A Map Is Not a Decision

🌐 Also in 28 languages

Español · 日本語 · Português (Brasil) · Français · Deutsch · Italiano · 한국어 · 简体中文 · Português (Portugal) · 繁體中文 · Bahasa Indonesia · Türkçe · Polski · Tiếng Việt · ไทย · Nederlands · العربية · עברית · Українська · हिन्दी · Русский · Svenska · Dansk · Norsk bokmål · Suomi · Español (México) · Français (Canada) · Deutsch (Schweiz)

Image generated with AI

Module 4 of 6 in the chapter Master Data and Identity, chapter 6 of the Data Architect course. Each module stands on its own. Previous: Matching, Merging and Survivorship · Next: When Authority Is Shared

The question this module answers: Is a unified profile the same thing as a golden record?

The same customer sits in three systems. Salesforce knows Daniel Reyes, d.reyes@example.com, Dallas. E-commerce knows Dan Reyes, the same email, Dallas, TX. Support knows Daniel Reyes, 214-555-0100, Dallas. Data 360 decides all three are the same person and builds a unified profile. Excellent. Do we have a golden record? Not necessarily. We know the records belong together; we haven’t decided which source owns the truth. This can be confusing for people: identity resolution creates a unified individual, but not a golden record. So I teach the difference head-on:

A unified profile is a map. A golden record is a decision.

Identity Resolution Answers “Who?”

Without identity resolution, one person shows up as disconnected profiles: marketing sees one, sales another, service another. Data 360 ingests them and uses identity resolution rules to decide which belong together. Then we can say these records appear to be the same person. That solves identity. It doesn’t solve mastering.

And the unified profile keeps the map back to its sources. In Salesforce’s own description, a unified profile isn’t a golden record: it’s a set of keys, with identity link objects tying each source record’s ID to the unified profile’s ID (Data 360 Architecture (opens in a new tab)). Unified person 10042 maps to Salesforce contact 003…, e-commerce customer EC-48372 and support customer SUP-9128. The originals don’t disappear. That’s very close to module 2’s registry.

Click to enlarge

A unified profile is a map back to the source identities.

Match Rules: Strictness Is a Risk Decision

How does Data 360 decide who gets unified? Match rules. Profiles match when all the criteria in a rule are met, and if you have several rules, a match on any one creates the unified profile. Salesforce’s own guidance: use more criteria for accuracy (more distinct profiles), fewer for consolidation (Identity Resolution Match Rules (opens in a new tab)). So adding criteria to a rule makes it stricter; adding alternative rules widens the ways profiles can match. Configuring identity resolution isn’t “turn on matching.” It’s defining what evidence is enough to say these profiles are one person.

More matching isn’t automatically better. Ten million source profiles become eight million unified ones. Loosen the rules, and now six million. Better? Maybe you found more duplicates. Or maybe you started combining different people. A smaller count isn’t proof of better data. The goal isn’t the fewest profiles; it’s the most defensible identities for the purpose. And each criterion has a match method: Exact, Exact Normalized, or Fuzzy at high, medium or low precision. Salesforce’s own warning is the point: a less precise method raises the consolidation rate, “but can result in undesirable matches.” (Fuzzy isn’t available for Account fields, and the normalized values are used only to compare; they aren’t stored in the unified profile, which is module 3’s harmonizing step at work: Identity Resolution Match Methods (opens in a new tab).) The looser the method, the more the identity itself becomes a judgment. If it’s wrong, do we merely personalize a page badly, or show one person’s support history to another, or let an agent act on two people’s data?

That’s my position: match strictness is a risk decision. It depends on how uniquely the matching fields identify a person and what a false match costs. Where blending is costly, under-match on purpose. Two profiles for one person is inconvenient. One profile holding two people can be dangerous.

Reconciliation Still Isn’t Mastering

Once profiles are unified, the sources still disagree: Salesforce says Dallas, e-commerce Plano, support Dallas. What does the unified profile show? Reconciliation rules decide: Last Updated, Most Frequent or Source Priority, set per object and overridable per field (Identity Resolution Reconciliation Rules (opens in a new tab)). That should sound like module 3’s survivorship.

Here’s where teams get into trouble. The unified profile shows Dallas, and someone says, “Great, Dallas is our golden value.” Slow down. E-commerce still says Plano. Salesforce is explicit that reconciliation doesn’t delete or change source data or update the source systems, and it isn’t meant to pick a “best” value for every application. It decides only what the unified profile shows (same source). Presentation isn’t authority.

