Data services / Data Analytics
Dashboards whose definitions, access and freshness can be checked.
Plan executive, sales, marketing, finance, operations and product analytics around approved sources and business definitions. Tableau, Power BI, Looker, dbt and Airflow are implementation options, selected to fit the source systems, users and decisions.
Discuss Data Analytics requirements βContents Β· Data Analytics
Begin with the decision, not the chart.
A useful analytics brief names the audience, decision, metric owner and consequence of a wrong number. Inventory CRM, ERP, product, payments, marketing and operational sources. Check whether an existing report already answers the question before proposing a new pipeline or dashboard. A visual template cannot resolve conflicting business definitions or missing source records.
Delivery scope can include source integration, modelled data, KPI definitions, access rules, visualizations, refresh and operating instructions. Agree and test each part against the actual accounts, datasets and audience. Assign source, metric and operating owners, with clear acceptance and unresolved issues.
Keep analytics scope distinct from application delivery. A dashboard is not automatically a mobile application, a live streaming system or an audited financial report. Define the audience, metrics, sources, access, freshness and decisions for each analytics product below.
Define the reporting contract.
- Name each source, data owner, permitted purpose, API/export capability and history available.
- Write the grain, formula, filters, timezone, currency and business owner for every KPI.
- List viewers, customer/region boundaries, edit rights, exports and embedded surfaces.
- Specify source freshness, refresh completion, cache behavior and the visible as-of timestamp separately.
- Agree expected data volumes, concurrent viewers, comparison periods, failure behavior, operating roles and acceptance evidence.
Six analytics solution concepts, not generic clone applications.
Executive Dashboard
Recorded scope: revenue, growth, churn and operational KPIs across five-plus sources, with drill-down in Tableau. Confirm the actual sources, joins, audience and licence mode. Define whether a drill-down exposes personal or commercially restricted records, and reconcile totals before leaders rely on the view.
Sales Analytics Hub
Recorded scope: CRM and ERP data in Power BI, covering pipeline, conversion, forecast accuracy and regional performance. Define opportunity stages, reopened deals, effective dates, quotas and ownership changes. Sales forecasts and booked revenue need separate definitions and evidence; a combined chart does not make them interchangeable.
Marketing Attribution
Recorded scope: multichannel campaigns, cohorts, channel mix and acquisition/lifetime-value measures, with Looker. Agree attribution model, consent, campaign identity, lookback, cross-device limits and missing events. Distinguish reported attribution from causal impact; integrating ad platforms does not prove incremental return.
Financial Reporting Platform
Recorded scope: profit and loss, cash flow, transaction drill-down and close workflows, with Snowflake and dbt. Confirm accrual versus cash, refunds, taxes, fees, FX and adjustments. A payment-provider payout is not automatically recognized revenue or a complete general ledger.
Real-Time Operations Monitor
Recorded scope: queue depth, service-level indicators, incident metrics, streaming signals and threshold alerts. State the event-to-view delay, sampling, missing-data behavior and operator response. A scheduled refresh or a green job status is not evidence the view is current or an intervention occurred.
Product Analytics Suite
Recorded scope: funnels, retention, feature adoption and experimentation using Python and Plotly/Dash. Define event contracts, identity resolution, test accounts, exposure and cohort rules. Experiments need design and statistical review beyond dashboard construction; no uplift or decision-quality outcome is asserted.
Company delivery, internal ownership or packaged reporting.
| Criterion | Scoped supplier engagement | Internal team | Packaged BI / reporting |
|---|---|---|---|
| Scope | Agree pipelines, semantics, access, visualization and handover as explicit deliverables. | Check available engineering, analysis, design and operations capacity. | Verify the actual features, supported sources and customization limits. |
| Change control | Name reviewers, acceptance owners and change-request boundaries. | Document priority and ownership across reporting and operational work. | Check upgrade behavior, configuration export and plan restrictions. |
| Cost | Separate implementation, licences, cloud, connectors and support in the quote. | Include staffing, tooling, infrastructure, review and incident coverage. | Include licences, integration, maintenance, usage and migration costs. |
| Integration | Test source objects, history, permissions, semantics and failure behavior. | Assess the same access and compatibility requirements; local knowledge is useful but not sufficient. | A prebuilt connector may not support the required object, deletion or history. |
| Fit | Useful where an agreed cross-system scope and receiving team can be defined. | Useful when continuing questions and operating ownership match team capacity. | Useful when standard reports genuinely answer the decision with acceptable access and governance. |
The source tableβs blanket setup intervals, salary ranges and cost rankings are not adopted. None of these models is inherently fastest, cheapest or most accurate. Compare the actual scope, people, dependencies and exit path. A consultant can handle a focused tool/report task while other teams retain platform ownership.
