Google Cloud Digital Leader Practice Questions & Answers
A selected foundation path based on the August 12, 2026 exam guide, with original business decisions and explicit coverage gaps.. Selected foundation practice, not a complete exam simulation.
Try 10 original samples with answers below, or start a personal practice set.
Free access · No sign-up · No personal progress recorded on this page · Coverage & review notes
TRY A QUESTION. FOLLOW THE REASONING.
Sample questions with answers
Work out your answer first, then compare the explanation. This stable selection uses distinct task families across the available learning domains.
QUESTION 01Digital Transformation with Google Cloud
Business outcomes and cloud arrangements
A parcel company wants dispatchers to test new route-planning features every week. It currently buys servers for each experiment, although most trials are abandoned. The board asks which proposed measure best connects a cloud initiative to this business objective. Assume data controls and service suitability will be evaluated separately.
Measure how quickly a team can provision an experiment, evaluate it with dispatchers, and retire unused resources.
Measure average processor utilization across the server fleet before and after moving the experiments.
Measure the reduction in quarterly infrastructure invoices without recording how long experiments take.
Measure how many dispatch teams complete the provider's platform-training course during the first month.
Answer: Measure how quickly a team can provision an experiment, evaluate it with dispatchers, and retire unused resources.
The stated objective is a faster learning cycle, not maximum infrastructure consumption. A measure covering provisioning, user evaluation and retirement tests whether teams can experiment and stop paying for unnecessary capacity. Cloud services can support this agility, but a deployment label alone does not demonstrate business value. Cost, access and reliability still need their own evidence; this measure makes no automatic claim about them.
Compare the other answer choices
Measure average processor utilization across the server fleet before and after moving the experiments.
Processor utilization can help investigate infrastructure efficiency, but it does not measure whether dispatchers can evaluate useful experiments sooner.
Measure the reduction in quarterly infrastructure invoices without recording how long experiments take.
Invoice reduction can support a cost objective. It cannot distinguish faster experimentation from simply running fewer experiments or removing useful capacity.
Measure how many dispatch teams complete the provider's platform-training course during the first month.
Training can build skills needed for adoption, but attendance does not establish whether the provisioning, evaluation and retirement cycle has improved.
QUESTION 02Exploring Data Transformation with Google Cloud
Choosing storage and databases
A field-inspection company has two separate requirements. It must upload and retrieve original inspection photographs as files. Its scheduling application must retain its PostgreSQL relational schema while a provider handles routine database maintenance. There is no demand for a global database redesign or historical analytics in this phase. Which TWO service assignments fit these requirements? Select exactly TWO.
Select all answers that apply.
Store the original photographs in Spanner transactional rows.
Store the original photographs as Cloud Storage objects.
Replace scheduling with SQL analysis of historical exports in BigQuery.
Keep scheduling records in a managed PostgreSQL database on Cloud SQL.
Replace PostgreSQL scheduling records with independent objects in Cloud Storage.
Answer: Store the original photographs as Cloud Storage objects.; Keep scheduling records in a managed PostgreSQL database on Cloud SQL.
Split the requirements before selecting products. Original photographs are whole-file objects, so Cloud Storage matches their upload and retrieval pattern. Scheduling records have a separate relational requirement: preserve PostgreSQL and delegate routine database maintenance, which points to Cloud SQL.
The pair gives each workload an appropriate access model rather than forcing files and operational records into one product. Spanner or BigQuery could have other roles after different requirements are established, but neither proposed replacement preserves this project's stated PostgreSQL scheduling requirement.
Compare the other answer choices
Store the original photographs in Spanner transactional rows.
Spanner could manage transactional inspection metadata, but storing original photographs in relational rows is not the stated whole-file object storage requirement.
Replace scheduling with SQL analysis of historical exports in BigQuery.
BigQuery could later analyze inspection history, but replacing scheduling with historical analysis does not preserve its operational PostgreSQL application and schema.
