Salesforce CDP Genie Features 2026: An APAC Retail Reality Check

Quick Answer: Genie is now Salesforce Data Cloud, rebranded again as Data 360 in 2025. Versus the legacy Marketing Cloud CDP, the 2026 feature set adds streaming ingestion, zero-copy federation with Snowflake and Databricks, market-scoped identity rulesets, Data Graphs and Agentforce grounding on unified profiles.

Success looks like this: a beauty retailer with 140 stores across Hong Kong, Singapore, Kuala Lumpur, Bangkok, Manila and Ho Chi Minh City runs one segment — "lapsed VIP, high AOV, fragrance-led, no purchase in 120 days" — and it resolves consistently in every market. The Hong Kong member who shops in Singapore on holiday doesn't show up as two people. The WhatsApp opt-in from Malaysia doesn't get burned by a Thai campaign send. The store associate's iPad shows the same lifetime value the marketer sees in the campaign builder, calculated from the same data, at the same moment.

Related reading: Salesforce Marketing Cloud Next AI Agents: The APAC Ops Reality

Work backwards from that and you understand why the Salesforce CDP Genie features 2026 conversation matters more to APAC retail operators than to almost anyone else. Multi-market is the default here, not an expansion phase. Six markets means six identity patterns, six messaging channels, six privacy regimes, and — in most of the groups I've worked with — six different people who each believe they own the customer record.

Related reading: Google's $40 Billion Anthropic Investment: What APAC Teams Do Now

Related reading: E-Commerce Platform Replatforming Guide 2026: The APAC Playbook

Related reading: DeepSeek v4 LLM Performance Benchmark: The APAC Cost Math

Related reading: Shopify Plus vs Adobe Commerce 2026 Cost: The APAC Verdict

From Genie to Data Cloud to Data 360: what actually changed

First, the naming, because it causes real confusion in vendor evaluations.

"Genie" was announced at Dreamforce 2022 as a real-time customer data platform built on the Salesforce Hyperforce infrastructure (Salesforce newsroom, September 2022). It was renamed Salesforce Data Cloud in 2023. At Dreamforce 2025, Salesforce consolidated its data and agent branding again — Data Cloud became part of the Data 360 family, sitting underneath Agentforce 360 (Salesforce press materials, October 2025).

So if your search history includes "salesforce genie vs data cloud" or "data cloud to data 360," the honest answer is that these are three names for an evolving product line, not three products. The legacy Salesforce CDP — the one that shipped inside Marketing Cloud around 2020 as Customer 360 Audiences — is genuinely a different animal, and that's the comparison worth making.

The architectural break that mattered was zero-copy. Legacy CDP ingested and duplicated data into a proprietary store. Data Cloud/Data 360 supports zero-copy federation with Snowflake, Databricks, BigQuery and Amazon Redshift, meaning a Data Cloud data model object can point at a table that never leaves your warehouse (Salesforce developer documentation). For a regional group whose transaction warehouse sits in Singapore for tax and audit reasons, that's not a performance footnote — it's the difference between a governance approval and a governance rejection.

The 2026 capabilities that actually move retail operations

Strip out the keynote language and there are five capability clusters that change day-to-day work for a multi-market retail team.

Real-time data streams and streaming insights

Web, app and POS events land as streams rather than nightly batches, and calculated insights can be evaluated on the stream. Practically: a cart abandonment in Manila can trigger inside the session rather than the next morning. The trade-off nobody mentions on stage — streaming consumption burns credits faster than batch, so you have to be deliberate about which events are genuinely real-time worthy. Most aren't.

Zero-copy federation and data sharing

Bi-directional sharing with Snowflake and Databricks means your data science team keeps modelling where they already work. If your CLV model lives in Databricks, you surface the score into Data 360 as an attribute rather than rebuilding the model in Einstein.

Identity resolution rulesets

Match rules and reconciliation rules are configurable per ruleset, and you can run more than one — for example a strict ruleset for marketing consent and a looser one for analytics. This is the single most important feature for Southeast Asia, and I'll come back to why it strains.

Agentforce and Einstein on top of unified profiles

The 2026 positioning is agentic: agents that read the unified profile, retrieve grounded context, and act. Salesforce reported Data Cloud and AI annual recurring revenue passing $1.2 billion, up roughly 120% year over year, in its Q2 FY2026 results (Salesforce investor relations) — evidence that the data-plus-agent bundle is where budget is actually landing, not just where marketing is pointed.

Data Graphs and vector search

Data Graphs pre-materialise related profile data for low-latency retrieval, and unstructured data support (documents, transcripts, reviews) via vector indexing lets agents ground answers in content, not just records. For a group running Cantonese, Bahasa and Vietnamese service transcripts, this is where the multilingual customer-service backlog finally becomes queryable.

