audience in plain English. Customer.io's AI reads your workspace's actual event names and attribute definitions (from the Data Index) and generates the matching segment conditions. Crucially, it only processes field names—never actual customer values—so no PII passes through the AI model. You can then preview the segment membership before saving and refine the conditions conversationally.
What's the best way to implement RFM scoring in Customer.io?
The most robust approach is to calculate RFM scores in your data warehouse (BigQuery, Snowflake, Redshift) and sync the results back to Customer.io via API or reverse ETL as a profile attribute (rfm_segment = "champions"). You can then segment directly on that attribute. If you don't have a data warehouse, you can approximate RFM using native Customer.io conditions based on event frequency, event recency, and lifetime value attributes—as long as you're tracking those signals.
How do you prevent dormant users from hurting email deliverability?
Create a dedicated dormant segment and exclude it from all regular campaigns. Run a structured 3-email re-engagement sequence. Anyone who doesn't engage after the sequence should be suppressed from future sends. Sending to large volumes of unengaged addresses will damage your sender reputation and hurt inbox placement for your entire list—including your active users. Our email deliverability guide covers the full technical framework.
Can segments trigger journeys automatically in Customer.io?
Yes. You can configure Customer.io journeys to trigger when a user enters or exits a segment. This means segment transitions—like moving from "Active" to "At-Risk"—automatically enroll users in the appropriate journey. You don't need to manually review lists or schedule campaigns. The system responds in real time as behaviour changes.
How often do dynamic segments update in Customer.io?
Customer.io segments update continuously as new events fire and attribute values change. There's no manual refresh required. When a user performs an action that meets or breaks a segment condition, their membership updates in real time. This is what makes behaviour-triggered journeys possible—and why segment entry/exit is a reliable trigger mechanism.
What's the difference between a predictive segment and a behavioural segment?
A behavioural segment reacts to what a user has already done. A predictive segment anticipates what a user is likely to do next, based on patterns across your entire customer dataset. For example, a behavioural segment might catch users who haven't logged in for 21 days. A predictive segment might identify users who will likely stop logging in within the next two weeks—before the silence even begins—based on subtle early signals like decreasing feature depth, shorter session times, and declining email engagement. Predictive segments let you intervene earlier, when it's easier to turn behaviour around.
How do you structure segments for a B2B SaaS with multiple users per account?
This is where custom objects become essential. Model each company account as a custom object with attributes like seat_count, active_users, plan_tier, and renewal_date. Link users to their account object via a relationship. Now you can segment at the user level based on account attributes—for example, reaching account admins whose team adoption is falling below 50%, or notifying champions at accounts approaching their renewal date. Without custom objects, you'd have to flatten account data onto user profiles, which gets messy and hard to maintain.
What data do you need to build effective segments from day one?
You need at minimum: a unique identifier (email or user ID), a status attribute (free/paid/churned), a signup date, and at least one core action event that represents meaningful product engagement—whatever "using the product" means for your specific product. From there you can build basic active, inactive, and never-activated segments. Add feature-usage events, email engagement tracking, and lifecycle stage attributes progressively to increase segmentation depth.
How does segmentation affect email personalisation?
Segmentation determines who gets a message. Personalisation determines what that message says. They work best in tandem. A user in the "At-Risk" segment might receive a different subject line, body copy, and CTA than a user in the "Power Users" segment—even if they're receiving the same campaign. Customer.io's Liquid templating lets you personalise content dynamically based on segment membership or any profile attribute. Our Customer.io Liquid tutorial covers the full syntax.
Can you A/B test segments in Customer.io?
You can A/B test the messages sent to a segment, but you can't directly A/B test the segment definition itself through the UI. For segment testing, you'd need to create two variant segments with different conditions and track their respective journey performance metrics in separate campaigns. This is a more manual process, but it's the most reliable way to validate whether one segmentation approach outperforms another.
What's the biggest segmentation mistake marketers make in Customer.io?
Treating "no purchase in 90 days" the same as "never purchased." These two groups look similar from a recency perspective, but they require completely different messaging. A lapsed buyer needs a win-back message that acknowledges the relationship and offers a reason to return. A non-buyer needs a first-time purchase incentive. Confusing the two—which happens constantly when segmentation is too shallow—produces low conversion rates and wastes offer budget.
How do you keep segments manageable as your account scales?
Use nested segments aggressively. Build modular "foundation" segments (like "Engaged Users," "Paid Accounts," or "Mobile App Users") and reference them inside more specific segments rather than rebuilding conditions every time. Document every segment's purpose and owner in a shared naming convention. Audit and archive unused segments quarterly. The teams that manage large Customer.io accounts well treat their segment library like a codebase—with version control habits and clear ownership.
The Bottom Line
Dr. John Snow didn't need more data than his colleagues. He needed to segment it differently. The deaths were the same data. The pump was the same pump. But by grouping people by behaviour—what water source they used—he found the signal hiding in the noise.
Your Customer.io data is the same way. The events are all there. The attributes exist. The custom objects are waiting to be modelled. The difference between campaigns that convert and campaigns that get ignored is how precisely you group the people you're talking to.
Start with the four core segments. Layer in RFM scoring. Test the AI segment builder on your trickiest audience definitions. Build custom objects for any relationship-based data that doesn't fit a user profile.
Done well, advanced segmentation isn't just a deliverability or engagement tactic. It's how you reduce customer churn, accelerate upgrades, build advocacy, and turn Customer.io into the revenue engine it's capable of being.
Need help getting your segmentation architecture right? NerveCentral is a Customer.io Certified Partner. We've built segment libraries and lifecycle automation systems for businesses at every stage. Get in touch and let's build yours.
Citations
-
Snow, J. (1854). On the Mode of Communication of Cholera. Documented by the Royal College of Surgeons of England. rcseng.ac.uk
-
G2 / Humanic AI. (2024–2025). 32 AI Email Marketing Statistics for 2024–2025. humanic.ai/blog
-
Recurly. (2024). Customer Churn Rate Benchmarks (n = 1,200+ subscription sites, Jan–Dec 2023). recurly.com/research/churn-rate-benchmarks
-
Customer.io. (2025, November 10). Lifecycle Marketing Trends 2026. Author: Molly Evola. customer.io/learn/lifecycle-marketing/lifecycle-marketing-trends-2026
-
Customer.io. (2025, August 13). AI-Powered Segmentation for Better Targeting. Author: Matt Johnson. customer.io/learn/lifecycle-marketing/ai-powered-segmentation-for-better-targeting
-
Sakhawalkar, A. & Pawar, S. (2024). Impact of Behavioral Segmentation on Customer Satisfaction: A Conceptual Review. Sinhgad Institute of Management, Pune. researchgate.net/publication/381096525
-
ScienceDirect / Elsevier. (2024). Predictive analytics in customer behavior: Anticipating trends and personalizing experiences. Results in Engineering, Volume 22. sciencedirect.com/science/article/pii/S2666720724000924
-
Instapage. (2024). Personalisation Statistics: The Data Behind Why It Works. instapage.com/blog/personalization-statistics
-
SmartReach. (2024). Email Segmentation Guide. smartreach.io/blog/email-segmentation-guide
-
Customer.io Documentation. (Updated: December 15, 2025). AI Segment Builder. docs.customer.io/ai/ai-segment-builder