Replace PostgreSQL scheduling records with independent objects in Cloud Storage.
Cloud Storage suits the photographs, but replacing relational scheduling records with independent objects discards the PostgreSQL requirement rather than satisfying it.
QUESTION 03Innovating with Google Cloud Artificial Intelligence
AI problem framing and data quality
A support director wants AI-generated case summaries to reduce the time supervisors spend preparing handovers. Before selecting a model, the team must decide what successful output means. A summary that omits an unresolved safety complaint is unacceptable even if it reads fluently. Which evaluation plan most directly connects the AI task to this business objective?
Rank summaries by word count reduction and use average brevity as the acceptance measure.
Test preservation of required case facts and measure supervisor review effort on representative handovers.
Rank models by a public general-language benchmark and adopt the highest-scoring candidate.
Measure tokens generated per second and select the model with the highest output throughput.
Answer: Test preservation of required case facts and measure supervisor review effort on representative handovers.
Frame success around what the summaries must preserve and what work they should reduce. The omitted complaint is a specific failure, so the evaluation needs task-specific criteria checking required facts alongside a measure of the supervisor's actual review effort.
Brevity, general benchmarks, and throughput can inform secondary tradeoffs, but none directly establishes a safe, useful handover. The Gen AI evaluation service supports tailored criteria and custom evaluation logic; the team must define and test those criteria rather than treating a general model score as evidence for this workflow.
Compare the other answer choices
Rank summaries by word count reduction and use average brevity as the acceptance measure.
Short summaries may save reading time, but brevity alone cannot reveal whether the unresolved complaint or another required case fact was omitted.
Rank models by a public general-language benchmark and adopt the highest-scoring candidate.
A public benchmark can shortlist models, but its tasks and scoring do not directly test this organization's required handover facts or review effort.
Measure tokens generated per second and select the model with the highest output throughput.
Throughput matters for some deployments, but producing summaries faster does not show that they preserve required facts or reduce supervisory correction work.
QUESTION 04Modernize Infrastructure and Applications with Google Cloud
Compute choices and migration decisions
A fictional surveying firm must move a licensed desktop-processing server out of a closing office within one month. Its vendor supports the existing Windows application only when administrators can install operating-system components and maintain the machine configuration. Rewriting or containerizing the application is outside this migration's scope. Which initial Google Cloud approach best meets these constraints?
Move the application into a Cloud Run service and adapt its execution model.
Run the supported Windows environment on Compute Engine virtual machines.
Package the application as Pods and deploy a managed GKE cluster.
Replace the application with a subscribed SaaS workflow during the move.
Answer: Run the supported Windows environment on Compute Engine virtual machines.
Start with the stated compatibility and timing constraints. This move must preserve a vendor-supported Windows environment with administrator control over operating-system components; application redesign is excluded. Compute Engine supplies self-managed VMs, including Windows environments, so it fits that initial migration. Managed containers or a SaaS replacement could be evaluated later, but require application or workflow changes outside this scope. The firm still needs to manage its guest configuration and application security.
Compare the other answer choices
Move the application into a Cloud Run service and adapt its execution model.
Cloud Run would require adapting the application to a supported managed execution model; the stem explicitly excludes that application change.
Package the application as Pods and deploy a managed GKE cluster.
GKE manages Kubernetes workloads, but packaging this application as Pods requires the containerization excluded from the migration scope.
Replace the application with a subscribed SaaS workflow during the move.
A SaaS replacement could remove server administration later, but changing the firm's supported application and workflow exceeds this initial move.
A fictional publisher has an automated reporting workload that must read a particular dataset. A suitable read-only role can be granted on that dataset. The job does not need to modify data or administer other projects, and it should have its own workload identity. Which access proposal best follows the stated requirements?
Grant the workload's service account the suitable read-only role on the dataset.
Grant the workload's service account a broad project Editor role for convenience.
Run the job using a human administrator's identity and existing project permissions.
Grant the workload's service account read access at the organization level instead.
