Segmentation and activation in Data 360: building an audience that actually reaches the target

A segment looks like the simplest thing in Data 360 (the platform formerly known as Data Cloud, renamed at Dreamforce in October 2025): drag a few attributes onto a canvas, get a count, publish. Then the count doesn’t match what you saw in Data Explorer, the segment “won’t publish,” half the audience has no email to send to, and someone asks why a suppressed customer got the campaign anyway. None of these are bugs. They’re the parts of segmentation and activation that the demo skips — the difference between defining an audience and actually reaching it.

This post is the whole path: what a segment is and how you build one, the membership-refresh cadences that decide freshness and cost, and then activation — the part where the audience leaves Data 360 for a channel, governed by an identity model, a related-attribute rule, and a consent gate that each quietly decide who really gets the message. If you’ve built segments that looked right and landed wrong, the reasons are almost all in here.

What a segment actually is

A segment is a dynamic population of profiles drawn from your unified data. “Dynamic” is the operative word: it re-evaluates on a schedule, so membership changes as the underlying data changes — someone crosses a spend threshold and joins, someone lapses and drops. You’re defining a rule, not freezing a list.

The first real decision is the segment-on object — the DMO the segment counts and produces rows on. The best-practice choice is almost always the Unified Individual DMO, the identity-resolved profile, because it dedupes source records through your identity resolution rules. Segment on the source-level Individual instead and you lose that dedup — the same real person, present in three source systems, counts as three. Which means your audience quality is inherited from your match rules: over-merge two customers upstream and they become one segment member; under-merge one and they’re three. The segment faithfully serves whatever identity resolution handed it.

You build the population in the Segment Builder canvas, which has two sides that matter: an include side that defines who’s in, and an exclude side for suppression — the customers who match your population but must not be contacted anyway. Criteria are organized into containers, each a condition on an attribute, combined with AND / OR logic and nested to express real rules like “high-value AND (recently active OR high loyalty tier).” You can also nest whole segments: reference an existing audience as an inner segment inside a new outer one, so a well-built “engaged customers” segment becomes a reusable building block instead of logic copied everywhere.

The distinction that decides what’s even possible is direct vs. related attributes. A direct attribute lives on the segment-on object itself. A related attribute is reached by walking a relationship to another object — a customer’s orders, say. For segmentation, a one-to-many relationship (one person, many orders) generally forces an aggregation: you can’t filter on “an order,” you filter on “count of orders > 3” or “sum of order value in the last year.” Hold onto the related-attribute idea, because it comes back with teeth at activation time.

Membership refresh: the cadence that sets freshness and cost

A segment is only as fresh as its last publish — the refresh action that recalculates membership. Data 360 gives you two lanes, and picking wrong is either a stale audience or a bigger bill.

Standard publishing refreshes every 12 or 24 hours and looks back across a wide window of engagement history — up to roughly two years. It’s the right default for most audiences, which don’t meaningfully change within a day.

Rapid publishing refreshes far faster — on the order of every 1 or 4 hours — but it trades away reach to do it: rapid segments look back only over the last 7 days of engagement data, and they come with hard constraints worth knowing before you design around them. There’s a limit on how many rapid segments an org can run, rapid segments take scheduling priority over standard ones, you can’t convert an existing standard segment into a rapid one, and a rapid segment can only activate to a narrow set of targets (Marketing Cloud Engagement and file-storage destinations). Reach for rapid only when the freshness genuinely drives an action inside that 7-day window; otherwise the 2-year look-back of a standard publish is the feature, not the limitation.

And you can choose not to schedule at all — a static, one-time evaluation. That’s the correct setting for a test segment you’re just validating, and it matters for cost: publishing consumes credits, so a scheduled test audience quietly meters every cycle for an audience nobody activates.

Publishing is asynchronous. A segment doesn’t recalculate the instant you save it; it queues, runs, and completes on the platform’s clock. Most “my segment is stuck” reports are really “publishing hasn’t finished yet” — and most “the count is wrong” reports are something more interesting.

Why your count never matches Data Explorer

This one deserves its own section because it burns hours. A segment’s member count frequently won’t equal what the same filter returns in Data Explorer, and it’s not a defect. The count reflects the segment’s delta processing — after a refresh, the reported membership reflects records evaluated as currently qualifying, not a raw row count of everything that ever matched. Layer on date filters, time-zone handling, and identity-resolution changes between runs, and two “identical” numbers legitimately diverge. Salesforce documents this directly (there’s a knowledge article on segment counts not matching). The practical takeaway: treat the published segment count as the source of truth for what will activate, reconcile against Data Explorer with the delta and time-zone behavior in mind, and don’t chase a discrepancy that’s expected by design.

Activation: where the audience leaves the building

Segmentation defines the audience. Activation is the act of publishing that segment to an activation target — a saved destination that holds the authentication and authorization for an external platform. This is the step that turns a population in Data 360 into an audience a channel can act on, and it’s where several new decisions appear.

The targets cover the obvious surfaces and then some: Marketing Cloud Engagement (down to the business unit), Marketing Cloud Personalization, B2C Commerce, back into Salesforce CRM / Data 360 itself, and the advertising and data-partner ecosystem — Google Ads and Display & Video 360, Meta, LinkedIn, Amazon Ads, and partners like LiveRamp, Criteo, and The Trade Desk — plus file-storage destinations. Each target stores its own auth, so activation is “publish this audience here,” repeatable across many destinations.

