Customer Data Platform Management Flashcards

Question 1

What is a customer data platform, and what is its main purpose?

Accepted Answer

A customer data platform is a system that collects customer data from many sources, combines it into a persistent, unified profile for each individual, and makes those profiles available to other systems for marketing, service, and analysis. Unlike a warehouse built mainly for reporting, it centers on individual customers and fast activation.

Question 2

What is a customer profile in the context of a customer data platform?

Accepted Answer

A customer profile is a persistent record that gathers all known attributes, identifiers, consented preferences, and behavioral events linked to one individual or household. It serves as the single reference point other systems query when deciding what to show, send, or recommend. Good profiles also note when each attribute was last updated.

Question 3

Which core components typically make up a customer data platform architecture?

Accepted Answer

A typical architecture includes data ingestion connectors that pull or receive events from websites, apps, and business systems, an identity layer that matches records to people, a profile store that holds unified data, and an activation layer that sends audiences and attributes to destinations. Some platforms also include analytics and a query interface.

Question 4

What is the difference between first-party, second-party, and third-party data?

Accepted Answer

First-party data is collected directly by an organization from its own customers and interactions, such as website activity. Second-party data is another organization's first-party data shared under agreement, and third-party data is aggregated from external providers who did not collect it directly. Platforms prioritize first-party data because the organization controls its collection and consent.

Question 5

What is an event in customer data collection, and why does it matter?

Accepted Answer

An event is a timestamped record of a single action a customer or device takes, such as viewing a product, submitting a form, or completing a purchase. Events usually carry a name, properties describing the action, and an identifier. Because they capture behavior as it happens, events feed profile enrichment, segmentation, and triggered messages.

Question 6

What is a unique identifier, and why do platforms use several of them?

Accepted Answer

A unique identifier is a value that points to one person, device, or account, such as an email address, login ID, or cookie. Platforms use several because no single identifier covers every interaction; a customer may browse anonymously, then sign in on another device. Linking them reveals which interactions belong to the same person.

Question 7

What does it mean for a customer data platform to be persistent?

Accepted Answer

Persistent means the platform stores customer data over time, rather than only holding information for a single session or campaign. Because records remain available, profiles grow richer with each interaction, and historical behavior can inform segmentation, measurement, and personalization long after the original event occurred.

Question 8

How does a customer data platform differ from a customer relationship management system?

Accepted Answer

A CRM mainly records the organization's interactions with known contacts, such as sales activity, support tickets, and deal stages, usually entered by staff. A customer data platform ingests data from many systems and sources, including digital behavior, and builds unified profiles that other tools can use. The two often work together.

Question 9

What is identity resolution, and what problem does it solve?

Accepted Answer

Identity resolution is the process of determining which records across sources refer to the same person or household and linking them together. It solves the problem of fragmented customer views, where one shopper may appear as several unrelated profiles because each system stores different identifiers.

Question 10

What is the difference between deterministic and probabilistic matching?

Accepted Answer

Deterministic matching links records only when they share an exact identifier, such as the same verified email address. Probabilistic matching estimates whether records belong to the same person by comparing several attributes and assigning a confidence score. The first is precise but narrower, while the second reaches further with more risk of error.

Question 11

What is a tracking pixel or SDK, and how does it collect data?

Accepted Answer

A tracking pixel is a small invisible image or script placed on web pages, while an SDK is a software development kit installed inside mobile or other apps. Both send events to the platform when a user loads a page or performs an action, typically including identifiers and properties.

Question 12

Why does consistent event naming and schema matter for collection?

Accepted Answer

Consistent event naming and schema give every team and system the same meaning for the same action. When one team records a purchase as "order_complete" and another uses "checkout_success," reports split, segments break, and analysis becomes unreliable. A tracking plan documents event names, properties, and data types so that collection stays uniform over time.

Question 13

What is a tracking plan, and who should help create one?

Accepted Answer

A tracking plan is a shared document that lists every event to be collected, along with its purpose, properties, allowed values, and the system that triggers it. Analysts, marketers, product managers, engineers, and privacy staff should all contribute, because each group sees different needs and risks. Reviewing it together before release prevents gaps and duplicates.

Question 14

What is a deterministic identity graph, and what does it store?

Accepted Answer

An identity graph is a database of linked identifiers in which each node is an identifier and each edge records how two identifiers are connected. A deterministic graph joins identifiers only through verified matches, so every link has a known basis. Platforms use the graph to find every identifier belonging to one profile.