Answer: Grant the workload's service account the suitable read-only role on the dataset.
Identify the principal, the allowed action and the resource scope independently. A service account represents the workload, the read-only role supplies the needed permissions, and the dataset is the narrow resource on which that role can be granted. Broader project or organization grants add access not requested by the job. A human administrator identity also fails the explicit workload-identity requirement. Effective access should still be checked for inherited grants before accepting the configuration.
Compare the other answer choices
Grant the workload's service account a broad project Editor role for convenience.
A project Editor role supplies modification and broader project permissions even though this workload needs only dataset reading and no administration.
Run the job using a human administrator's identity and existing project permissions.
A human administrator identity carries unrelated privileges and fails the explicit requirement that the reporting job have its own workload identity.
Grant the workload's service account read access at the organization level instead.
An organization-level read grant can inherit across resources beyond the particular dataset, exceeding the resource scope explicitly needed by this job.
A fictional architecture practice wants early spending visibility for its experimental project without automatically interrupting designers' work. Finance wants a monthly planned amount, notifications to a nominated project manager, and a warning based on forecasted costs before the month ends. Which TWO settings belong in an alerts-only Cloud Billing budget for this requirement? Select exactly TWO.
Select all answers that apply.
Scope the budget to the experimental project and set its monthly planned amount.
Replace spending thresholds with a request-rate quota on one design-related API.
Configure a programmatic response that disables project billing when a threshold is reached.
Use a spend-cap budget as the intervention mechanism instead of an alerts-only configuration.
Configure a forecasted-cost threshold and a notification channel reaching the project manager.
Answer: Scope the budget to the experimental project and set its monthly planned amount.; Configure a forecasted-cost threshold and a notification channel reaching the project manager.
Translate the request into budget scope, period, amount and notification behavior. An alerts-only budget can cover a specific project, use a monthly planned amount, and alert when forecasted costs cross a selected percentage. A suitable notification channel can reach the nominated project manager. These settings give visibility without the automatic interruption the scenario excludes. A rate quota addresses a particular resource's consumption, while disabling billing or selecting a spend-cap intervention changes the requested operating behavior.
Compare the other answer choices
Replace spending thresholds with a request-rate quota on one design-related API.
A request-rate quota limits a particular API rather than providing the required monthly project spending forecast and manager notification workflow.
Configure a programmatic response that disables project billing when a threshold is reached.
Disabling project billing is an automatic cost-control intervention, which conflicts with the explicit requirement not to interrupt the designers' work.
Use a spend-cap budget as the intervention mechanism instead of an alerts-only configuration.
The requirement specifically asks for alerts-only visibility without automatic interruption; substituting a spend-cap intervention changes that requested behavior.
QUESTION 07Digital Transformation with Google Cloud
Service models, regions and networks
A regional ticketing service has chosen a Google Cloud region that meets its documented location requirements. Its current design places the application and its critical database in one zone. Management now requires the service to remain usable during a single-zone outage; surviving an entire regional outage is outside this project. Which TWO actions directly address and validate this requirement? Select exactly TWO.
Select all answers that apply.
Distribute the application and its critical data dependencies across separate zones with failover.
Move the existing single-zone deployment to a larger virtual machine in that same zone.
Simulate a zone outage and verify that the service continues through its failover path.
Choose a different region while keeping the application and database in one zone there.
Add application replicas in the current zone while leaving the sole database there.
Answer: Distribute the application and its critical data dependencies across separate zones with failover.; Simulate a zone outage and verify that the service continues through its failover path.
Treat a zone as a failure domain and trace every critical dependency. A second application copy cannot preserve service if both copies or their only database depend on the failed zone. Placement and data availability must therefore support operation across separate zones.
The design must also be exercised. Simulating a zone outage checks that traffic and dependencies actually move to a working path. These actions target the specified zonal event; they do not establish protection against loss of the whole region, which would need a separate recovery design.
Compare the other answer choices
Move the existing single-zone deployment to a larger virtual machine in that same zone.
