Informatica Data Engineering Foundation flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

Mapping vs Mapping Task

A mapping is the reusable source-to-target transformation definition. A mapping task is a runnable instance that binds a mapping to a runtime environment and execution settings.

Active transformation

A transformation that can change the number of rows passing through it, such as Aggregator, Filter, Router, Joiner, or Rank. Contrast with passive transformations like Expression.

Connected vs unconnected lookup

A connected lookup receives input from the pipeline and returns one value per row. An unconnected lookup is invoked from an Expression using :LKP and can return multiple outputs.

Pushdown optimization

A technique that converts transformation logic into SQL executed by the database, reducing data movement. Full pushdown requires every transformation to be pushdown-compatible.

SCD Type 2

A slowly changing dimension technique that preserves history by inserting a new versioned row with effective dates and a current-record flag when an attribute changes.

Incremental load watermark

A stored high-water mark, such as a timestamp or sequence value, used to select only new or changed source rows on each run and avoid full reloads.

Session recovery strategy

A setting that determines how a failed session resumes: restart from the beginning, resume from the last checkpoint, or fail without recovery.

Data quality dimensions

Standard measurable characteristics of data quality: completeness, accuracy, consistency, timeliness, uniqueness, and validity. Scorecards aggregate these per dataset.

Quality score formula

Per-rule quality score equals passing records divided by total records, multiplied by 100. Dataset scores are weighted aggregates across rules and dimensions.

Exception record

A record that failed a data-quality rule and was routed to a remediation queue with owner, severity, and status such as open, in review, or resolved.

Data observability pillars

Five monitoring signals for datasets: freshness, volume, schema, distribution, and lineage. Alerts fire when a monitored signal breaches its threshold.

Freshness SLA breach

A condition where a dataset's latest update time is older than its agreed freshness threshold, indicating delayed delivery to downstream consumers.

Column profiling baseline

A pre-rule statistical summary of a column, including null percentage, distinct count, min/max, and pattern frequency, used to set evidence-based quality thresholds.

Severity vs priority

Severity reflects business impact if the issue is unresolved. Priority determines remediation order. They are separate fields and should not be conflated in alert routing.

Schema drift

An unexpected change to a dataset's structure, such as added, removed, or retyped columns, which can break downstream pipelines and should trigger an observability alert.

MDM implementation styles

Four common approaches: registry (index only), consolidation (read-only golden record), coexistence (bidirectional synchronization), and centralized (single system of entry).

Match threshold bands

Score ranges that classify candidate record pairs: above the upper threshold auto-match, between thresholds route to manual review, below the lower threshold no match.

Survivorship trust score

A per-source, per-attribute weight that determines which source value populates the golden record when multiple systems supply conflicting values for the same entity.

Cross-reference (XREF)

A stored link between a master record and its originating source-system identifiers. XREFs enable traceability and reversible merge and unmerge operations.

Merge and unmerge

Merge combines duplicate records into one master record while retaining source links. Unmerge reverses the operation, restoring the original source records.

Data steward vs custodian

A steward owns business meaning, quality remediation, and approval decisions for a domain. A custodian owns technical storage, pipelines, and access implementation.

Golden record

The single trusted version of a master entity, assembled from source records using match rules and survivorship rules, and used as the authoritative reference.

Party master data

Master data describing people and organizations, including customers, prospects, suppliers, and employees, typically deduplicated and linked into hierarchies.

Fuzzy matching

A record-linkage technique that scores similarity rather than requiring exact equality, using methods such as edit distance, Jaro-Winkler, or phonetic encoding.

Governance operating model

The assignment of decision rights: policy owners in the business, domain data stewards, IT custodians, and a cross-functional council for enterprise-wide decisions.

Data classification tiers

Labels such as public, internal, confidential, and restricted that determine access controls, masking requirements, retention, and permitted AI training use.

Impact vs root-cause lineage

Impact analysis traces downstream consumers affected by a changed column. Root-cause analysis traces upstream sources that produced a bad value.

Model card

A governance document describing a model's intended use, training data provenance, evaluation metrics, known limitations, and monitoring plan for drift.

Legal hold

A directive that suspends normal retention deletion for specified data because of litigation or investigation. It overrides standard retention expiry rules.

Business glossary term

A governed definition of a business concept with an owner, approved meaning, synonyms, and links to related assets. It differs from physical column metadata.

Certified asset

A catalog asset flagged as trusted after review by a data steward. Uncertified assets should not feed regulated reporting without additional validation.

Column-level lineage

A mapping that traces a target column back to its specific source columns through transformations. It is more precise than table-level lineage but harder to capture automatically.