A compact example of what ingestion actually looks like — this is the shape of the work, not a keynote slide:

1# Ingestion API: push POS events into a Data Cloud stream
2curl -X POST \
3  "https://${MY_DOMAIN}/api/v1/ingest/sources/POS_Retail_APAC/transactions" \
4  -H "Authorization: Bearer ${ACCESS_TOKEN}" \
5  -H "Content-Type: application/json" \
6  -d '{
7    "data": [{
8      "transaction_id": "HK-0042-9917",
9      "member_id": "MBR-HK-118204",
10      "store_code": "HK-CWB-02",
11      "market": "HK",
12      "currency": "HKD",
13      "net_amount": 1840.00,
14      "category": "fragrance",
15      "occurred_at": "2026-01-14T11:22:05+08:00"
16    }]
17  }'

Then the segment logic runs as SQL-like query against unified profiles:

1SELECT ssot.Id__c, ssot.PartyIdentification__c
2FROM UnifiedIndividual__dlm ssot
3JOIN CLV_Insight__dlm clv ON clv.UnifiedId__c = ssot.Id__c
4WHERE clv.LifetimeValue__c > 8000
5  AND clv.DaysSinceLastPurchase__c > 120
6  AND ssot.MarketConsent__c = 'WHATSAPP_OPT_IN'
7  AND ssot.HomeMarket__c IN ('HK','SG','MY');

Note the consent field sitting in the segment definition itself. In APAC that isn't optional hygiene — it's the control that keeps a Malaysian opt-in from being spent on a cross-market blast.

Is Genie just the legacy CDP with better branding?

No, and the difference shows up in three measurable places.

Latency. Legacy CDP operated on batch ingestion and scheduled segment refreshes, typically measured in hours. Data 360 supports streaming ingestion with real-time segment evaluation. For a flash-sale market like Singapore or a live-commerce push on Shopee, hours versus minutes is the entire campaign.

Scope of activation. Legacy CDP lived largely inside Marketing Cloud. Data 360 profiles are available to Sales Cloud, Service Cloud, Commerce, Tableau, Flow and Agentforce. That matters because the store associate and the call-centre agent were always the weakest link in APAC personalisation — they had the customer in front of them and the worst data.

Data gravity. Copy-everything architectures forced a choice between completeness and cost. Zero-copy lets you keep the 400-million-row transaction table in Snowflake and federate it. Gartner's ongoing coverage of the CDP category has consistently flagged warehouse-native architecture as the direction of travel (Gartner research on customer data platforms).

Where I'd push back on the marketing: the agentic layer is only as good as the underlying data model. In a retail group with three POS systems and a loyalty platform bought in 2014, the constraint is never the AI. It's that "member_id" means something different in Taiwan than it does in Vietnam.

Identity resolution in Southeast Asia is where the model strains

This is the part APAC buyers should test hardest in a proof of concept, because the default identity assumptions are North American.

Email is a weak identifier across much of Southeast Asia. Mobile number is the primary key in practice — but numbers get recycled, prepaid churn is high, and a single household frequently shares one number for loyalty enrolment. Meanwhile the actual conversation happens on channels that vary by border: LINE dominates in Taiwan, Thailand and Japan; WhatsApp in Hong Kong, Singapore, Malaysia and Indonesia; Zalo in Vietnam; WeChat for mainland-visiting shoppers. Each channel hands you a different identifier with different persistence.

Four things I'd insist on in the build:

• Separate match rules by market. A phone-plus-name fuzzy match that works in Australia over-merges in Vietnam, where name distributions are far narrower. Run market-scoped rulesets rather than one global ruleset.
• Normalise phone to E.164 before ingestion, not inside Data Cloud. Country-code inconsistency is the single largest source of false negatives I've seen in regional profile unification.
• Treat name transliteration as a data problem, not a matching problem. Chinese, Thai and Vietnamese name handling in Latin script produces duplicates that no reconciliation rule will fix downstream.
• Keep consent at market level and channel level. Not profile level. This is the difference between a survivable regulator enquiry and an unsurvivable one.

On a Greater China multi-brand retail engagement, the pattern we hit repeatedly was that the CDP itself was configured correctly and the upstream loyalty enrolment form was the problem — optional email, no country code validation, free-text name. The mechanism to fix is boring: tighten capture at the point of enrolment, then backfill. There's no AI shortcut past a bad form.

Is Salesforce Data Cloud free, and what does it actually cost?

