SocialHub.AI

One ID & Tags · Segments

One audience engine — four ways to build, every one grounded in real members.

Describe an audience in plain language, build it with rules over your behavioral and value data, find more members like your best ones, or let AI surface the plays you're missing. Whichever way you start, the audience comes sized and validated against your actual members.

Create an audience4 ways

Describe

Say it in plain language

Build

Build it with rules

Look-alike

Find more of your best members

Discover

Let AI surface what you're missing

every one checked against real members

One grounded audience

sized, validated, ready to activate

The problem

A blank condition builder only helps if you already know what to build.

Blank-page segmentation

Most tools wait for you to describe an audience. The opportunities you forget to describe stay invisible.

Data locked away from marketers

The recency, spend and category signals that make a good segment sit in analytics — not in the hands of the person building the campaign.

Guesswork on reach

Without a real count you can't tell whether an audience is worth a campaign, too small to bother, or accidentally your whole base.

Four ways to build

Start however you think — the engine does the rest.

The same audience engine meets you where you are: type it, click it, clone your best members, or let AI propose it. Every path ends at a validated, sized, ready-to-activate audience.

1

Describe

Say it in plain language

Type “lapsed VIPs in Shanghai who bought skincare” and the AI drafts real, ready-to-check rule candidates — validated against your fields, sized against your members, explained in plain English. You review and save; nothing is guessed.

2

Build

Build it with rules

A visual builder over your governed data — lifecycle, points and tier, purchase recency and frequency, spend and average order value, category and brand affinity, store, channel and consent. Every rule reads certified data, so the audience means the same thing everywhere.

3

Look-alike

Find more of your best members

Pick a winning segment, a tag, or a handful of members as a seed and Flash finds the members who most resemble them — inside your own base, weighing shared traits and buying rhythm alongside spend tier, favorite product category and churn outlook. Every match comes with a “Why they're similar” summary. Never uploaded to an ad platform.

4

Discover

Let AI surface what you're missing

The advisor audits which standard retention plays you already run and proposes the ones you've left on the table; auto-discovery finds natural groupings in your member base you'd never think to describe. Each one arrives sized and ready to adopt.

What you can segment on

Behavioral, value and category data — in the marketer's hands.

The signals other stacks bury in an analytics module are first-class rules here. Segment on how recently, how often and how much a member buys, what categories they favor, where they shop, how they can be reached — even attributes unique to your business.

Lifecycle & identity

New, active, lapsing or dormant · enrolment date · sign-up channel · birthday month

Loyalty

Points balance · tier · lifetime points earned · redemptions · referrals

Purchase behavior

Recency (days since last purchase) · order count · net spend · average order value · return rate

Category & brand affinity

Bought a category or brand · primary category · breadth of categories · replenishment timing

Store & channel

Home store · store visit recency & count · preferred channel · orders by channel

Consent & reachability

Email marketing consent · push-reachable · do-not-sell · suppression status

Value & engagement tiers

Historical value tier · engagement score · recency/frequency/value bucket · intent signal

Predicted next actions

Days until expected next purchase (negative = overdue) · click propensity (0–100 likelihood of clicking a campaign)

Your own custom fields

Any attribute you added to members — membership number, preferred store, a child's birthday — segmentable the moment it exists

These read the same certified recency, frequency, value and category data the rest of the platform uses — so a “high-value, at-risk skincare buyer” means exactly the same thing in a segment, a dashboard and an agent's decision.

The two predicted conditions stay honest. An expected next purchase date is each member's own buying rhythm projected forward, with a ± window from how regular they actually are — it only exists once a member has four or more purchases and a reasonably regular rhythm, and overdue shows as overdue, never a made-up date. Click propensity is learned nightly from your own recent sends — clicks only, never opens, which mail privacy inflates — and only members actually emailed recently get a score. “Days until expected next purchase between −14 and 3” is a ready-made replenishment-reminder audience, in one filter.