A larger machine can address capacity constraints, but remains inside the same failure domain and cannot serve traffic when that zone is unavailable.
Choose a different region while keeping the application and database in one zone there.
Changing regions may serve location or latency objectives, but a single-zone design still loses its critical components during an outage of that zone.
Add application replicas in the current zone while leaving the sole database there.
Same-zone replicas can tolerate some instance failures, but they share the zone outage risk and still depend on a database in that zone.
QUESTION 08Exploring Data Transformation with Google Cloud
Governed analytics and streaming pipelines
A delivery company currently loads vehicle events into its reporting system once each night. Dispatchers need to notice route disruptions while deliveries are underway. The team proposes continuous event ingestion, managed transformation, analytical storage, and dashboards. Which assignment of Google Cloud services best matches these four roles in that order?
Dataflow for ingestion, Looker for transformation, Pub/Sub for analytical storage, and BigQuery for dashboards.
BigQuery for ingestion, Pub/Sub for transformation, Looker for analytical storage, and Dataflow for dashboards.
Pub/Sub for ingestion, Dataflow for transformation, BigQuery for analytical storage, and Looker for dashboards.
Looker for ingestion, BigQuery for transformation, Dataflow for analytical storage, and Pub/Sub for dashboards.
Answer: Pub/Sub for ingestion, Dataflow for transformation, BigQuery for analytical storage, and Looker for dashboards.
The business change is from delayed reporting to a continuously updated analytical pipeline. In the documented service pattern, Pub/Sub receives external messages, Dataflow reads and transforms them, BigQuery provides the analytical store, and Looker presents business intelligence from that data.
These roles form a pipeline rather than interchangeable names for one database. The arrangement can support timely decisions, but actual end-to-end freshness still depends on configuration and operating behavior. Selecting the appropriate services does not establish a measured latency target or prove that a disruption will be detected.
Compare the other answer choices
Dataflow for ingestion, Looker for transformation, Pub/Sub for analytical storage, and BigQuery for dashboards.
Dataflow can read sources, but this assignment puts Pub/Sub in the warehouse role and Looker in the transformation role, breaking the proposed analytical pipeline.
BigQuery for ingestion, Pub/Sub for transformation, Looker for analytical storage, and Dataflow for dashboards.
BigQuery can ingest data, but Pub/Sub is not the transformation engine and Looker is not the analytical store in the documented architecture.
Looker for ingestion, BigQuery for transformation, Dataflow for analytical storage, and Pub/Sub for dashboards.
Looker consumes analytical data for BI; assigning it initial event ingestion and Pub/Sub dashboard presentation reverses the required business and messaging responsibilities.
QUESTION 09Innovating with Google Cloud Artificial Intelligence
Generative AI, agents and human oversight
A benefits organization is piloting an assistant that drafts explanations for individual eligibility decisions. Errors could materially affect applicants' rights. Staff will remain responsible for final decisions, and the pilot must assess whether the assistance is safe for this context. Which TWO measures most directly support that plan? Select exactly TWO.
Select all answers that apply.
Have responsible staff inspect the evidence and draft before issuing each final eligibility decision.
Send drafts to applicants at once, with trained staff auditing a randomly selected sample the next month.
Use provider benchmark results and a latency comparison to choose the model for the eligibility pilot.
Evaluate realistic and difficult eligibility cases against factual and safety criteria before expansion.
Let high-confidence drafts proceed without pre-decision review and send lower-confidence drafts to staff.
Answer: Have responsible staff inspect the evidence and draft before issuing each final eligibility decision.; Evaluate realistic and difficult eligibility cases against factual and safety criteria before expansion.
The case retains staff responsibility for rights-sensitive final decisions and needs evidence of contextual safety. Staff need a meaningful opportunity to inspect relevant evidence, correct the draft and intervene before the decision is issued. Realistic and difficult eligibility cases also need factual and safety evaluation before expansion.
