Cohorts

A cohort is a saved, living group of users — “Power users”, “Churn risk”, “Enterprise trials”, “Everyone who dropped at checkout”. You define it once, and Replayfy keeps its membership current for you: people join and leave the cohort as their behavior and properties change. From then on, a cohort is a lens you point at your data — open its members, drill into their session replays, and target the exact audience you care about.

Cohorts group users, not sessions. Every member is an end user you identify from your SDK (or an anonymous-but-tracked visitor). The richer the traits and events you send, the more precisely you can slice your audience — see Feed cohorts with the right data.

Two ways to define membership

Every cohort is either auto-updated or manual. You choose this when you create it.

Membership	How it works	Best for
Auto-updated	You write rules. Anyone who matches is a member; anyone who stops matching drops out — automatically, as new data arrives.	Living segments — “active this week”, “on the Enterprise plan”, “fired `payment_success`”.
Manual	You hand-pick a fixed set of users. Membership only changes when you edit it.	Curated lists — VIP accounts, a hand-built beta group, a point-in-time snapshot.

Auto-updated cohorts stay fresh on their own — you never have to press a button to keep them accurate. As sessions and events flow in, membership is re-evaluated continuously (typically within a few minutes). You can still force an immediate refresh with Recompute from any cohort’s ⋯ menu.

Attribute vs behavioral cohorts

Auto-updated cohorts are built from rules. Rules fall into two families, and you can mix them freely in one cohort:

Attribute cohorts match on who a user is — the traits and properties you send with identify. Plan, device, browser, OS, country, city, email, name. These change rarely, so these cohorts settle quickly.
Behavioral cohorts match on what a user does — their activity and the events you track. “Has ≥ 12 sessions”, “seen in the last 7 days”, “fired checkout_started”, “currently online”. These are re-evaluated as activity happens.

Supported rule conditions

Each rule is a field, an operator, and a value. Add as many rules as you like — a user must match all of them to be a member.

Field	Type	Operators	Example
Plan	Attribute	is / is not	Plan is `Enterprise`
Device	Attribute	is / is not	Device is `Mobile`
Browser	Attribute	is / is not	Browser is `Safari`
OS	Attribute	is / is not	OS is `iOS`
Country	Attribute	is / is not	Country is `United States`
City	Attribute	equals / contains / starts with / ends with	City contains `London`
Email	Attribute	equals / contains / starts with / ends with	Email ends with `@acme.com`
Name	Attribute	equals / contains / starts with / ends with	Name contains `Admin`
Total sessions	Behavioral	≥ / > / =	Total sessions ≥ `12`
Last seen	Behavioral	within the last N days / more than N days ago	Last seen within the last `7` days
Currently online	Behavioral	is true / is false	Currently online is true
Did event	Behavioral	fired / did not fire	Did event fired `payment_success`

“Did event” reads straight off the events you send with track. Use it for any “users who completed X” question — fired for users who did it, did not fire for users who never have.

Feed cohorts with the right data

Attribute and behavioral rules only work if your SDK is sending the underlying signal. Identify your users with the traits you want to segment on, and track the events you want to build behavioral cohorts from.

import { initReplay } from "@replayfyapp/browser";

const replay = initReplay({
  apiKey: "rpl_pk_YOUR_KEY",          // publishable key — safe to ship
  apiHost: "https://us.replayfy.app", // your ingest host
});

// Attributes → power "Plan is Pro", "Email ends with @acme.com", etc.
replay.identify("user_8f21", {
  email: "[email protected]",
  name: "Ada Lovelace",
  plan: "Pro",
});

// Events → power "Did event fired payment_success"
replay.track("payment_success", { amount: 4900, currency: "usd" });

<!-- Once the async loader from Web setup has put the `Replayfy` global on the
     page, the same methods are available on it (no returned controller to hold). -->
<script>
  Replayfy.identify("user_8f21", { email: "[email protected]", plan: "Pro" });
  Replayfy.track("payment_success", { amount: 4900, currency: "usd" });
</script>

Only ship the publishable key (rpl_pk_…) in client apps. Server keys (rpl_sk_…) are for backend calls and must never appear in shipped code. See Track events for naming conventions and Web setup for the full install.

Create a cohort

Open the builder

Go to Cohorts and click New cohort (or press N then C). You’ll land on the template gallery.

Start from a template or from scratch

Pick a ready-made audience to pre-fill the builder, or choose Start from scratch for a blank cohort. Templates are grouped by goal:

Category	Templates
Engagement	Power users · Currently online · Recent purchasers
Lifecycle	Churn risk · Dormant users
Monetization	Enterprise accounts · Free power users
Environment	Mobile users · Safari users

Name it and choose membership

Give the cohort a clear Name and an optional Description (“What is this segment, and how should the team use it?”). Then choose Auto-updated or Manual.

