Segments

The Segments page is where Sift's AI groups your audience into meaningful, reusable cohorts. While Overview highlights individual users worth your attention, Segments zooms out: it answers "Who is in my community?" by automatically classifying every user into a hierarchy of segments and personas based on how they behave across every connected platform.

At a glance

What you're looking at

Sift continuously rebuilds this taxonomy as new activity comes in, so the segments you see always reflect your community as it is today, not how it was when you last set up a static filter. That makes Segments useful for both analysis (understanding the shape of your audience) and routing (finding everyone who matches a real-world description).

How segments are built

Sift's models analyze each user's posts, replies, sentiment, lifecycle stage, and engagement patterns and classify them into a two-level structure:

• Segments: broad, top-level cohorts such as a customer role, a risk category, or a value tier.
• Personas: finer-grained groupings nested under each segment that capture a more specific behavioral pattern.

Every user can belong to multiple personas with a confidence score, so the same person might appear as both a Power user and a Brand advocate. This is what makes Segments useful for both routing and analysis: you can find every customer who matches a real-world description, not just those who match a hand-built filter.

Headline KPIs

A four-tile strip at the top summarizes the health of your segmentation:

• Classified users: Total number of users placed into at least one segment.
• Assignments: Total number of user-to-persona assignments. A single user contributes multiple assignments when they match multiple personas.
• Confidence: Median classification confidence across all assignments. Higher means Sift is more certain about its placements.
• Last updated: Timestamp of the most recent taxonomy rebuild, so you know how fresh the data is.

How to read it: if Confidence is low, treat the finer personas with more skepticism and verify before routing on them. A stale Last updated means recent activity isn't reflected yet.

Trust & risk

A horizontal chip strip surfaces the categories your trust-and-safety team will reach for first. Each chip shows the count of users currently carrying that risk label:

• Impersonators: Accounts flagged as posing as someone else.
• Spam / scam: Accounts identified as posting spam or scam content.
• Detractors: Users currently classified as expressing negative sentiment toward your brand.
• Frequent complainers: Users with a sustained pattern of complaints.
• At-risk: Users showing signals of potential disengagement or churn.
• Churning: Users who appear to be actively leaving the community.

The strip only renders chips that have a non-zero count, so its presence on the page is itself a signal of where attention is needed.

Segment-type mix

Beneath the trust strip, a stacked bar visualizes how your overall assignments break down across the high-level segment-type classifiers. Sift groups every segment into one of the following categories:

Each row shows the absolute number of assignments and its percentage of the total, giving you a fast read on the composition of your community.

Segment & persona tree

The bulk of the page is the segment tree: a searchable, expandable list of every segment Sift has built for your organization, with the personas inside each segment nested underneath. Each segment row shows its name, classifier icon, total user count, and a chevron to expand or collapse it. A search box at the top filters by segment name, persona name, summary, or label, and an Expand all / Collapse all toggle makes it easy to scan the full taxonomy.

Inside each segment, persona cards carry the detail:

• Name and short summary: A human-readable description of the persona.
• Users: How many users are currently assigned.
• Confidence: Median assignment confidence for this persona.
• Value: Average value score of the users in the persona.
• Sentiment: Average sentiment, color-coded.
• Trend: Whether the persona is rising, stable, or declining based on the activity trajectory of its members.
• Top labels: The badges most frequently associated with users in this persona.
• Lifecycle mini-bar: A compact stacked bar showing how members are distributed across the lifecycle stages (Engaged, Active, New, Casual, Lapsing, Dormant).

Clicking any persona card opens its dedicated detail page.

How Segments fits with Overview and People

• Overview tells you which individuals need attention right now.
• Segments tells you which groups exist and how they're shaped.
• People is where you actually browse and act on the members of any segment.

Use Segments to validate that Sift's taxonomy matches how you think about your audience, to spot risk concentrations, and as the jumping-off point into People when you want to work with the underlying users.

Related