Segment Intelligence in Data 360: closing the loop between the audience you activated and the revenue it drove

Every marketing team can build a segment. Point-and-click in Data 360, drag in a few criteria, publish it to Meta and Google, and you’ve reached your audience. The demo ends there, and so does most of the value — because the question the CFO actually asks is not “did you build the segment,” it’s “which of these audiences made us money, and which one burned budget?” That answer doesn’t live in Data 360. It lives inside the ad platform, in the campaign’s click and conversion data, on the other side of a wall the activation crossed on the way out and never crossed back.

That gap — activation goes out, performance never comes back — is the single most expensive blind spot in a Data 360 marketing stack. You can spend a year perfecting how you build and reach an audience and still have no idea which segments deserve more spend, because the loop was never closed. Segment Intelligence is the feature built to close it: pull engagement and conversion data back into Data 360, harmonize it against the unified profile, and attribute it to the segment that produced it. This post is about the round-trip — what Segment Intelligence measures, how the data comes home, and the licensing and credit realities nobody puts on the slide.

Activation is a one-way street by default

Here’s the mechanic that traps people. When you activate a segment, Data 360 sends a list — hashed emails, device IDs, a LiveRamp identifier — to a destination like Meta, Google Ads, Amazon Ads, The Trade Desk, LinkedIn, or one of the hundred-plus other connectors and activation partners. The ad platform matches that list against its own users and serves them impressions. Everything that happens next — the impressions served, the clicks, the cost, the conversions the platform attributes — is recorded in the ad platform, keyed to the ad platform’s identity graph, not yours.

So you have two halves of a story that never meet. Data 360 knows exactly who was in the segment. The ad platform knows exactly what those people did. Neither side can tell you the thing that matters: did the high-value segment I so carefully built actually outperform the cheap lookalike audience, on revenue, per dollar spent? Most teams answer this by exporting CSVs from three ad consoles into a spreadsheet once a quarter, reconciling by hand, and calling it attribution. It’s slow, it’s stale, and it breaks the moment identity doesn’t line up across platforms.

Segment Intelligence exists because the unified profile is the only place the two halves can meet. You already resolved identity to build the segment; the same identity resolution that unified the profile is what lets performance data flow back and attach to the right person, and from there to the right segment.

What Segment Intelligence actually does

Strip away the marketing gloss and Segment Intelligence is a measurement layer that does three concrete things:

• It brings performance data back. Engagement signals — sends, opens, clicks, impressions, spend — and downstream conversion data are ingested back into Data 360 as engagement data, alongside the profile data you already model. Salesforce positions Segment Intelligence as the capability to track, compare, and optimize audience engagement across marketing channels, which only works if the engagement is physically present in the platform to be tracked.
• It ties that performance to the segment and the profile. Because the data lands against unified individuals, Data 360 can roll it up by segment, by channel, by campaign, or by a custom goal — and compare them on the same footing. That roll-up is what turns “we ran three campaigns” into “segment A returned $4 for every $1, segment B returned $0.80, kill segment B.”
• It surfaces the metrics a budget decision needs. Return on ad spend (ROAS) is the headline driver; cost per acquisition, conversion rate, and reach efficiency sit underneath it. You set the KPIs that matter to your business and monitor performance at the level you care about — audience, campaign, channel, or goal.

The important reframe: Segment Intelligence is not another way to build an audience. It’s the “did it work” layer that sits after activation. Segmentation decides who you talk to; Segment Intelligence decides who you keep talking to, using money as the tiebreaker. If you’ve read our piece on calculated versus streaming insights, think of Segment Intelligence as the applied, marketing-shaped consumer of exactly that kind of computed metric — ROAS is a calculated insight with a paycheck attached.

