Destination catalog, field mapping, cadence, suppression
Destination activation matrix
Governance, Labels & DUP
Data usage labels, governance policy enforcement, access control
Governance policy set, audit log config
Launch & Monitor
Go live, monitor audience/activation/destination health
Activation dashboard, incident runbook
Step-by-step implementation
Do these 11 steps in order, in a development environment. Each step lists where to go, what to do, what to check, what you must know, and the questions interviewers ask about it.
Step 1
Confirm the data foundation
Where
Data management > Datasets; Customer > Profiles
What you do
Verify every dataset you need is Profile-enabled and its identities were validated in the AEP build.
Tips
A dataset without the Profile toggle is invisible to audience building.
Check before moving on
Data readiness checklist signed with the AEP lead.
Must know
RTCDP sits on top of the AEP foundation. If schemas, identities and Profile-enabled datasets are not validated, every audience built on them is wrong.
Confirm the edition (B2C, B2B, B2P) and package before design, because it decides which features and destinations you have.
Check profile counts and identity coverage before building any audience.
Interview questions tap to reveal a model answer
What has to be in place before you build audiences in RTCDP?
Validated schemas and identities, Profile-enabled datasets with data flowing, merge policies agreed, and consent attributes populated.
How does RTCDP relate to AEP?
RTCDP is an application built on AEP: it uses AEP's profiles and identity graph to build audiences and activate them to destinations.
What would you check if audience sizes look too high?
Over-merged identity graphs, the wrong merge policy, duplicate datasets, or a missing exclusion such as consent.
Screenshot Dataset list showing Profile statusAdd screenshots/rtcdp/step-01.pngStep 2
Agree consent rules
Where
Consents and Preferences field group on the profile schema; legal workshop
What you do
Build a region-by-region consent matrix and confirm which consent field gates each channel.
Tips
Consent must be populated on profiles before any audience is activated.
Check before moving on
Consent values present on test profiles.
Must know
Consent has to be captured as data (for example the Consents and Preferences field group) before it can be enforced.
Consent rules differ by region (opt-in in the EU under GDPR, opt-out in many US states), so agree them with legal per channel.
Consent policies can block activation automatically, but availability depends on your licence; check your package. Where they are not available, exclude non-consented profiles in every activated audience.
Interview questions tap to reveal a model answer
How do you enforce marketing consent in RTCDP?
Capture consent in the profile, then enforce it with consent policies where licensed, or exclude non-consented profiles in every audience used for activation.
Opt-in vs. opt-out regions: how does design change?
Opt-in regions need explicit consent before marketing; opt-out regions allow it until the person objects. Store region and consent status so audiences can apply the right rule.
Where is consent stored in XDM?
In consent and preference fields on the profile (for example the Consents and Preferences field group), updated from the CMP or preference centre.
Choose identity stitching and attribute merge (timestamp ordered or dataset precedence); set the default policy.
Tips
Use dataset precedence when CRM values must win over web values.
Check before moving on
5–10 profiles with conflicting values resolve as expected.
Must know
A merge policy decides which profile fragments join (ID stitching: private graph or none) and which value wins when they conflict (timestamp ordered or dataset precedence).
Each schema class can have one default merge policy per sandbox; audiences use the default unless you pick another.
Changing the default merge policy does not move existing audiences to it; they keep the policy they were built with.
Interview questions tap to reveal a model answer
What is a merge policy?
Rules for assembling a profile: whether to stitch identities via the graph, and how to resolve conflicting attribute values from different datasets.
Timestamp ordered vs. dataset precedence?
Timestamp ordered keeps the most recently updated value; dataset precedence trusts datasets in the order you rank them (e.g. CRM over web forms).
When would you use "None" for ID stitching?
When each identity must stay separate, for example in regulated cases where profiles must not be joined through the graph.
Drag attributes and events into Audience Builder (formerly Segment Builder), add time constraints, choose the merge policy, save.
Tips
Name audiences with the use case and business owner.
Check before moving on
Estimated size is in the expected range.
Must know
Audiences are built from profile attributes, events, other audiences and time constraints in Audience Builder (the new, generally available builder; the old Segment Builder is due to be deprecated in early 2027).
Always tag audiences with owner and use case, and keep naming consistent; audience sprawl is a common governance problem.
Preview estimated size and sanity-check it against Analytics or CJA numbers.
Interview questions tap to reveal a model answer
How would you build a cart abandonment audience?
People with an add-to-cart event in the last hour and no purchase event after it, excluding anyone without marketing consent.
Attributes vs. events in audience rules?
Attributes describe who someone is (tier, country); events describe what they did (searched, purchased), usually with a time window.
How do you validate an audience?
