Unified Governance Across 3,000+ Retail Touchpoints

Situation

L’Occitane operates a luxury retail network of more than 3,000 stores across 90+ countries, with three Salesforce clouds wired in behind it. Sales Cloud held wholesale accounts, Service Cloud handled support across regions, Marketing Cloud powered loyalty and CRM. Each had grown into its own shape over a decade of acquisitions, regional buildouts, and platform decisions that were locally correct at the time.

The visible consequence was three identities for the same person. A customer who bought in-store in Paris, ordered online in Tokyo, and contacted support in New York existed as three unconnected records, with different IDs, partial histories, and no shared context. Marketing campaigns were built on 30 to 40 percent of the available customer data. Store associates couldn’t see online purchase history. Service agents resolved cases without knowing the customer’s product portfolio.

Across 12 APAC markets the picture was tighter and more painful. Regional Salesforce implementations had diverged from the global architecture. Each market had developed its own integration patterns, field customisations, and data model. The result was a multi-cloud estate that was technically connected and operationally fragmented.

Challenge

The clouds talked to each other. The customer record didn’t. Sales Cloud keyed on account ID, Service Cloud on case contact, Marketing Cloud on email. No systematic identity resolution sat between them, so any “single view” was assembled by hand for each report.

Integration was nightly batch and point-to-point. Marketing Cloud synced from Sales Cloud overnight. Service Cloud pushed cases to Sales Cloud via a custom REST API. Commerce sat outside the Salesforce estate entirely. Latency ran 24 to 48 hours, every dependency was a place a future outage could start, and no surface could activate on what a customer did this morning.

The data model carried years of inconsistency on top. Field naming differed across clouds. “Customer” was modelled three different ways. There was no canonical data dictionary, no owner per entity, and no change-control process. For the Agentforce programme already on the executive roadmap, that foundation was insufficient. Intelligent agents need a complete, current, unified customer view. They can’t reason on fragments separated by a 24-hour batch window.

What I told the steering committee in week three

The agents will land. The hard problem isn’t the agent. It’s that the customer record they’re meant to act on doesn’t exist yet.

Action

The architectural call was to make Salesforce Data Cloud the customer data layer for the whole estate, rather than build another integration between clouds. Data Cloud would ingest from Sales, Service, Marketing, and the commerce platform, resolve identities, hold the canonical model, and activate back to every surface from a single source.

With the data foundation in place, activation was rebuilt across three surfaces. Marketing Cloud reconnected to Data Cloud for real-time segmentation; segments shifted from 24-hour-old contact lists to live unified profiles. Experience Cloud powered a new store-associate mobile app, putting a Customer 360 view at the point of sale. Service Cloud gained Data Cloud context cards, surfacing purchase history and product ownership directly inside case management.

The governance layer was what made it stick across markets. A canonical data dictionary defined every shared entity, named owners, set change-control rules. Quality dashboards measured completeness, accuracy, and freshness per cloud. For the 12 APAC markets, a harmonisation process aligned regional customisations to the global model, set up regional data stewards, and installed drift monitoring so divergence would show up in days, not at the next migration.

Result

The estate now operates on one customer record across 3,000+ touchpoints, in-store and digital combined. Identity resolution reduced duplication by 85 percent and produced a single customer graph that every activation works from.

Marketing felt it first. With 2.5x more complete data and near-real-time segmentation, targeted campaigns lifted conversion by 40 percent in the first two quarters. Segment build time dropped from hours to minutes; the team stopped batching weekly waves and started shipping daily.

Store associates adopted the Customer 360 app fast. Associates with full customer history closed 15 percent more incremental sales through relevant recommendations. Service case resolution dropped 25 percent as agents stopped escalating to find context that was now sitting on the case page.

The Agentforce programme that depended on this foundation arrived on schedule. Three pilots launched on the unified profile, with complete history, real-time activation, and the same governance the rest of the estate runs on. The customer record is the constraint Agentforce always exposes, and L’Occitane crossed it before the agents went live, not after.

Reflection

This pattern works when there are three or more clouds sharing a customer entity, when activation latency matters, and when leadership accepts that the foundation has to land before the next layer ships. It is the right call when an Agentforce or Customer 360 programme is already on the roadmap, because the data layer is the binding constraint either way.

It works less well when the org has only one cloud and one market live. The canonical model still helps, but the cost of Data Cloud relative to the benefit at that scale is harder to justify until a second market or a second cloud is imminent.

Worth doing earlier: the governance layer. Markets diverge fastest in the first 18 months after a regional rollout. Canonical dictionary, named owners, and quality monitoring at month three save the harmonisation engagement at month thirty.