Where it sits: the Marketing Intelligence family on Tableau Next

Segment Intelligence isn’t a standalone island. It’s part of Marketing Intelligence, Salesforce’s marketing-analytics solution that stitches together Data 360, Agentforce, Einstein, and Tableau Next to turn third-party marketing performance data into something you can act on without hand-building dashboards. The visualization surface — cross-channel ROI, attribution, and journey impact — is a Marketing Intelligence dashboard rendered on Tableau Next, the agentic BI layer that replaced the old CRM Analytics story.

That lineage matters for two reasons. First, it tells you where to look in Setup: the feature installs a Tableau Next workspace and provisions the data spaces it needs. Second, it tells you what you’re buying into — a stack where the analytics, the data platform, and the agent all share one governed semantic layer, so the ROAS number the dashboard shows and the ROAS number an agent reasons over are the same number. That’s the whole argument for grounding agents on a semantic layer instead of letting each surface invent its own math.

The round-trip, end to end

Here’s the data flow, and it’s worth reading slowly because each hop is a place a real deployment breaks:

1. Segment built in Data 360        (unified profiles, profile + engagement criteria)
        │  activate
        ▼
2. Ad platform / destination        (Meta, Google, Amazon, Trade Desk, LinkedIn…)
        │  campaign runs — impressions, clicks, spend, conversions recorded here
        ▼
3. Engagement + conversion data     ingested BACK into Data 360 as engagement data
        │  harmonize + identity-resolve against the unified profile
        ▼
4. Attribution roll-up              performance rolled up by segment / channel / campaign / goal
        │
        ▼
5. Marketing Intelligence dashboard (Tableau Next) + Agentforce recommendations

Step 3 is the one people underestimate. Getting performance data back in is a real ingestion job — a connection to the ad platform’s reporting, a mapping of its schema onto your engagement model, and the same modeling discipline you’d apply to any source. A conversion event with a sloppy identity key attaches to the wrong profile and quietly corrupts every segment it rolls into. Segment Intelligence changes where the loop closes; it doesn’t excuse you from getting the mapping right.

Setting it up: the honest version

The entry point is Marketing Intelligence. In Setup, use Quick Find to open Marketing Intelligence, and the guided setup installs the Tableau Next workspace, provisions the required data spaces, and enables the feature. From there the work is the part the wizard doesn’t do:

1. Connect the performance sources. Each ad platform whose spend and conversions you want to measure needs a connection back into Data 360. Native connectors and trusted partners cover the major destinations, but “activation connector” and “performance-ingestion connector” are not always the same object — confirm both directions exist for the platforms you actually run on before you promise a dashboard.
2. Model the engagement data. The incoming performance data lands as engagement data and has to map onto your data model to be attributable. Decide the grain — event per impression, per click, per conversion — and the key that ties it to a unified individual.
3. Define the goals and KPIs. Segment Intelligence measures against the KPIs you set. ROAS is the default headline, but if your business optimizes for pipeline influenced or trial-to-paid conversion, model that as the goal so the roll-up answers your question, not a generic one.
4. Assign access. The dashboard lives in Tableau Next and the underlying data lives in Data 360, so the humans who need it require access to both. This is a frequent day-one stumble: the analyst can see the segment but not the performance, or vice versa, because the two grants were treated as one.

None of that is exotic, but all of it is real configuration, and the demo skips every step.

Where Agentforce enters — and where it shouldn’t

The reason Salesforce folds Einstein and Agentforce into this stack is that a ROAS table is only useful if someone acts on it, and the acting is repetitive. Marketing Intelligence generates AI campaign summaries, and Agentforce can autonomously identify underperforming campaigns and recommend improvements — reallocate budget away from the segment returning $0.80, lean into the one returning $4. This is a genuinely good fit for an agent: high-volume, rule-shaped, tied to a metric with a clear direction.

But keep the human-in-the-loop line exactly where it belongs. An agent recommending “shift spend from segment B to segment A” is assistive and safe. An agent executing a five-figure budget reallocation across live campaigns without a human approving it is a different risk class, and it’s the kind of consequential action that belongs behind an approval gate, not inside an autonomous loop. The same discipline we argued for in Agentforce for Marketing applies here: let the agent assemble the recommendation with the numbers attached; let a marketer own the money decision.