Six questions before dashboard implementation.
What does implementation include?
Separate source access, extraction, transformations, semantic definitions, visual design, publishing, permissions and handover. Agree which dbt models, Airflow jobs, workbook files, dictionaries and runbooks are included. Existing infrastructure may reduce scope, but a chart is not proof the layers beneath it are present or tested.
Which BI tool fits?
Compare the audienceβs existing environment, exploration needs, embedded use, accessibility, exports, governance, support and licensing. Tableau, Power BI, Looker, Googleβs Data Studio/Looker Studio and custom web dashboards have different operating models. Lookerβs semantic platform is distinct from the reporting product historically called Looker Studio. Verify current product names and the actual plan rather than silently treating them as one tool.
Why define KPIs before visual design?
Revenue, active customers, churn and recurring revenue can each have legitimate definitions. Write the grain, time window, exclusions and owner, then test approved examples. Finance and sales may need different measures with visible labels; forcing them into one unexplained number can hide a real business distinction.
How are delivery effort and running cost assessed?
Review source quality, model complexity, access, data volume, refresh, migration and review dependencies before setting milestones. Budget separately for BI licences/capacity, warehouse, connectors, hosting and support. A custom dashboard is not automatically licence-free or cheaper at scale; compare the complete operating cost and maintenance responsibility.
Consultant, agency or internal analyst?
A focused report change, a cross-system platform and an ongoing business-analysis function can need different people. Agree domain ownership, code review, incident coverage and handover. Avoid treating job titles or supplier geography as evidence of quality, speed or cost.
Analytics, engineering or data science?
Analytics communicates defined measures and patterns. Engineering moves and prepares data reliably; science develops and evaluates predictive or experimental methods. The boundaries can overlap. Use the current data and decision to choose scope, rather than require every project to follow an identical sequence.
Four layers with explicit contracts.
Ingestion, transformation, storage and visualization
Document how approved source records become a model and then a rendered view. Keep provenance and the business grain visible across those boundaries. Decide whether a semantic layer is shared across workbooks and what happens when a definition changes. A folder hierarchy is not a compatibility guarantee.
dbt and Airflow have different responsibilities
dbt can version SQL models, dependencies and tests; Airflow can coordinate scheduled work. The implementation must define selected resources, run order, retries, access and publication. An imported library or function named refresh does not create a DAG, schedule or consumer-visible guarantee.
Query performance and incremental refresh
Measure actual warehouse queries and rendered views with representative data, roles, filters and concurrency. Use the platformβs supported optimization methods. Indexes are not universal across warehouse engines; materialized views, partitioning, clustering and caching have their own limits. No fixed dashboard response time is claimed.
Build a metric dictionary that survives source changes.
For each measure, record its business question, formula, grain, timezone, currency, inclusion rules, source fields and owner. Explain how refunds, reversals, taxes, fees, discounts, late events and deleted records affect it. Compare appropriate source totals rather than assume a Stripe payout equals revenue. Define whether historical periods are restated after corrections.
For subscription metrics, distinguish active accounts, paid accounts, recurring revenue, trial status and cancellation dates. Define cohort entry, retention period and identity changes. A many-to-many join can multiply values even when every table is individually valid. Include fixtures that expose duplication, missing keys and boundary dates.
A lineage graph can show declared table/model dependencies but may omit workbook calculations, custom SQL, exports or manual adjustments. State where tracing stops. Traceability to a source table is not proof that a business measure is correct or that every number can be reconstructed from one source row.