Grounded by default

No audience leaves the builder unchecked.

Real reach, not a guess

Every audience returns a live count against your actual members — how many, and what share of the base — so you know whether it's worth a send before you build the campaign.

Caught before you send

A built-in check flags audiences that match no one, barely filter anyone, contain an impossible rule, or duplicate one you already have.

Audiences that subtract

Target one audience except another — say “win-back candidates, but not anyone already in the VIP flow” — in a single definition.

Measured value, never invented

When Flash shows a cohort's value, it's real historical spend — not a predicted lifetime value or an uplift model. The numbers on your screen are the numbers an agent acts on.

The advisor

It also tells you which audiences you forgot to build.

Beyond building on demand, Flash audits your program: it compares the standard retention plays — new, lapsing, dormant, VIP, high-frequency, birthday — against what you already run, and proposes only the gaps. Each suggestion is checked against your members and is one click to adopt.

AI proposes the plays; if it's unavailable, built-in preset plays take over. The advisor always returns a real, usable answer.

Standard retention plays

have vs. missing
  • New Members

    Enrolled in the last 30 days

    covered
  • VIP / High-Value

    Lifetime earned over threshold

    covered
  • Lapsing

    No activity in 90+ days

    suggested
  • Dormant win-back

    Long-inactive, re-engageable

    suggested
  • High-Frequency

    Repeat buyers

    suggested
  • Birthday this month

    Time-based outreach

    suggested

Start with a running program

Ready-made plays, installed in a click.

A library of proven retention audiences ships with the platform — install them in one click and they size themselves against your base. Adjust, or use as-is.

New Members (30 days)Inactive 90+ daysBirthday This MonthChampionsAt RiskHibernatingHigh SpendersStore Captures — No Purchase+ more

Built to be used

An audience isn't the finish line.

Straight into campaigns

Pick a segment as a campaign audience and it resolves live at send — decrypting contact info, honoring consent, and subtracting any audience you chose to exclude.

Visible on every profile

Open a member and see exactly which segments they belong to right now — no stale, snapshotted lists.

One engine with Tags

The same rule builder powers Tags, and tags are a rule you can segment on — so audiences and labels stay perfectly composable.

AI & innovation

AI as a retention strategist, not a query box.

AI drafts audiences, finds lookalikes of your best members, discovers cohorts you'd never describe, and flags the plays you're missing — but it never decides sizing on its own. Every number is measured, and the rule builder, the advisor, lookalikes and discovery keep working with deterministic logic even when the AI is unavailable.

Proactive, not reactive

Instead of waiting for a request, it analyzes coverage and recommends the next audience to add.

Grounded in your data

Cohort sizes and value are real historical measurements of your own members — no invented audience sizes, no predicted lifetime value.

Degradation-safe

The rule builder, advisor, lookalike and discovery paths fall back to deterministic rules and presets, so the core stays useful even when the AI is unavailable.

Lookalikes, without the leak

Find more of your best members — in your own base, not an ad network's.

Ad-platform lookalikes require handing your customer list to a third party. Flash finds resemblance inside the members you already own — weighing the traits and buying rhythm your best members share, plus their spend tier, their favorite product category and their churn outlook. You get a sized, deliverable audience, and nothing is uploaded anywhere.

And the result explains itself: a “Why they're similar”summary shows exactly what each match is built on — for example “81% in the same VIP spend tier · 55% shop the same top category” (illustrative). Anything Flash doesn't have data for is skipped honestly, never faked.

What changes for the business

Segmentation stops being a blank page — audiences get described, built, discovered and sized, then flow straight into the campaigns that use them.

Four ways

describe, build, look-alike, or let AI propose

Real reach

every audience sized against your actual members

Ready to run

from audience to a live campaign, consent honored

Each audience checked against your real members for estimated reach.

Build the audiences you have — and the ones you're missing.

400M+
50+
12+