The gotchas that don’t make the keynote

This is a consumption feature, and measurement isn’t free. Ingesting engagement data back in, harmonizing it, and computing calculated metrics all consume Data 360 credits — the same credit meter that everything else in the platform runs on. Performance data is high-volume by nature (every impression is a row), so a naïve “ingest everything at event grain, recompute hourly” configuration can run up a bill that dwarfs the ad spend you were trying to optimize. The credit-optimization playbook applies directly: aggregate where you can, match refresh frequency to how often a budget decision is actually made, and don’t stream what a daily batch would answer.

Packaging and naming are moving. “Data Cloud” is now Data 360, “Marketing Cloud” marketing analytics has consolidated under Marketing Intelligence, and the exact SKUs, connector availability, and feature gating in this area have changed release to release. Treat any capability here as something to verify against current release notes and your own contract before you design around it — especially which ad platforms have bidirectional connectors, because that list is where the round-trip physically succeeds or fails.

Attribution is still hard, and this doesn’t make it a solved problem. Segment Intelligence tells you what happened against the segment you built; it does not resolve the eternal argument about credit — last-touch versus multi-touch, view-through versus click-through, the platform’s self-reported conversions versus your own. Cross-channel attribution and journey impact are on the dashboard, but the model behind them is a choice you own. Don’t let a clean ROAS tile lull you into treating a hard measurement problem as if it were now automatic.

The takeaway

Building and activating a segment is the visible, demo-friendly half of a Data 360 marketing stack, and it’s the half that decides nothing on its own. The decisions — where the next dollar goes, which audience to double down on, which lookalike to kill — depend on performance data coming back and attaching to the segment that produced it. Segment Intelligence is the layer that closes that loop, sitting inside the Marketing Intelligence family on Tableau Next, tying ad-platform engagement and conversion data to your unified profiles so ROAS becomes a number you can see per segment instead of a spreadsheet you reconcile per quarter. Set it up with the ingestion and identity discipline any source demands, mind the credit meter because performance data is voluminous, keep the budget decision human, and verify the current packaging against your contract. Do that, and the loop that was open by default finally closes — which is the only way marketing spend ever gets honestly optimized.

Understanding the basics

What is Segment Intelligence in Data 360?

Segment Intelligence is a Data 360 capability for tracking, comparing, and optimizing the performance of activated audience segments across marketing channels. It brings engagement and conversion data back into Data 360 after activation, harmonizes it against the unified profile, and rolls it up by segment, channel, campaign, or goal so you can measure metrics like return on ad spend (ROAS) per segment. It’s part of the Marketing Intelligence solution, which combines Data 360, Agentforce, Einstein, and Tableau Next, and its dashboards render on Tableau Next.

How is Segment Intelligence different from segmentation and activation?

Segmentation and activation are the outbound half — you build an audience from unified profiles and push it to destinations like Meta, Google, or LinkedIn. Segment Intelligence is the inbound, measurement half: it pulls campaign performance back in and attributes it to the segment that produced it. Segmentation decides who you reach; Segment Intelligence decides, using revenue and spend, which segments were worth reaching. You need both to close the loop.

Does Segment Intelligence cost extra to run?

Yes, in two senses. It’s part of the Marketing Intelligence packaging, so confirm licensing against your contract, since Salesforce’s naming and SKUs in this area have changed release to release. And operationally it consumes Data 360 credits: ingesting high-volume engagement data, harmonizing it, and computing performance metrics all meter against your credit balance. Because impression- and event-grain data is voluminous, match ingestion grain and refresh frequency to how often you actually make budget decisions rather than defaulting to real-time everything.

---

Trying to get the measurement loop closed — engagement data ingested cleanly, identity resolved so performance attaches to the right profile, and the credit math under control before it dwarfs your ad spend? Talk to us. Getting the data foundation right so the numbers can be trusted is exactly the work we do.