Aureus Global Systems

Data Platform Readiness
Assessment

20 questions · 5 sections · ~5 minutes

0/20 answered

Section 1

Data Infrastructure

How reliable are your data pipelines?

Pipelines fail regularly, manual fixes weekly

Mostly stable, but incidents happen monthly

Generally reliable with some edge case failures

Highly reliable, automated alerting in place

99.9%+ uptime, self-healing pipelines, full observability

How fresh is the data your business teams consume?

Data is days or weeks stale

Daily batch — T+1 is typical

Multiple batches per day (every few hours)

Near real-time (under 30 minutes)

Real-time streaming with sub-minute latency

Can your platform handle 3× your current data volume today?

We'd need weeks of work to scale

Possible but it would require significant manual effort

We could scale with some planned work

Largely elastic, would handle it with minor changes

Fully elastic — auto-scales without intervention

How well do you govern cloud infrastructure costs?

We have little visibility into what drives our bills

We review costs monthly but lack granular control

We have tagging and basic budgets in place

We actively optimize with rightsizing and scheduling

Full FinOps practice — cost per pipeline, per team, per workload

How quickly can you recover from a data incident?

Recovery is manual and takes days

We can recover but it takes many hours

We have runbooks, recovery takes 1-4 hours

Automated recovery for most scenarios, under 1 hour

Tested DR procedures, recovery under 15 minutes

Section 2

Data Quality & Governance

Do you have a searchable catalog of your data assets?

No catalog — finding data means asking colleagues

Informal documentation (Confluence/Notion) that's usually outdated

Partial catalog covering core datasets

Comprehensive catalog with ownership and descriptions

Full catalog with lineage, quality scores, and active governance

Do you have automated data quality monitoring?

Quality issues are discovered by end users complaining

Some manual checks run occasionally

Basic automated checks on critical pipelines

Automated checks across most pipelines with alerting

Full DQ framework — expectations, anomaly detection, SLA tracking

Can you trace any metric back to its source data, end to end?

No — lineage is unknown for most metrics

We can trace some critical metrics manually

Partial lineage for key datasets

Most data assets have documented lineage

Full automated lineage — click any metric, see the full chain

Is data access governed by role, with audit trails?

Access is largely open or managed ad hoc

Basic access controls, no consistent audit trail

Role-based access on sensitive datasets

RBAC across the platform with regular access reviews

Fine-grained access control, full audit trail, compliance-ready

Q10

Is there one agreed definition for key business entities?

Every team defines KPIs differently — constant debates

Informal alignment exists but it breaks down regularly

Core KPIs are documented and mostly agreed upon

Certified metrics with a governance process

Formal data contract layer — single source of truth for all entities

Section 3

Analytics & Business Intelligence

Q11

Can business users get the data they need without filing an IT ticket?

Every data request goes through IT — queue is weeks long

Some users can self-serve on limited datasets

Most common requests are self-serve, edge cases go to IT

Broad self-service with a semantic layer and governed datasets

Full self-service — any user can safely query governed data in minutes

Q12

Does every team use the same definition of your core business metrics?

Metrics are defined differently in every team's spreadsheets

Alignment exists informally but breaks down regularly

Core metrics are documented and mostly agreed upon

Certified metrics with a clear owner and governance process

Contractual metric definitions — disputes are resolved in minutes, not meetings

Q13

How long does it take to deliver a new business report from scratch?

Weeks — it requires significant engineering work

About a week including reviews and sign-off

2-3 days for a standard report

A few hours for most requests

Self-serve — business teams build their own in minutes

Q14

What fraction of your decision-makers actively use data dashboards weekly?

Under 20% — most decisions are made on gut or anecdotes

Around 40% — dashboards exist but aren't trusted or habitual

Around 60% — regular users, occasional trust issues

Around 80% — dashboards are the default for most decisions

Over 90% — data-driven decisions are the cultural norm

Section 4

AI & ML Readiness

Q15

Is your historical data clean and complete enough to train production ML models?

Our data has significant gaps, inconsistencies, and quality issues

Core data is okay but would need significant prep work

Most data is usable with moderate cleaning effort

Data is largely model-ready with minor preprocessing

Clean, versioned, documented datasets ready for training

Q16

Do you have a consistent process for building and managing ML features?

Feature engineering is ad hoc — each data scientist does it differently

Some shared notebooks/scripts but no formal process

Informal standards, features reused within teams

Feature store with some shared features across teams

Mature feature platform — discoverable, versioned, monitored features

Q17

Do you have infrastructure to train, serve, and monitor ML models in production?

No ML infrastructure — models live in notebooks

We can train models but deployment is a manual process

Some models in production, deployment is partly automated

ML platform with automated training, deployment, and basic monitoring

Full MLOps — automated retraining, drift detection, model governance

Q18

Has your organisation shipped any ML model to production in the past 12 months?

No — ML is still in research/POC stage

One or two pilots that never fully productionised

A few models in production, maintained manually

Multiple production models with proper monitoring

ML in production is routine — models ship regularly and reliably

Section 5

Team & Process

Q19

How structured is your data team?

No dedicated data team — engineers do data work as a side task

One or two data generalists, no clear roles

Small dedicated team (data engineers + analysts), limited leadership

Structured team with data engineering, analytics, and a data manager

Full data organisation — engineering, analytics, science, and an executive sponsor

Q20

What portion of your data team's time goes to building new things vs. fixing existing ones?

Over 70% is firefighting and maintenance

About 50/50 — half the week is reactive

Around 60% building, 40% maintaining

Around 75% building, 25% maintaining

Over 85% building — the platform largely runs itself