Later sample audits can discover patterns, but they cannot correct each issued decision beforehand. General provider results do not evaluate this particular eligibility workflow. Model confidence is not a demonstrated guarantee that a consequential draft is factually correct or safe. These controls support the stated pilot; they do not establish legal compliance or calibrated risk.
Compare the other answer choices
Send drafts to applicants at once, with trained staff auditing a randomly selected sample the next month.
A later audit does not give staff an opportunity to correct every consequential draft before the associated decision reaches an applicant.
Use provider benchmark results and a latency comparison to choose the model for the eligibility pilot.
Provider benchmarks and latency comparisons do not evaluate this organization eligibility facts or the stated factual and safety criteria.
Let high-confidence drafts proceed without pre-decision review and send lower-confidence drafts to staff.
A confidence threshold alone does not establish correct eligibility reasoning, and the high-confidence path removes meaningful staff review before some final decisions.
QUESTION 10Modernize Infrastructure and Applications with Google Cloud
Modern applications, APIs and portability
A fictional research network will run an application in GKE and in a second environment that already provides Kubernetes. Both environments have compatible operating systems and processor architectures. The team wants to reuse its application packaging and Kubernetes deployment workflow, while accepting that identity, networking, and external services still need adaptation. Which TWO design choices directly support that aim? Select exactly TWO.
Select all answers that apply.
Package the application and its runtime dependencies in compatible container images.
Make a saved Compute Engine VM image the application's sole deployment artifact.
Use Kubernetes workload definitions through the Kubernetes interfaces in both environments.
Replace the shared deployment workflow with calls specific to the Cloud Run management interface.
Make every release depend on manually rebuilding an environment-specific VM.
Answer: Package the application and its runtime dependencies in compatible container images.; Use Kubernetes workload definitions through the Kubernetes interfaces in both environments.
There are two reusable layers in the stated objective: the application package and the Kubernetes deployment workflow. Container images package the application for compatible container environments, while Kubernetes workload definitions retain the common orchestration interface. GKE implements Kubernetes and deploys containers as Pods. These choices support reuse; they do not prove that credentials, networking, storage or managed-service dependencies transfer unchanged. The scenario explicitly reserves that integration work, avoiding a claim of effortless portability.
Compare the other answer choices
Make a saved Compute Engine VM image the application's sole deployment artifact.
A Compute Engine VM image is an infrastructure artifact, not the shared container package requested for the two existing Kubernetes environments.
Replace the shared deployment workflow with calls specific to the Cloud Run management interface.
A Cloud Run-specific management interface replaces the Kubernetes deployment interface that the team explicitly wants to reuse in both environments.
Make every release depend on manually rebuilding an environment-specific VM.
Manual environment-specific VM rebuilding introduces a separate release path rather than reusing compatible container packages and the existing Kubernetes workflow.
Six domain-level foundation groupings; no complete bullet-by-bullet coverage claim.
Learn service boundaries, data and AI decisions, security and operations through a free foundation path with original cases.
64 original fixed tasks (52 single choice and 12 select-TWO), not a full-length official examination or score predictor.
10 public examples64 question variants in the studio64 task families
Review status: Independent educator review and learner calibration are pending. Use these selected foundation examples alongside the coverage map and current official references.
Start with the lesson linked to an unfamiliar topic, try a focused set in the practice studio, then return to the review cards. Your practice progress stays in this browser.
Catalog scope record: . This records the saved content mapping, not a guarantee that every external rule is current.
Selected foundations across six domains of the official guide launched August 12, 2026; not complete bullet-by-bullet syllabus coverage.
64 original fixed tasks (52 single choice and 12 select-TWO), not a full-length official examination or score predictor.
Difficulty and reading time are editorial estimates; no official passing score, calibrated readiness or score prediction is offered.
Independent educator review remains pending; this foundation path has not undergone a learner trial.
Check the exam provider's current information for registration, test rules and official preparation materials. The exam guide collects the relevant official references.