The concept that governs what you can actually send is activation membership — the grouping of DMO fields chosen for a given activation. It decides two things that determine whether the message lands:

• Which contact points go out. Only contact points mapped into the activation membership are selectable, and when a profile has several candidate values (three emails across three sources), a source priority order decides which one is sent. An audience with no mapped, valid contact point is an audience that reaches no one — the silent failure behind “we activated but nobody got it.”
• Which attributes ride along for personalization. Here’s where the related-attribute rule bites: in activation, a related attribute is only usable when every relationship on the path is one-to-one or many-to-one. No one-to-many fan-out. You can carry a customer’s single loyalty tier or their account’s region; you can’t naively carry “their orders,” because one person to many orders is exactly the fan-out activation won’t resolve. Architects who modeled a segment fine and then can’t attach the field they wanted are almost always hitting this rule.

Two more boundaries shape every activation. First, data spaces: segments and activations are scoped to a data space, and you can’t cross that boundary within a single activation — the partition that decides who unifies with whom also decides who can be activated together. Second, consent. Consent Management for Data 360 gates who may be contacted, and consent rules apply suppression at activation time — the mechanism that’s supposed to keep an opted-out customer out of the send. If suppression seems to “miss,” the usual cause is a consent record that hasn’t propagated or a rule not wired into the activation; verify the consent path is actually in the activation, not just present in the data.

Segments are for channels; agents ground on something else

A useful boundary to draw, because teams building both blur it: a segment is a marketing-facing audience meant to be activated to a channel. It is not how you ground an Agentforce agent. When an agent mid-conversation needs “everything about this one customer, joined, now,” that’s a data graph serving a pre-joined record in milliseconds — retrieval, not audience-building. Same unified data underneath, two entirely different consumption surfaces: segment → activate to a platform versus data graph → ground an agent’s answer. Don’t try to make a segment do an agent’s grounding, or a data graph do a marketer’s targeting.

There’s also a near-real-time cousin to activation worth knowing. A batch activation is scheduled, carries a rich attribute payload, and targets marketing and ad platforms. A data action (and the broader family of Data 360-triggered automation) fires near-instantly on an event with a leaner payload, to a platform event, webhook, or Marketing Cloud. If the requirement is “publish an audience to an ad platform nightly,” that’s activation. If it’s “the moment this signal crosses a threshold, fire something,” that’s a data action, not a faster segment. Picking the right surface here is the same latency-vs-payload trade we map in the automation decision guide.

Watching the loop close: Segment Intelligence

One capability closes the gap between “we activated” and “did it work.” Segment Intelligence pulls campaign and engagement metrics back from activation targets into Data 360, so segment performance and ROI become data you can analyze and act on rather than a report living in the ad platform. It’s the difference between firing audiences into the dark and running a measured loop — and it’s the natural place to notice that the rapid-publish audience you spun up isn’t outperforming the standard one enough to justify its cost.

What to actually do

Segment on the Unified Individual so your audience inherits deduped identities, and remember the audience is only as good as the match rules beneath it. Build the population with clear include/exclude logic, use nested segments to reuse audiences instead of copying logic, and default to standard publishing unless a genuine within-7-days action justifies rapid — leaving true test segments unscheduled so they don’t meter. Expect the published count to diverge from Data Explorer and know why. At activation, treat activation membership as the real design surface: map a valid contact point or the send reaches no one, respect the one-to-one/many-to-one path rule for any attribute you want to carry, keep everything inside one data space, and confirm consent suppression is wired into the activation itself. Then close the loop with Segment Intelligence so you’re measuring reach, not assuming it.

Do that and a segment stops being a number on a canvas and becomes an audience that actually arrives — the right people, reachable, permitted, and measured. Get any one of those wrong and the segment can look perfect and still land on no one, or on the wrong someones. The building is easy; the reaching is the engineering. If you’re standing up segmentation and activation on Data 360 and want the identity, contact-point, consent, and cost calls made deliberately, talk to us — designing the data foundation so audiences reach the right people is exactly the work we do.

Understanding the basics

What is the difference between a segment and an activation in Data 360?

A segment is a dynamic population of profiles defined from your unified data — a rule that re-evaluates on a schedule to decide who’s in the audience. An activation is the act of publishing that segment to an activation target, an external destination (Marketing Cloud, an ad platform, a data partner, file storage, or back into CRM) that holds its own authentication. Segmentation decides who; activation decides where they go and what fields and contact points travel with them. A segment that’s never activated reaches no one, and an activation with no valid mapped contact point also reaches no one.

Why doesn’t my Data 360 segment count match Data Explorer?

Because the two count differently, by design. A published segment’s membership reflects delta processing after a refresh — records currently qualifying — rather than a raw count of everything that ever matched, and date filters, time-zone handling, and identity-resolution changes between runs shift the number further. Salesforce documents this behavior directly. Treat the published segment count as the source of truth for what will activate, and reconcile against Data Explorer with the delta and time-zone behavior in mind rather than assuming a defect.

When should I use rapid publishing instead of standard?

Only when the freshness drives an action inside a short window. Standard publishing refreshes every 12 or 24 hours and looks back across a wide history (up to about two years). Rapid publishing refreshes far faster (on the order of every 1 or 4 hours) but only looks back over the last 7 days of engagement data, is limited in how many rapid segments an org can run, takes priority over standard segments, can’t be created by converting an existing standard segment, and can only activate to a narrow set of targets. If the audience doesn’t need to act on something within days, standard publishing’s longer look-back is the better choice — and publishing consumes credits either way, so leave test segments unscheduled.

---

Standing up segmentation and activation on Data 360 and finding audiences that look right but land wrong — missing contact points, attributes that won’t attach, counts that won’t reconcile, suppression that leaks? Talk to us — building the data foundation so segments reach the right people, permitted and measured, is exactly the work we do.