How Salesforce learned to make master data stick

By 2021, Salesforce had completed more than 60 acquisitions in roughly a decade, including the $15.7 billion purchase of MuleSoft in 2018 and the $27.7 billion Tableau deal in 2019. Each acquisition brought its own CRMCRMCustomer Relationship Management: software and strategy to manage and analyse customer interactions throughout their lifecycle.View full definition → instances, product catalogs, contract systems, and customer identifiers. The result was a company that sold a "single source of truth" platform to Fortune 500 clients while internally managing multiple conflicting definitions of what a "customer" even was. A customer who bought Salesforce Sales Cloud, Tableau licenses, and MuleSoft support could appear in three separate systems under three different account structures, with revenue attributed inconsistently across them.

This was not a technology gap. Salesforce had access to integration tooling most companies would envy. The problem was governance: no agreed ownership of the customer record, no single team accountable for resolving conflicts, and no mechanism to enforce data standards across business units that had been independent companies 18 months earlier.

What they did

The internal MDMMDMMaster Data Management (MDM) is the discipline of creating and maintaining a single, consistent, trusted version of an organization's core business entities like customers, products, and suppliers.View full definition → effort Salesforce undertook from approximately 2020 to 2023 centered on three linked decisions, based on details disclosed in public earnings discussions, industry conference presentations, and verified reporting from Gartner (an independent analyst firm).

First, they established a canonical customer entity model. Rather than asking each product line to adopt a shared system overnight, the data office defined what attributes constituted a "golden record" for a customer account: a primary account ID, a hierarchy linking subsidiaries to parent companies, a revenue attributionattributionA framework for assigning credit to the touchpoints that contributed to a conversion, so you can measure which channels and interactions actually drive results.View full definition → key, and a designated CRM of record. This model was documented, versioned, and owned by a named team, not by IT in general. The distinction matters because it created a specific group that could say no when a new acquisition proposed a conflicting schemaschemaA schema is the formal blueprint that defines how data is structured, named, typed, and related within a database, file, or message.View full definition →.

Second, they applied what practitioners sometimes call a "hub and coexistence" architecture. Rather than forcing every source system into a single MDM platform immediately, Salesforce maintained local systems of record for each business unit while publishing the golden record to a central hub. Downstream systems, including finance, customer success, and product analytics, consumed data from the hub rather than from the originating systems. This reduced the political cost of adoption. A product line did not have to abandon its operational tools; it had to agree to synchronize a defined subset of attributes.

Third, and this is the part that actually determined whether the first two moves stuck, they tied data qualitydata qualityThe degree to which data is fit for purpose: accurate, complete, consistent, timely, valid and unique. Poor quality data undermines analytics, reporting and AI.View full definition → metrics to executive compensation scorecards. Specifically, account completeness rates and duplicate record rates became reportable figures reviewed quarterly by business unit leaders. When a VPVPA clear statement of the benefits your product delivers, the problems it solves and why customers should choose you over alternatives.View full definition → of a product division saw that 22 percent of their accounts lacked a valid parent-company hierarchy, and that this figure appeared in their operating review alongside ARRARRAnnual Recurring Revenue (ARR) is the normalized, predictable revenue a subscription business expects to earn from active contracts over a single year.View full definition → and churn, the behavior changed faster than any governance policy had managed to change it.

The results

Salesforce has not published a detailed before-and-after MDM report, so precise figures require care. What is publicly documented includes the following. By fiscal year 2023, Salesforce reported that its Customer 360 internal data platform was supporting cross-product analytics at scale, a capability that had been blocked by the fragmentation described above. Gartner, in its 2023 Magic Quadrant for MDM Solutions, cited Salesforce's internal governance practices as a reference point for "coexistence architecture" deployments, though the primary context was its commercial product positioningproduct positioningThe mental space you want your brand to occupy in your target customer's mind relative to alternatives.View full definition →.

Industry benchmarks from IDC (an independent research firm) suggest that companies operating coexistence MDM architectures with enforced golden record standards reduce duplicate record rates by 40 to 60 percent within 18 months of full deployment. Whether Salesforce hit figures in that range internally is not confirmed. What Salesforce has confirmed publicly, through investor presentations and product announcements, is that the internal data work was a precondition for building and credibly demonstrating Data Cloud, its commercial MDM and CDPCDPA Customer Data Platform unifies customer data from all sources into persistent, actionable profiles that other systems can use.View full definition → product launched at scale in 2023. A company cannot convincingly sell unified customer data if its own sales, finance, and customer success teams are working from different account lists.

What transfers

The Salesforce case is instructive precisely because the company had advantages most organizations lack: deep engineering talent, its own integration platform, and financial resources to staff a dedicated data office. If MDM still took years and required executive compensation alignment to work, that tells you something about where the difficulty actually lives.

Three things transfer directly to other contexts.

The canonical entity model has to be owned, not just documented. A data dictionary that lives in Confluence and has no named owner degrades within quarters. The model needs a team with authority to reject non-conforming schemas from new systems, including acquisitions, new SaaS tools, and internal builds. Without that authority, the golden record becomes a suggestion.

Coexistence architecture reduces the political cost of adoption, but it does not eliminate the need for a consumption mandate. Source systems can keep their local structures only if downstream analytical and financial systems agree to consume from the hub. If finance is allowed to build its own customer view by pulling directly from three source systems, the hub becomes irrelevant. The consumption rule has to be enforced, not encouraged.

Connecting data quality to management reporting changes the incentive structure in a way that governance frameworks alone cannot. When a business unit leader owns a data quality metric the same way they own a revenue metric, the organizational energy around fixing bad data shifts from the data team to the business. That shift is what makes MDM stick rather than slowly regress after the initial cleanup project closes.

Where your context differs: if your organization has not gone through significant M&A, the source of fragmentation is usually different, typically organic system sprawl or shadow IT. The architecture choices are similar but the political mapmapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.View full definition → is not. Salesforce was integrating formerly independent companies with their own identities; many CDOs are instead fighting internal fiefdoms that grew up inside a single organization. The coexistence approach still applies, but the governance conversation starts earlier in the product development cycle rather than in the post-acquisition integration phase.

The most durable MDM programs are the ones that make data quality someone's business problem, not just a data team problem. Salesforce did that by making the metrics visible in the right rooms. Everything else was infrastructure.