Question 15

How should a platform handle conflicting attribute values from different sources?

Accepted Answer

The platform should apply explicit survivorship rules that decide which value wins for each attribute. Common rules prefer the most recent update, the most trusted source, or the value confirmed directly by the customer. Keeping the source and timestamp of every value lets teams audit decisions and correct rules when results look wrong.

Question 16

Why do platforms store raw events separately from derived profile attributes?

Accepted Answer

Raw events preserve every detail exactly as collected, so they can be reprocessed when definitions change or new questions arise. Derived attributes, such as lifetime value or last purchase date, are calculated summaries that are fast to query for segmentation and activation. Keeping both lets teams trace each summary back to its evidence.

Question 17

What is a schema in a customer data platform, and why is it useful?

Accepted Answer

A schema defines the fields, data types, and relationships that records must follow when they enter the platform. It is useful because it catches malformed or unexpected data early, keeps definitions consistent across sources, and lets downstream tools rely on predictable structure.

Question 18

What is the difference between a customer object and an event object in data modeling?

Accepted Answer

A customer object describes a person or account and its stable or slowly changing attributes, such as name, region, or signup date. An event object describes one action at one moment, such as a page view or order, and links back to a customer. This keeps history separate from current state.

Question 19

What are data quality dimensions, and which ones matter most for profiles?

Accepted Answer

Data quality dimensions are measurable qualities such as accuracy, completeness, consistency, timeliness, validity, and uniqueness. For customer profiles, uniqueness and accuracy often matter most, because duplicate or wrong records lead to repeated messages and misdirected offers. Timeliness matters too, since stale attributes can misrepresent current preferences.

Question 20

How can teams detect duplicate customer profiles in the platform?

Accepted Answer

Teams can detect duplicates by monitoring the number of profiles that share the same email, phone number, or device identifier, and by reviewing merge logs for unusual patterns. Regular sampling, where analysts manually compare a set of profiles, catches errors that automated rules miss.

Question 21

What is data lineage, and why do customer data teams need it?

Accepted Answer

Data lineage is a record of where each piece of data came from, how it was transformed, and where it was sent. Teams need it to troubleshoot discrepancies, answer audit questions, and assess the impact of changing a source or definition. Without lineage, tracing a broken audience back to its faulty field is slow.

Question 22

What is a canonical customer data model, and what should it support?

Accepted Answer

A canonical customer data model is a structure that maps each source's fields to shared names and meanings. It should support questions about who the customer is, what they did, and how they relate to other people or accounts. Designing it around business questions, rather than source systems, keeps it useful as sources grow.

Question 23

What is the purpose of a data quality monitor on a customer data pipeline?

Accepted Answer

A data quality monitor automatically checks incoming and stored data against expected rules, such as required fields, value ranges, and volume patterns. Its purpose is to alert owners when something drifts, so problems are fixed before they reach reports, audiences, or customer-facing messages.

Question 24

Why should teams version their schemas and data definitions over time?

Accepted Answer

Versioning records each change to a schema or definition, so teams know which rules applied to data collected at a given point. Without versions, historical reports and audiences can silently change meaning when a field is renamed or redefined. Clear notes also help downstream users adapt before a change breaks them.

Question 25

What is consent management, and why is it central to a customer data platform?

Accepted Answer

Consent management is the practice of recording what each person has agreed to, such as marketing emails or analytics tracking, and enforcing those choices in every downstream use. It is central because a platform spreads customer data widely, so a withdrawn permission must be honored everywhere the data travels.

Question 26

What is data minimization, and how does it apply to customer profiles?

Accepted Answer

Data minimization means collecting and keeping only the personal data needed for a stated purpose. For customer profiles, this means defining which attributes each use case requires, avoiding collection of sensitive fields without need, and deleting or aggregating data once it is no longer useful.

Question 27

What is role-based access control, and how should it apply to profile data?

Accepted Answer

Role-based access control gives people and services permissions based on their job function rather than granting broad access to everything. For profile data, it means marketers may see audience segments while only a small set of approved staff can view identifiers or sensitive attributes. Access should be reviewed regularly and removed when roles change.

Question 28

How do retention policies apply to customer data stored in a platform?