The source refresh function is not a publication guarantee.
The original analytics_pipeline.py executable sequence is retained below. It runs dbt build for selected revenue ancestors, signs into Tableau using environment-based PAT credentials, chooses the first filtered workbook, requests a refresh and prints a message. Scheduling and absolute-publication comments are qualified; no function invocation, DAG, task or schedule is supplied.
A nonzero dbt process stops this function, but selected tests can have warning severity, freshness may be a separate check and other resources are outside the selection. Models can be materialized before a later test fails. This is not a cross-model transaction or a guarantee that other dashboards never see changed data. Define an accepted-dataset promotion step if that boundary matters.
Selecting [0] can fail or choose an unintended workbook when names are missing or ambiguous. Verify account, site, project and stable workbook ID, plus the operatorβs access. The PAT name airflow is a label, not proof of a scheduler identity or least-privilege role. Dependencies, configuration, database credentials and model files are not included.
Tableauβs refresh operation returns a job. The sample ignores it, so the print statement does not prove the refresh completed, succeeded, used current data or rendered the intended KPIs. Poll using supported job behavior with timeout/failure handling, then verify source timestamps, dataset version, filters and view values. The separate job and dbt build do not form one atomic transaction.
Technical example
Source sample, not an installed backend or verified deployment. The limitations above apply.
# Refresh step for the Executive Dashboard pipeline
# Scheduling and dependencies are not supplied in this function
import os
import subprocess
import tableauserverclient as TSC
def refresh_revenue_dashboard():
"""Build and test the revenue models, then refresh the Tableau workbook."""
# A nonzero dbt process stops this function; this is not an atomic publication gate
subprocess.run(["dbt", "build", "--select", "+bi_revenue"], check=True)
auth = TSC.PersonalAccessTokenAuth("airflow", os.environ["TABLEAU_PAT"], site_id="analytics")
server = TSC.Server(os.environ["TABLEAU_URL"], use_server_version=True)
with server.auth.sign_in(auth):
workbook = server.workbooks.filter(name="Executive Dashboard")[0]
server.workbooks.refresh(workbook)
print("Executive Dashboard refreshed from tested models")Show freshness and failure honestly.
Distinguish event time, latest source availability, model completion, extract refresh, server cache and client render. An hourly-looking chart or a daily task name does not establish any of these. Display an approved as-of indicator and explain whether a view uses a live query, extract or cached result. Consumers need to know when the last accepted dataset is stale.
Define what happens when a source fails, tests warn, a refresh times out or only some models finish. A previous good view may remain visible and become stale. Decide whether to block, label, fall back or notify, and test the actual path. Ensure an all_done leaf cannot make an upstream failure look like a successful overall release.
Retries need bounded behavior and overlap control. A second refresh request after a lost response may create another job; a recovered task may find a prior write already committed. Record output identity and job state. Separate operational alerts from approval to send reports or share restricted data with a wider audience.
Power BI roles, embedded identity and source credentials differ.
Microsoft documents that Power BI row-level security restricts workspace Viewers, not Admin, Member or Contributor roles. Test with actual intended viewers and negative cross-tenant cases, including exports and underlying-model permissions. Hiding a field in a visual is not a substitute for model-level restrictions. Review object-level controls when columns also need protection.
In embedded flows, the applicationβs authentication, embed token, effective identity, roles and semantic model must align. A service principal or a token alone does not establish the intended customer boundary. Test actual embed tokens and identities, not only a desktop preview or a generic Test as role session.
Googleβs reporting data-credential modes also matter. Ownerβs Credentials can authorize queries using the credential ownerβs access; this is not the same as each viewerβs warehouse identity. Viewer credentials and supported service-account modes have different setup and limitations. Sharing a report does not automatically apply the warehouseβs per-user policy. Document and test the selected mode.
Review Tableau site/project/workbook rights, PAT ownership, token expiry and extract credentials. Environment variables keep the sample from hardcoding a value, but do not prove safe issuance, secret storage, log handling or rotation. Keep permitted secrets out of code and exports, and revoke access at the end of the agreed work.