A limited entitlement of Data Cloud is included with several Salesforce enterprise editions — Salesforce has publicised a starter allocation of credits and profiles bundled with qualifying Enterprise and Unlimited licences (Salesforce Data Cloud pricing page). Treat that as a sandbox, not a production budget.

Real consumption is credit-based across distinct meters: ingestion, profile unification, segmentation, activation, calculated insights, and increasingly agent/inference actions. Three things drive APAC bills higher than the initial estimate:

1. Segment refresh frequency. Hourly refresh on 40 segments across six markets is a different order of magnitude from daily refresh on eight.
2. Number of markets, not number of customers. Market-scoped rulesets and market-scoped segments multiply object counts.
3. Real-time streams left on for events nobody activates. Audit quarterly. I've yet to see a first-year deployment where every configured stream earned its keep.

Model consumption before signing, and negotiate credits against a forecast you've actually built — not the vendor's. In competitive sport you don't sign up for a distance you've never run; the same discipline applies to consumption contracts.

Governance and residency across a fragmented privacy map

APAC has no GDPR equivalent — it has seven regimes moving at different speeds.

Hong Kong's Personal Data (Privacy) Ordinance is enforced by the PCPD and notably still lacks a commenced cross-border transfer restriction, though the PCPD publishes recommended model clauses (PCPD guidance). Singapore's PDPA, administered by the PDPC, imposes transfer conditions and a mandatory breach notification regime. Indonesia's PDP Law, Vietnam's Decree 13 and Personal Data Protection Law, and Australia's Privacy Act reforms all impose their own consent and localisation expectations.

What this means operationally for Data 360:

• Confirm which Hyperforce region your org and Data Cloud instance run in, and whether that satisfies your most restrictive market. Availability varies — verify against current Salesforce documentation rather than assuming Sydney, Tokyo or Singapore is available for your edition.
• Use consent objects and data spaces to partition markets. Data spaces are the cleanest mechanism for keeping Vietnamese and Australian data logically separated inside one org.
• Document purpose limitation per segment. The regulator's question is never "do you have a CDP" — it's "on what basis did you use this data for this message."

A build order that survives contact with reality

For a regional retail group, sequence beats ambition. The order I'd run:

1. Pick two markets, not six. Ideally one mature (Singapore or Hong Kong) and one messy (Indonesia or Vietnam). If the model survives both, it generalises.
2. Fix capture before unification. Enrolment forms, POS member lookup, country-code validation.
3. Stand up identity resolution with market-scoped rulesets and measure match rate and false-merge rate per market. These are your two scoreboard numbers for the quarter.
4. Activate to one channel end-to-end before adding more. WhatsApp or LINE, with consent enforced in the segment query.
5. Only then layer agents. Grounded service agents and next-best-action are multipliers on a clean profile and amplifiers on a dirty one.
6. Audit consumption at 90 days. Turn off what nobody used.

Staffing is the quiet constraint. Data Cloud work needs a data engineer who understands the Salesforce metadata model — a genuinely scarce profile across APAC, and one where distributed hiring across Manila, Ho Chi Minh City and Kuala Lumpur is often the only realistic path to a team that can actually maintain the thing after go-live.

Where this is heading

The Salesforce CDP Genie features 2026 story is really the story of the CDP dissolving into infrastructure. The segment builder stops being the product and becomes a query surface; the profile becomes retrieval context for agents; the warehouse becomes the system of record and Salesforce becomes the activation and reasoning layer on top. Expect the branding to keep moving — Genie to Data Cloud to Data 360 in three years suggests a fourth name is plausible — while the underlying direction stays fixed: zero-copy, real-time, agent-consumed.

For APAC retail operators, the competitive gap in 2026 and 2027 won't be who licensed which platform. Everyone in the mall will have one. It will be who solved phone-number identity across six markets, who kept consent enforceable at channel level, and who built a team that can maintain the pipeline after the implementation partner goes home. Those are operational problems with operational answers, and they reward the same thing they always have: disciplined repetition, measured weekly.

If you're scoping a multi-market Data 360 build and need a team that understands both the Salesforce data model and the identity realities of Southeast Asia, talk to Branch8 about your CDP roadmap.

Sources

• Salesforce — Data Cloud product overview
• Salesforce Newsroom — announcements and product releases
• Salesforce Developers — Data Cloud documentation
• Salesforce Investor Relations — quarterly results
• Gartner — customer data platform research
• PCPD Hong Kong — Personal Data (Privacy) Ordinance guidance
• PDPC Singapore — Personal Data Protection Act
• Snowflake — secure data sharing documentation
• Salesforce Ben — Salesforce Data Cloud coverage