Accepted Answer

A retention policy sets how long each category of customer data is kept and what happens afterward, such as deletion or anonymization. Platforms should apply retention rules per data type, since event history, identifiers, and consent records usually differ in their requirements. Deletion jobs should be tested and logged.

Question 29

What is a data processing agreement, and when does a platform vendor need one?

Accepted Answer

A data processing agreement is a contract that sets out how a service provider may handle personal data on behalf of an organization, including purposes, security duties, and deletion terms. A vendor that stores or processes customer data for the organization generally needs one before data flows begin. Legal teams should confirm local requirements.

Question 30

Why do customers' rights of access and deletion create operational work for platform teams?

Accepted Answer

Access and deletion requests require the platform to find every record linked to one person across all sources, including identifiers in archives and downstream copies. Because identity links may be spread across systems, teams need reliable identity resolution and documented data maps to respond accurately and within the deadlines set by applicable rules.

Question 31

What is pseudonymization, and how does it differ from anonymization?

Accepted Answer

Pseudonymization replaces direct identifiers with a token or key, so records can still be linked for analysis but cannot be attributed to a person without a separately protected key. Anonymization removes the link entirely so that individuals cannot reasonably be identified. Pseudonymized data is still personal data, while truly anonymized data generally is not.

Question 32

How should a platform manage consent records themselves as data?

Accepted Answer

Consent records should be stored as data with the identifier, the purpose, the channel, the time, and the source of each choice. Teams should be able to show what a person agreed to at any moment. Keeping this history protects the organization when anyone asks how permission was obtained.

Question 33

What is audience activation, and how does it work in a customer data platform?

Accepted Answer

Audience activation is the process of building a segment from unified profiles and sending it to destinations such as advertising channels, email tools, or onsite personalization systems. The platform typically syncs members and attributes on a schedule or in near real time, and destinations use that list to target or suppress people.

Question 34

What is a segment, and how does it differ from a one-time list export?

Accepted Answer

A segment is a rule-based group of profiles that updates automatically as profiles change, such as customers who viewed a category in the last week. A one-time list export is a snapshot taken at one moment, so it goes stale as behavior changes. Live segments keep activation accurate, while snapshots suit one-off campaigns.

Question 35

What is a suppression list, and why is it important for activation?

Accepted Answer

A suppression list identifies people who should be excluded from a campaign, such as customers who opted out, recently purchased, or are already in a conflicting audience. It is important because sending to suppressed people can violate consent choices, waste spend, and damage the customer relationship. Apply suppression before data reaches each destination.

Question 36

What is a real-time personalization use case powered by profiles?

Accepted Answer

A common example is a website that checks a visitor's profile when the page loads, then shows product recommendations based on browsing, saved preferences, and past purchases. The same profile can drive an email trigger when a customer leaves items in a cart. Speed matters, since the profile must return before the page renders.

Question 37

How do you measure the health of a customer data platform program?

Accepted Answer

Measure health through indicators such as profile match rate, data freshness, the share of profiles with consented fields, the rate of data quality failures, and the number of activated audiences that teams actually use. Trends matter more than single readings, so review these measures on a regular cadence with owners.

Question 38

What is attribution, and why is it difficult for customer data platforms?

Accepted Answer

Attribution assigns credit for a conversion to the marketing touchpoints that contributed to it. It is difficult because customers move across devices, channels, and anonymous sessions, and many touchpoints lack a stable identifier that links them. Gaps in identity resolution therefore lead directly to incomplete or biased credit.

Question 39

How should teams evaluate whether an activated audience is performing well?

Accepted Answer

Compare the audience against a holdout group that is similar but receives no message, so any difference can be attributed to the activation. Track outcomes such as conversion, retention, or revenue over a defined window, not just clicks or opens. Document the hypothesis and holdout size before launch so results are judged fairly.

Question 40

What operating model helps a customer data platform program succeed over time?

Accepted Answer

A successful program has a clear business owner, a cross-functional team spanning engineering, analytics, marketing, and privacy, and a roadmap tied to use cases. A governance group maintains shared standards for naming, consent, and access, while a prioritized backlog guides new requests. Regular reviews keep the platform aligned with its intended outcomes.

Customer Data Platform Management Flashcards

1Foundations of Customer Data Platforms

2Collection and Identity Resolution

3Data Modeling and Quality

4Governance, Privacy, and Consent

5Activation and Measurement

Learning Resources