Incremental refresh is a model and source policy.
Power BI incremental refresh uses RangeStart and RangeEnd filters and a supported policy. Verify source-side filtering or query folding, historical partitions, late updates, deletions, timezone boundaries and the real-time partition mode if used. A configured parameter does not by itself prove the source query avoids scanning full history.
For dbt incremental models, confirm keys, nulls, duplicates, adapter strategy, schema evolution and lookback. The model should not silently miss a correction outside the refresh window. Run representative backfill and migration tests and compare accepted measures before promoting a new dataset.
Refresh success is distinct from data quality and business freshness. A completed BI job can read an old extract, a differently filtered source or a dataset whose upstream test only warned. Check output timestamps, record counts, approved KPI examples and access with the actual consumer roles.
Versioning code does not restore overwritten facts.
Git tags and workbook exports can preserve definitions. dbt snapshots can record supported row history when configured and run. Neither is a universal backup of source data, permissions, warehouse state or externally emailed reports. Document retention, restore routes and what cannot be undone before describing rollback as a simple deployment switch.
A schema or metric change may require data migration, historical restatement and coordinated consumers. Test compatibility with the prior workbook and model versions. Reconcile after restoration, confirm access and refresh status, and show the receiving team how to identify the last accepted dataset.
If reports have already reached people, changing a dashboard does not recall their copies. Plan corrections and recipient handling as separate authorized communication. Authorize the recipients, contents, schedule and access for any report distribution separately from dashboard implementation.
Three analytics engagement shapes.
Dashboard delivery package
Proposed scope may include approved sources, transformations, a semantic definition, dashboards, refresh and handover. Confirm the exact audience and views, repository and workbook access, licences and acceptance. Historical package price or duration is not a current offer.
Custom analytics platform
Define custom architecture, KPI models, embedded or role-based views, operational needs and migration. Name engineering, analyst, QA and delivery responsibilities. A custom label does not mean every connector, source or security mode is supported.
Ongoing analytics work
Agree new reports, source/model changes, quality reviews, maintenance and incident coverage. Separate enhancements from defects and record support terms, fees, cancellation and access revocation. No standing team or unlimited service is created here.
Seven quality principles, with observable evidence.
Architecture
Keep ingestion, transformation, semantic and visual layers documented, with explicit owners and change boundaries.
Data quality
Test grain, keys, relationships, nulls, completeness, freshness and the approved business examples. A count check alone is insufficient.
Performance
Profile representative queries and rendered views, including filters, roles, concurrent viewers and cache states. Agree measurable acceptance rather than a universal speed claim.
Validation
Reconcile staging and published measures with appropriate source records and explain timing or definition differences.
Reliability
Test scheduling, job polling, warning/failure behavior and the visible stale-data policy, with named operators.
Security
Verify actual roles, source credentials, customer boundaries and export permissions. A login page does not prove row isolation.
Governance
Keep the dictionary, dependency/lineage limits, ownership and change record available to the receiving team.
Six analytics delivery gates.
SQL and pipeline review
Review changed SQL, DAGs, workbook calculations and infrastructure with responsible owners. Inspect actual permissions and CI configuration; the policy described here does not prove enforcement.
Automated validation
Agree schema tests, freshness checks, semantic fixtures and reconciliation thresholds. Set test severity and publication behavior explicitly. Verify that a failing check does what the audience expects.
Staging sign-off
Use approved representative volume and distributions. Inspect visual totals, filters, accessibility, render performance and role-specific views. Include missing-source and stale-data cases, not only the successful refresh.
Handover package
Specify the dbt project, DAGs, workbook exports, dictionary, access matrix, configuration and runbook. Transfer credentials through approved secret handling. Verify that the receiving team can deploy and troubleshoot the delivered system.
Version and restore rehearsal
Record compatible model, data and workbook versions and test a supported restore path. Include overwritten facts, retention and permission changes. A snapshot or Git tag is not automatically a complete rollback.