Add rules (auto-updated cohorts)

Click Add rule and pick a field, operator, and value. Every rule you add narrows the audience — members must match all of them. As you edit, a live preview shows exactly how many users match right now, with sample avatars of who they are.

Create

Click Create cohort. An auto-updated cohort is populated immediately from its rules; a manual cohort starts empty, ready for you to add members.

The live preview count is always real — it runs your rules against your actual users. If it reads 0 users match right now, loosen a rule before creating.

Build a manual cohort

Choose Manual membership when you want a fixed, hand-curated list rather than a rule-driven one. Create the cohort first, then add members from anywhere you see a user:

On the Users page, open any user’s ⋯ menu and choose Add to cohort.
In a user’s detail panel, use Add to cohort and toggle the cohorts they belong to.

Only manual cohorts accept hand-added members. An auto-updated cohort recomputes from its rules, so a manually added user would be dropped on the next refresh — that’s why the picker only lists manual cohorts.

Use a cohort across Replayfy

A cohort is a saved segment, so you reach for it wherever you work with users.

View the members

From the cohort’s row (or its ⋯ menu), choose View members. This opens the Users list scoped to that cohort — a banner names the cohort so you always know what you’re looking at.

Watch their sessions

From the scoped user list, open any member to see their profile and jump straight into their session replays — so “show me what churn-risk users are actually doing” is two clicks away.

Combine with other filters

The scoped Users list still honours your other filters — country, platform, online status, last-seen window — so you can drill from a broad cohort down to an exact slice without leaving the page.

The Cohorts page itself gives you an at-a-glance read of every segment: member count, whether it’s Auto-updated or Manual, and when it was last computed. Top-of-page stats sum your active cohorts and total members.

Create a cohort from a funnel drop-off

One of the most useful cohorts is “everyone who abandoned here”. You can build it directly from a funnel: open a step where users dropped off and choose Create cohort from drop-off. Replayfy first shows you how many identified users reached the previous step but not this one, then materializes them as a manual cohort — a point-in-time snapshot of that exact group.

A drop-off cohort is a snapshot: it captures who dropped in the analyzed window and won’t change as new users drop off later. It includes only identified users (anonymous visitors can’t be added to a cohort), so its size can differ from the session-level drop-off shown on the funnel.

Cohorts created by Replayfy AI

You don’t have to open the builder at all. Ask Replayfy AI to assemble a segment for you — for example, “create a cohort of Pro users who fired payment_success in the last 7 days” — and it builds the rules and saves the cohort. Anything the assistant creates carries a Created with Replayfy AI badge, both in the Cohorts list and in the scope banner when you view its members, so it’s always clear which segments were AI-authored. From there they behave exactly like any other cohort — you can edit, recompute, or delete them.

Manage and clean up

Action	Where	What it does
Recompute	Cohort `⋯` menu	Forces an immediate refresh of an auto-updated cohort’s membership.
View members	Row click or `⋯` menu	Opens the Users list scoped to the cohort.
Search & filter	Cohorts page	Filter by Auto-updated / Manual, or search cohorts by name.
Delete	Cohort `⋯` menu	Permanently removes the cohort and its membership. This can’t be undone.

Funnels

Find where users drop off — then turn any step into a cohort.

Track events

Send the custom events that power behavioral cohorts.

Replayfy AI

Ask the assistant to build a cohort for you in plain English.

Get started

Self-hosting

Install the SDK

Products

Platform

Guides

Help

Two ways to define membership

Attribute vs behavioral cohorts

Supported rule conditions

Feed cohorts with the right data

Create a cohort

Build a manual cohort

Use a cohort across Replayfy

Create a cohort from a funnel drop-off

Cohorts created by Replayfy AI

Manage and clean up

Funnels

Track events

Replayfy AI

​Two ways to define membership

​Attribute vs behavioral cohorts

​Supported rule conditions

​Feed cohorts with the right data

​Create a cohort

​Build a manual cohort

​Use a cohort across Replayfy

​Create a cohort from a funnel drop-off

​Cohorts created by Replayfy AI

​Manage and clean up

Funnels

Track events

Replayfy AI

Two ways to define membership

Attribute vs behavioral cohorts

Supported rule conditions

Feed cohorts with the right data

Create a cohort

Build a manual cohort

Use a cohort across Replayfy

Create a cohort from a funnel drop-off

Cohorts created by Replayfy AI

Manage and clean up