Salesforce Certified Data 360 Consultant Study Guide

Domain 1: Solution Positioning

Key concepts you must know · 93 practice questions

The differentiator against a data warehouse is identity resolution plus activation inside the Salesforce platform, not storage or query performance.
Consumption is metered in credits against ingestion, unification, insight refreshes, segmentation and activation, so an inefficient design costs money continuously.
A data space is a logical partition for a brand, region or business unit, and identity resolution runs within one, so profiles do not merge across spaces.
Data 360 gives a unified view for activation and analysis; the source systems remain authoritative for their own records and nothing is written back to them.
Its contribution to AI is grounding: unified profiles supply accurate context and retrievers supply relevant content, neither of which trains the model.
A first use case succeeds when the data is already reachable and a named business metric is expected to move; breadth and source count are the wrong criteria.
Unification is pseudonymisation-adjacent territory: a deterministic hash still identifies a person consistently, so it remains personal data.
Data minimisation means ingesting the fields the use cases need, which reduces cost, regulatory exposure and the surface an erasure request must reach.

Domain 2: Data 360 Setup and Administration

Key concepts you must know · 85 practice questions

Permission set licences entitle a user to capability and permission sets grant the specific access; both are needed, which explains the half-configured user.
A storage credential wraps the cloud identity and an external location binds it to a path, replacing mounts that gave everyone on a cluster the same access.
Data Explorer browses records in an object, Profile Explorer opens one unified individual and its links; knowing which to open shortens most investigations.
A stream's run history names the failure directly, so diagnosis starts at the failing component rather than reasoning backwards from a downstream symptom.
Configuration is promoted through supported metadata or a data kit, and data never promotes, since production ingests from production sources.
A sandbox refresh copies production downward, so anything built in the sandbox and not promoted is lost.
An empty segment almost always means the criteria reference an attribute that is unpopulated or holds values in a different form than expected.
Alert on two things: a stream that failed or did not run, and consumption rising against the recent norm. Both fail silently otherwise.

Domain 3: Data Source Connection and Ingestion

Key concepts you must know · 119 practice questions

Refresh behaviour follows what the source sends: upsert for a change feed, append for immutable events, full refresh only when the source sends everything.
An incremental feed cannot convey a deletion, since a removed record is indistinguishable from an unchanged one; the source must signal it explicitly.
Identifiers with leading zeros must be ingested as text, or the values silently stop matching the source and every join and match rule degrades.
Zero-Copy federation queries the remote platform in place, so you inherit its performance, availability and query cost, and some operations still materialise.
Derive once on the way in: a formula for row-level values, a batch transform for anything spanning records or objects.
A transform writes to a new output object rather than modifying its input, which is what lets a corrected transform simply be re-run.
A missing file is a silent success, so monitor for the absence of expected data rather than only for failures.
Engagement data is usually ingested rather than federated: high volume, high frequency and needed at low latency.

Domain 4: Harmonization and Unification

Key concepts you must know · 113 practice questions

A ruleset runs match rules first to decide which records are one person, then reconciliation rules to choose each attribute's winning value.
Rules within a ruleset are alternatives, so any one linking is enough; attributes within a rule are requirements, so all must agree.
A shared placeholder value in a match attribute is the classic mass over-merge, which is why known junk values should be normalised to null at ingestion.
Unification links source records and leaves them intact, so lineage survives and a ruleset change can be re-run against the same data.
A unified profile's identifier can move when records regroup, so it is unsuitable as a durable key in an external system.
Grain is the promise an object makes about what one row represents, and a measure at a coarser grain repeats across rows and double counts.
Harmonization aligns meaning rather than names: two fields sharing a label and describing different concepts belong in separate model fields.
Match quality drifts because the data changes, so review it periodically and whenever a source is added, not only when somebody complains.

Domain 5: Data Enhancements, Sharing, and Analysis

Key concepts you must know · 119 practice questions

A calculated insight computes a per-profile metric on a schedule; a streaming insight evaluates a bounded window as events arrive, for in-session reaction.
An insight's dimensions set the grouping, so adding one produces rows per profile per dimension rather than a single profile attribute.
Insight cost is volume multiplied by frequency, which is why a metric over five years recomputed hourly is the most expensive shape available.
A data graph pre-joins related data for one profile so retrieval is a single low-latency call, which is what real-time consumers and agents need.
Define a metric once as an insight rather than letting each consumer aggregate, or the figures diverge and reconciliation costs more than governance would.
A feature must be available at the moment of prediction; one that encodes the outcome makes a model look excellent in evaluation and useless in production.
Grounding inserts profile data and retrieved content into a prompt for one request; the Trust Layer masks, enforces zero retention and logs the exchange.
An insight with no qualifying records usually has no value rather than zero, and what that absence means for consumers is a decision to make and document.