Monitoring and support
Test the selected refresh, job-failure, data-quality and usage signals. Agree recipients, response coverage and maintenance boundaries. Do not infer unconditional active monitoring or a support interval from the sourceβs old badge.
Sixteen analytics stack areas to qualify.
| Technology | Role and checks |
|---|---|
| Tableau | Publishing, extracts, jobs, roles, workbook identity and supported client versions. |
| Power BI | Semantic models, refresh, licences/capacity, Viewer RLS and embedded identity. |
| Looker | Semantic modeling and embedded analytics, with actual model and user-attribute access. |
| dbt | Selected SQL transformations, tests, contracts, freshness and model documentation. |
| Airflow | Orchestration, dependencies, timetable, retries and failure/leaf behavior. |
| Python | Custom analytics dependencies, environment, security and operating ownership. The original version label is historical, not a current runtime recommendation. |
| SQL | Grain, joins, aggregations, nulls and source-specific execution plans. |
| Snowflake | Warehouse roles, compute/storage and supported operational controls. |
| BigQuery | Storage, access, partition/cluster design and pricing-mode-specific query controls. |
| PostgreSQL | Operational or reporting source, with query load, permissions and transaction semantics. |
| Redis | Optional cache with expiry, tenant separation and invalidation policy. |
| Kafka | Optional event ingestion with tested offset/output and replay boundaries. |
| Grafana | Operational views and configured alerts, not proof of notification or intervention. |
| pandas | Data manipulation and exports with memory, types, nulls and privacy checks. |
| Plotly / Dash | Custom views with application authentication, access and hosting to implement. |
| Fivetran | Managed connectors whose actual source/plan/history and usage terms need checking. |
Five stages from brief to accepted reporting.
Brief and confidentiality
Name decisions, audiences, source owners and restricted data. Agree confidentiality and minimize samples before sharing sensitive records.
Scope and plan
Approve sources, KPI definitions, visual/tool choice, access, quote, milestones and exclusions. Identify licence and cloud dependencies.
Build and review
Version transformations and view definitions. Demonstrate approved examples, permissions and evolving decisions in a reviewable environment.
QA and reconciliation
Test representative data, cross-role access, query/render behavior, refresh failures and business totals. Record sign-off and unresolved limits.
Deploy and hand over
Use authorized accounts, verify the accepted output and provide the dictionary, code/workbooks and operating instructions. Confirm support and exit terms separately.
Quote the scope and the operating model.
Source count, source quality, historical data, real-time needs, custom calculations, RLS, embedded use, regional deployment and API integration affect the scope. Review these before an itemized current quote. The source pageβs price and delivery ranges are not repeated as present terms, and no fixed deadline or surprise-free billing guarantee is created here.
BI licensing, capacity, warehouse compute/storage, connectors, application hosting, monitoring and maintenance can remain ongoing costs. Verify the actual plan and who pays each service. Power BI sharing rules vary with licence and capacity; a custom Dash implementation also needs secure hosting, operations and maintenance. Do not assume a cost ranking from viewer count alone.
Treat ownership as project-specific contract language: distinguish original code, third-party software, vendor licences, source data rights and cloud accounts. Confirm export and exit routes before committing. An asserted source-code handover does not transfer rights to every connected source or commercial product.
SaaS executive dashboard: recorded scope, not a funding result.
The supplied example joins five named sources: PostgreSQL, Google Analytics/GA4, Stripe, Salesforce and HubSpot. It describes twelve KPI modules across revenue, growth, churn and operations, Tableau views, dbt transformations, Airflow scheduling, drill-downs, role-based access and weekly report distribution. These are unverified recorded requirements, not a confirmed client installation.
The original labels the example illustrative while also describing delivered work and a board/fundraising outcome. Its speed and decision-improvement claims are not accepted as evidence. Ask which client, datasets, metrics, identity rules and measurement period support the story and whether publication is permitted. No funding, productivity or performance result is asserted here.
Retain the useful acceptance questions: Which source owns each KPI? How do payment states and CRM opportunities reconcile? What is the accepted as-of time? Can the viewer inspect permitted detail without crossing customer boundaries? Are emailed copies and recipients separately approved? A successful refresh alone answers none of these.