Check the estimated size against benchmarks, inspect sample profiles that qualify, and test with a small export.
Screenshot Audience Builder with the rule and size estimateAdd screenshots/rtcdp/step-04.pngStep 5
Use streaming or edge for real-time triggers and batch for daily lists.
Tips
Some rule types cannot be evaluated by streaming — check eligibility.
Check before moving on
A synthetic event qualifies a test profile within the SLA.
Must know
Batch evaluation runs every 24 hours (or on demand); streaming evaluates in near real time as events arrive; edge evaluates instantly on the Edge Network for same- or next-page personalization.
Not every rule can be evaluated by streaming or edge; complex or long look-back rules may only run as batch.
Pick the method from the use case's latency need, not by default.
Interview questions tap to reveal a model answer
Batch vs. streaming vs. edge segmentation?
Batch: scheduled, all rule types. Streaming: near real time on incoming events, with rule limits. Edge: instant decisions at the edge for web personalization, with stricter limits.
Which method for a flight-delay alert?
Streaming (or an AJO event-triggered journey), because the message is only useful within minutes.
Why might an audience not be eligible for streaming?
Its rules use unsupported constructs, such as long multi-event look-backs or certain functions, so it falls back to batch.
Data management > Datasets > Data governance tab; Privacy > Policies
What you do
Apply data usage labels to sensitive fields and enable the governance policies that use them.
Tips
Label at schema/field level early — retrofitting later is disruptive.
Check before moving on
Policies show as enabled.
Must know
Data usage labels (contract, identity and sensitive labels) are applied to datasets and fields.
Marketing actions plus labels form policies; Platform enforces them when you activate to a destination.
Label early. Adding labels after audiences and destinations exist is disruptive.
Interview questions tap to reveal a model answer
What is a data usage label?
A tag on a dataset or field that says how the data may be used, for example a sensitive or contractual restriction.
How does policy enforcement work at activation?
When you map an audience to a destination, Platform checks the labels on the data involved against the destination's marketing actions and blocks violations.
How do you test governance?
Deliberately try an activation that should be blocked and confirm the policy stops it.
Screenshot Data governance labels on a datasetAdd screenshots/rtcdp/step-06.pngStep 7
Connect a destination
Where
Connections > Destinations > Catalog > Set up
What you do
Authenticate the destination (ad platform, ESP, S3, Marketo), name it, and choose its marketing actions.
Tips
Marketing actions are what governance policies check at activation.
Check before moving on
Connection shows as active.
Must know
Destinations differ: streaming vs. file-based (batch), audience export vs. profile export, advertising, email, CRM and personalization destinations.
Each destination has its own identity requirements (hashed email, mobile IDs) and rate limits.
Configure destination marketing actions so governance can be enforced.
Interview questions tap to reveal a model answer
Streaming vs. batch destinations?
Streaming sends audience changes continuously (ad platforms, Target); batch exports files on a schedule (cloud storage, email tools).
Why hash emails for ad destinations?
Ad platforms match on hashed identifiers so raw PII is never shared; Platform can hash on export where supported.
What do you check before connecting a destination?
Licence, authentication, identity requirements, required fields, rate limits and the marketing actions to apply.
Screenshot Destination catalog card and auth screenAdd screenshots/rtcdp/step-07.pngStep 8
How to use: do each step in your sandbox using the “How to do it” column, then read across the row to see how the same step changes by domain. Each domain column follows one scenario from top to bottom. Add a screenshot and what you learned in the last column.
Travel & Airlines · practice brand: SkyLumen Airways
Match rate at the ad platform; watch peak rate limits.
How (any domain): Small control audience first; check the destination; then monitor activation health.
Check: Activation dashboard and runbook in place.
Screenshot Test & monitor · E-commerce & RetailAdd screenshots/rtcdp/retail-09.png
Extra steps for this domain
In Phase 4, build abandoned-cart, browse-abandon, and post-purchase segments as the first three.
In Phase 5, set higher segment refresh frequency (near-real-time) given short purchase-decision windows.
Load-test destination activation at peak-season volume before Black Friday/Cyber Monday; confirm no rate-limit failures.
Healthcare
Lab use case: Check-up due audience
Profile: B2C (patients) + B2B (providers, payers). Compliance focus: HIPAA, HITRUST, consent-first data collection, no PHI in standard analytics tools without a BAA-covered environment.
What changes in this domain: Segment only on non-PHI derived attributes unless in a compliant sandbox
Special considerations: Extra governance labels required; limit destination list to compliant vendors
“Annual check-up due” from derived, non-PHI fields only; legal approves every attribute.
How (any domain): List audiences per use case (known, anonymous, account); build a region-by-region consent matrix with legal.
Check: Audience brief and consent matrix signed off.
