Final Assessment — Data Foundations Practitioner

20-question final exam covering the Building Data Foundations for AI course. Pass mark 70%.

Final Assessment

This is the final assessment for Building Data Foundations for AI.

20 multiple-choice questions
Pass mark: 70%
Covers all seven modules
Unlimited retakes
On passing, you earn your Data Foundations Practitioner certification

When you're ready to apply this to a real programme, our AI Enablement service embeds the data foundation work as one of the five enablement pillars — done together with the workflow redesign rather than as a separate stream, which is exactly how foundation programmes survive.

Good luck.

Data Foundations Practitioner Final Assessment

20 questions — Pass mark: 70%

Q1.What is the distinction between reporting data and action data?

AgeReporting data describes what happened; action data drives what happens next inside a workflowStorage locationEncryption

Q2.Why is reporting data inadequate for AI workflows?

Too detailedIt tolerates latency, missing fields, and inconsistent definitions — none of which a system can reason about safely in real timeToo cheapToo secure

Q3.What does 'captured at the point of action' mean?

Data is created inside the application or workflow as the work happensPhotographedManually enteredEncrypted

Q4.What is 'standardised at capture time'?

Definitions, identifiers, and field formats are enforced as data enters the system, not normalised in batch jobs laterAll data is standardised as JSONOnly one team writes dataEncrypted on capture

Q5.What is structured around the workflow, not the report?

A schema that can answer the question a downstream model will ask, without further joins or enrichmentFewer columnsJSON onlyAppend-only

Q6.What latency does an AI-ready data layer require?

Daily refreshPredictable, low-latency freshness — typically seconds or low minutes, never inconsistentMonthly closeLatency doesn't matter

Q7.Why does observable production lineage matter for AI?

Regulators like documentationBecause debugging, drift detection, and regulatory defensibility all require being able to walk the chain of dependencies in real timeIt is fasterIt saves storage

Q8.What is the right starting point for designing an action-data schema?

The reporting requirementsThe next decision the workflow needs to make and the data that decision depends onThe existing warehouse schemaWhat the data team prefers

Q9.What is the wide-row pattern?

A spreadsheet with many columnsA storage layout where each row contains all the precomputed context a downstream decision needs, so the workflow reads it in one operationA reporting tableAn audit log

Q10.Why does identifier discipline matter more for action data?

It doesn'tBecause automated systems join data on identifiers in real time and cannot tolerate inconsistency the way reporting joins canBecause identifiers are harder to encryptBecause regulators count identifiers

Q11.What is a data SLO?

A storage limitA service-level objective: a measurable commitment to a property of the data over a time windowA regulatory documentA compression option

Q12.Why do data incidents need to be treated like production incidents?

To make data engineers feel importantBecause automated systems are making decisions on the bad data in real time, so slow response means more bad decisionsRegulatory ruleCost reasons

Q13.What are the five pillars of data observability?

Speed, cost, scale, latency, throughputFreshness, volume, schema, distribution, lineageEncryption, backup, auth, audit, monitoringETL, ELT, reverse-ETL, CDC, streaming

Q14.What is OpenLineage?

A database typeAn open-source standard for emitting and consuming data lineage events from pipelinesA regulatory frameworkA SQL dialect

Q15.What is column-level lineage and why does it matter for AI?

Lineage at column granularity — necessary for understanding which exact field a model is using and how it was derivedA bar chartLineage in spreadsheet formA regulatory artefact only

Q16.What is a data flywheel?

A database engineA self-reinforcing loop where every action generates feedback that improves the system, which generates more accurate actions, which generates better feedbackA regulatory frameworkA compression algorithm

Q17.Why are AI models commoditising while data moats are not?

Models are slowerAnyone can buy the same model; nobody else has your specific operational data captured at the point of action over timeModels are less regulatedModels cost more

Q18.What is feedback curation?

Reading customer reviewsStructured capture of overrides, exceptions, and corrections so they become trustworthy training signalQA of data entryInternal audit

Q19.Why do most data foundation programmes fail organisationally?

Wrong toolsThey are scoped against an architectural ideal rather than a business workflow, so they get cut when no visible outcome landsToo smallToo cheap

Q20.What is the executive sponsor's primary job in a data foundation programme?

Sign technical decisionsHold political space and protect the team during the long invisible build phaseAttend daily standupsCode review

Certificate Locked

Complete all 7 modules and pass the final exam to earn your Data Foundations Practitioner certificate.