Domain 6: Data Activations and Utilization

Key concepts you must know · 131 practice questions

A segment is criteria plus a membership resolved when it runs, so a customer qualifying since the last run is not in it.
Mixing alternatives with a required condition needs the grouping made explicit, which is the commonest segmentation error and produces a plausible-looking count.
An activation is an audience published to a target on a schedule; a data action fires an event about one individual when a condition is met.
A condition expressing a continuing state rather than a transition fires repeatedly for the same customer, which floods the receiver.
Consent is not enforced implicitly: it must be a condition where the audience is defined or published, filtering on channel and purpose.
Each attribute an activation publishes is personal data disclosed to that destination, under that destination's controls and retention.
Mutual exclusivity between concurrent campaigns is built by ranking them and excluding higher-priority audiences from each lower one.
Event delivery can be lost or repeated, so a flow triggered by a data action should be idempotent by design.

Salesforce Certified Data 360 Consultant exam tips

Almost every question is a scenario with several workable answers, so read for the constraint. A stated latency requirement, a rule that consumers must not be disrupted, or a note that the data is already reachable is usually what eliminates three options.
When a question asks where something belongs, work out which stage of the pipeline it sits in. Formula and transform act on source-shaped data, insights act on unified profiles, and segments and activations act on the result. Most wrong answers put the work in the wrong stage.
Match the mechanism's latency to the decay of the opportunity. An abandoned cart needs a streaming insight and a data action; a next-day survey needs a scheduled segment. Choosing the more immediate mechanism where it is not needed is as wrong as choosing the slower one.
Remember which things the platform does not do: it does not write back to sources, does not resolve duplicates in non-party objects, does not enforce consent implicitly, and does not deduplicate across independent activations. Several questions turn on exactly these boundaries.
For cost questions, the answer is almost always volume multiplied by frequency. Refresh cadences and field selections inherited as defaults are where the waste sits, and reducing them rarely removes any capability.
Distinguish a wrong-value complaint from a wrong-person complaint. The first points at reconciliation rules, the second at match rules, and each has its own investigation. Confusing them is a fast way to lose a scenario question.

Study guide FAQ

What format and passing requirements does the Salesforce Certified Data 360 Consultant exam have?

The exam is 60 scored multiple-choice questions plus up to 5 unscored, with 105 minutes to complete it and a passing score of 70%. Registration is 200 USD plus applicable taxes, and there are no formal prerequisites.

Is Data 360 the same thing as Data Cloud?

Yes. The product shipped as Salesforce CDP, became Data Cloud, and is now Data 360. Nothing about the architecture changed with the name, so existing configuration and skills carry across, and the credential itself is now titled Data 360 Consultant. Expect to meet both names in documentation and in the same conversation.

Which sections carry the most weight?

Data Activations and Utilization is the largest at 20 percent, followed by Data Source Connection and Ingestion and Data Enhancements, Sharing, and Analysis at 18 percent each, then Harmonization and Unification at 17 percent. Solution Positioning is 14 percent and Setup and Administration 13 percent, so the exam is spread fairly evenly with a deliberate tilt toward acting on the data.

How much hands-on configuration knowledge does the exam expect?

Enough to reason about a configuration rather than to recall a screen. You should know what a data stream, data lake object, data model object, identity resolution ruleset, calculated insight, data graph, segment and activation each do, where each sits in the pipeline, and what happens when one is changed. It does not ask you to recite menu paths.

How much privacy and governance content is there?

More than candidates expect, and it is spread across every section rather than confined to one. Consent enforcement in segmentation and activation, what an erasure request has to reach, whether an attribute should be published to a destination at all, and who may create activations all appear as scenario questions. Treat governance as part of the design rather than a separate topic.

Related Salesforce resources

Test yourself: 660 Salesforce Certified Data 360 Consultant practice questions