Don't just ask "who should see this?" — ask "what has this person done or not done, and where are they in their journey?" This article walks through the priority order to use when segmenting audiences, and the mistakes to avoid.

The Appcues App is in beta and rolling out gradually. If you don't see it yet at app.appcues.com, it hasn't been enabled for your account. See Introducing the Appcues App for more.

Segmentation priority

Layer segment dimensions in this order — earlier dimensions carry more weight:

1. Behavior (what the user has or hasn't done)

Actions and inactions with features in your product — "hasn't completed onboarding," "used Feature A but not Feature B," "login frequency dropping." Reach for this first: behavioral segments don't go stale the way attribute-based segments do. If your Appcues data includes custom events, use them before anything else.

2. New vs. existing

How long a user has been in the product. New users need onboarding and guidance; existing users need less interruption and more awareness of what's changed. Use registration date or a "days since first seen" property to split these.

3. Power vs. casual

How intensely someone uses the product. Power users want less interruption and more awareness of new features. Casual users want more guidance. The more messages a power user gets that aren't specific to them, the more they tune out future messages.

4. Admin vs. end user

The bedrock distinction — but it's the floor, not the ceiling. Admins focus on setup and configuration. End users focus on education, value discovery, and habit-building. Never target end users with admin-only actions, like system configuration or integration setup.

5. Persona or job role

Role-based targeting — insurance brokers vs. policyholders, a product manager vs. a developer. Use this when the campaign's content varies meaningfully by what the person does, not just how they use the product.

6. Device or platform

Web vs. mobile vs. mobile web. Factor this in when experiences and triggers diverge across surfaces.

7. Plan tier

Trial vs. free vs. paid. Relevant for expansion and trial-conversion campaigns where the experience changes based on what the user has access to.

Why behavior comes first

Most under-segmentation problems are really "not using events" problems. Teams default to targeting by user property — plan tier, role, signup date — because those attributes are easy to set up. But two admins on the same plan can be in completely different stages of their journey: one is power-using the product daily, the other logged in once and disappeared.

Prioritize events over properties: events capture where the user is right now, not just who they are. Behavioral descriptions ("users who started the setup wizard but didn't finish") produce sharper targeting than attribute-only descriptions ("users on the Pro plan").

Three segmentation mistakes to avoid

Watch for three failure modes:

Targeting the wrong signal

Segmenting on traits when you should segment on outcomes. If the goal is feature adoption, the right signal is "hasn't used the feature," not "is an admin." Always work backward from the outcome you're trying to drive.

Over-segmenting

Creating overly complex segments — unique segments for every experience, "not a member of segment X" stacked across many rules, or CSV uploads treated as live segments. These slow down qualification, are hard to maintain, and give a false sense of precision. If you see this pattern, simplify the targeting.

Under-segmenting

Treating all admins (or all users) as one group when they're clearly in different lifecycle stages. Shipping a single experience to a broad, mixed audience and calling it targeted. Reach for event-based distinctions before settling for a broad attribute-based segment.

What to avoid

• CSV file uploads as a "live" segment. A CSV is a snapshot — it's stale the moment you upload it. Use behavioral or attribute-based logic instead.
• Stacking "not a member of segment X" across many segments. This slows qualification and becomes unmaintainable. Use one-off targeting or simpler logic instead.
• A unique segment per experience. Often unnecessary — one-off targeting directly on the experience is cleaner.

What good segmentation needs from your data

Segmentation is only as good as the data your product sends to Appcues. If you're passing very few user properties and no custom events, you won't be able to build behavioral segments — you'll need to work with what you have until that changes.

If you're passing a large number of properties and events, figure out which ones actually matter for the campaign. Too much data without a clear signal is as limiting as too little.

How to segment better

When you define your audience, include:

• What they've done or not done — "users who created a project but never invited a teammate"
• Where they are in their journey — "second week of trial," "been active for 90 days"
• What outcome you're segmenting toward — "I want to reach people most likely to upgrade"

The more behavioral context you give, the sharper the targeting.