Test knowledge. Know how Data 360 identity resolution decides and displays: in a match rule, all criteria must match, and any one rule matching unifies the profiles (more criteria for accuracy, fewer for consolidation); the three reconciliation rules are Last Updated, Most Frequent and Source Priority, set per object and overridable per field; and reconciliation doesn’t change the source data, and contact points aren’t reconciled.

Contact points make the point even clearer. A person can have a personal and a work email, a mobile and a home phone, all legitimate. Data 360 doesn’t reconcile contact points at all: every email and phone stays on the unified profile (same source). Sometimes several values is simply the truth. “Unified” doesn’t mean one value for everything.

A Golden Record Is a Decision

Compare a real mastering requirement. The business says the ERP is authoritative for the legal name, and the mastered value is “Oakline Logistics LLC.” Salesforce, billing and service must all use it, and if one disagrees, the mastering process pushes the governed value back. That’s a different architecture. We’re not choosing what one view displays; we’re deciding this is the value the enterprise uses, and every participating system converges on it.

So: a unified profile brings identities and information together. A golden record is a governed decision about which values are authoritative. A unified profile can make an excellent single view of the customer, assembled from authoritative sources. That doesn’t make it the system that masters the customer.

Click to enlarge

A unified profile is a map. A golden record is a decision.

Traceability

If the unified profile says Dallas, where did that come from: Salesforce, the ERP, e-commerce? Was it chosen as last updated, most frequent, source priority? That provenance matters for audit, privacy, troubleshooting, fixing data quality and agent decisions, and above all for knowing where a correction belongs. If Dallas is wrong, editing the assembled view doesn’t fix the source. You need the external IDs, the identity links and the lineage fields to trace back to the record that should change.

Three Tools for Three Jobs

This is how I separate them:

1. Duplicate management and merge, within one Salesforce org, when there’s a clear winner and you no longer need the duplicates (module 3).
2. Data 360 identity resolution, for one identity and a single view across sources without changing them: a registry.
3. A dedicated MDM platform, when values must be decided once and pushed back to every participating system: coexistence or centralized.

Click to enlarge

Three tools for three jobs: merge, link, master.

Don’t buy MDM when you need a map. “We want the service agent to see that this Salesforce customer is the same person who bought online” may need identity resolution and combined context, not centralized MDM. “The ERP, Salesforce, e-commerce and billing must all use the same governed legal name, and changes must flow back to all of them” is different: that’s mastering. Don’t solve the second when you have the first, and don’t pretend the first solved the second.

One Identity Isn’t One Permission

Identity resolution says these records are one person. That doesn’t mean everyone who can see one source record may see everything attached to the unified identity. A Retail Partner may see the deliveries it sent; another partner’s deliveries to the same Recipient are none of its business, even if the identity layer knows they’re related. The identity graph can be right while the access design is wrong. Unifying identity doesn’t erase security boundaries. That’s why over-matching becomes a privacy problem, and why the Steve Trucking checkpoint treats unifying Recipients so carefully: a map of who’s who doesn’t grant ownership, or access.

Beyond the exam: what AI changes. Agents need one identity to act on, and the unified profile gives it to them. That raises the stakes both ways. Over-matching is now a privacy incident: an agent acting for one person on another’s data. And an agent that “updates the customer” must write to the attribute’s system of record, using module 1’s authority map, not to the unified profile. The map tells the agent who the person is; the authority model tells it where a change belongs.

What an Architect Decides Here

• What evidence is enough to unify profiles, and how the cost of a false match sets the strictness.
• Which reconciliation rules decide what the unified profile shows.
• Whether the need is duplicate cleanup in one org, identity across sources, or true enterprise mastering.
• Which source stays authoritative for each attribute, even when Data 360 shows a reconciled value.
• How unified identity interacts with security, privacy, provenance and agent actions.

Data 360 can tell us these profiles belong together, give us a unified identity, reconcile what it shows and keep the links back to every source. That’s enormously valuable. But if Salesforce says one thing and the ERP another, someone still has to decide which system may change the enterprise truth. So when someone says “we have a unified profile, so we have our golden record,” ask: did we unify the data, or did we master it?

Explore

• Salesforce Architects, Data 360 Architecture (opens in a new tab), checked 2026-10-14.
• Salesforce Help, Data 360: Identity Resolution Match Rules (opens in a new tab) and Identity Resolution Match Methods (opens in a new tab) and Identity Resolution Reconciliation Rules (opens in a new tab), checked 2026-10-14.