Require references that match the analytics scope.
When previous dashboard work informs a decision, verify the client-authorized reference, service, team and actual workload. A personal-finance, security or billing product reference may concern a different scope. Inspect the source, metric definitions, access and operating evidence relevant to your reporting requirements.
Six supplier checks before a commitment.
Company identity
Ask for the contracting legal entity, registration evidence and matching invoice details.
Numbers and delivery claims
Ask for dated definitions and supporting records behind project counts, team size, prices and delivery promises. A historical site counter or a technology reference does not establish present capacity.
Independent reviews
Check the actual profile, reviewer, product, date and scope on Clutch or another review service. Check that a reference concerns the same service and obtain client-authorized evidence relevant to your workload.
A second review source
Compare a separate review source such as GoodFirms with the proposed team and scope. Resolve identity mismatches and ask whether the reviewer used the same service, not simply the same brand.
The product and handover
Request a permitted demonstration and inspect the repository, operational evidence and handover contents. A catalogue card, code sample or screenshot is not proof of an installed application, working integration or production load.
Client permission and relevance
Request references the client has authorized, tied to the actual data workload and team. Personal finance, security and billing product testimonials do not by themselves verify a pipeline or dashboard engagement.
Replace broad promises with reviewable terms.
People and responsibilities
Before a commitment, ask who owns data engineering, analytics, QA, delivery and incident response. Put access, review duties and communication arrangements in the contract with the actual named roles and capacity.
Evidence before claims
Ask for the agreed source inventory, acceptance tests, observed measurements, current licences and project-specific rights. Documentation supports technology behavior, not Omnistu delivery history or customer outcomes.
Written scope and handover
Agree milestones, acceptance, exclusions, dependencies, support and change control in writing. Specify custom and reusable-component rights, the receiving responsibilities and any ongoing service levels.
Eight industry contexts change reporting and access.
Healthcare and life sciences
Separate clinical, billing and research uses. Agree permitted data, identifiers, retention, access review, quality thresholds and human review. Dashboard outputs should distinguish missing records from a genuine clinical change; a chart or pipeline is not a clinical decision system.
Telemedicine
Join appointment, consultation and billing events only for approved purposes. Define visit status, cancellations and late corrections. Keep patient identifiers out of routine logs and email exports. Verify the rules of the actual care organization and region rather than claim generic compliance.
Fintech
Document transaction states, reversals, fees, settlement time, currencies and ledger reconciliation. Separate a fraud signal from a confirmed event. Access, audit evidence and retention need explicit owners; a streaming feed does not prove regulated readiness or financial accuracy.
Retail and e-commerce
Define order, shipment, return, refund, stock and marketplace seller identities. Reconcile gross and net measures, duplicate events and delayed fulfilment changes. Check consent and permitted uses before joining customer behavior with purchases or advertising data.
Media and entertainment
Separate content events, subscriptions, entitlements and payouts. Agree what a play, active viewer or conversion means across devices. Restrict personal viewing history and content rights metadata; review deletion propagation through extracts, aggregates and backups.
Creator economy
Distinguish creators, channels and audiences across platforms. Document attribution limits, payout adjustments and bot filtering. A public profile does not grant rights to reuse every event or asset. Check platform API terms and access before committing to a connector.
Transportation and mobility
Model trip states, cancellations, location freshness and operator/driver access. Separate GPS samples from completed trips and earned revenue. Minimize precise location retention, account for timezone boundaries, and verify which operational decisions may use delayed data.
Food and beverage
Join menu, ordering, kitchen, delivery and inventory events with clear business keys. Account for substitutions, refunds, waste and changing menus. Restrict customer and courier data; freshness alerts should distinguish a source outage from a real fall in demand.
Data analytics questions.
Which dashboard types are in scope?
The retained concepts cover executive, sales, marketing, financial, operations and product analytics. Define actual KPIs, sources, audiences and permissions for the engagement; no universal installed template or working demo is claimed.
What determines the quote?
Sources, history, model and KPI complexity, access, embedded use, refresh, migration, infrastructure and support determine the itemized quote. Keep vendor licences, usage and ongoing maintenance separate from implementation.
Can any data source be integrated?
No universal compatibility claim is made. Check permissions, terms, endpoint/object support, export history, pagination, rate limits and deletions. An APIβs existence does not guarantee approved or complete access.
How is delivery timing agreed?
Review dependencies, access, representative data, review capacity and acceptance before milestones. No historical fixed interval, app delivery guarantee or immediate-deployment promise is offered here.
Which BI tool should be selected?
Compare existing environment, governance, exploration, embedding, access, exports, licensing and maintenance. Tableau, Power BI, Looker, reporting products and custom views need workload-specific evaluation, not a universal winner.
Can reports and digests recur?
They can be proposed if the selected platform supports them. Agree recipients, permitted fields, formats, schedules and failure handling, then obtain the necessary communication approval.
What maintenance follows delivery?
Agree monitoring, data-quality reviews, fixes, new reports, source changes, incident coverage and cancellation in current terms. An ongoing retainer does not automatically include every enhancement or an uptime guarantee.
How is data secured?
Review actual IAM/RBAC, row and column restrictions, source credentials, encryption configuration, logs, retention, exports and embedded identity. Verify controls and any required audit; intended controls do not establish certification or blanket regulatory compliance.
How is accuracy checked?
Use approved metric definitions, grain-aware reconciliation, tests and representative business examples. Explain timing differences and lineage gaps. A successful job or source-row count does not guarantee that every business number is correct.
Can existing licences and workspaces be used?
Verify the actual accounts, plans, roles, capacity and publishing permissions. Existing subscriptions may fit, but do not assume every viewer or embedded scenario is covered. Validate an inherited report before choosing repair or replacement.
Can analytics be embedded for customers?
Assess the chosen BI or custom application mode, authentication, effective identity, tenant restrictions, exports, licences and operating costs. Test customer isolation with actual allowed and denied cases. No built-in isolation or lower-cost guarantee is asserted.
Primary references for technical qualification.
Official platform, framework and API documentation checked on 17 November 2026. These support the technical checks here, not Omnistu staffing, pricing, rights, delivery or client outcomes. Recheck installed versions and configuration before implementation.
- Best Practices - Airflow 3.4.0 Documentation - airflow.apache.org
- Dag Runs - Airflow 3.4.0 Documentation - airflow.apache.org
- About dbt build command | dbt Developer Hub - docs.getdbt.com
- severity, error_if, and warn_if | dbt Developer Hub - docs.getdbt.com
- Source freshness | dbt Developer Hub - docs.getdbt.com
- Estimate and control costs Β |Β BigQuery Β |Β Google Cloud Documentation - docs.cloud.google.com
- Configure incremental models | dbt Developer Hub - docs.getdbt.com
- Model contracts | dbt Developer Hub - docs.getdbt.com
- unique_key | dbt Developer Hub - docs.getdbt.com
- About incremental strategy | dbt Developer Hub - docs.getdbt.com
- API reference - tableau.github.io
- samples/refresh.py - github.com
- Row-level security (RLS) with Power BI - Microsoft Fabric | Microsoft Learn - learn.microsoft.com
- Permission tokens needed to embed a Power BI app - Power BI | Microsoft Learn - learn.microsoft.com
- Data credentials Β |Β Data Studio Β |Β Google Cloud Documentation - docs.cloud.google.com
- Configure incremental refresh and real-time data for Power BI semantic models - Power BI | Microsoft Learn - learn.microsoft.com
- Add snapshots to your DAG | dbt Developer Hub - docs.getdbt.com
- Model versions | dbt Developer Hub - docs.getdbt.com
- Share and Collaborate on Power BI Reports and Dashboards - Power BI | Microsoft Learn - learn.microsoft.com
- Filter and Sort - tableau.github.io
Bring the domain, dependencies and receiving-team needs.
Describe the records, users, permitted actions, integrations, existing code and operating constraints. Agree base/custom boundaries, acceptance evidence, current quote and handover before implementation.
Discuss Data Analytics requirements β