Data catalog trong thực tế: implementation guide và best practices

"Chúng tôi đã mua Alation với giá $80K/năm. Nhưng sau 6 tháng, chỉ có 15% datasets được documented và không ai dùng."

Đây là câu chuyện chúng tôi nghe rất nhiều. Các doanh nghiệp đầu tư vào data catalog tools - Atlan, Alation, Collibra, hoặc DataHub - nhưng adoption rate thấp đến đáng thất vọng.

Root cause: Họ treat data catalog như một technology project, không phải organizational change initiative.

Trong bài trước về data catalog, chúng tôi đã giải thích data catalog là gì và tại sao nó quan trọng. Bài viết này đi sâu vào "how": Cách triển khai data catalog thành công trong thực tế.

Tại Carptech, chúng tôi đã giúp 8 doanh nghiệp triển khai data catalog với average adoption rate 70%+ trong 3-6 tháng. Trong bài này, chúng tôi chia sẻ playbook đã proven:

• Pre-implementation checklist (những gì cần chuẩn bị trước)
• 8-week implementation roadmap (từng phase chi tiết)
• Adoption strategies (làm sao để users thực sự dùng)
• Common pitfalls và cách tránh
• Tool comparison (Atlan vs Alation vs DataHub vs Collibra)
• Measuring success (metrics để track)

Recap: Data catalog là gì?

Quick recap (chi tiết xem bài trước):

Data catalog = "Google cho dữ liệu" của công ty bạn.

Key capabilities:

• Search: Tìm datasets by keywords, tags, descriptions
• Understand: See schemas, sample data, lineage, quality scores
• Access: Request permissions if you don't have
• Trust: See certifications, ownership, freshness
• Collaborate: Comments, ratings, Q&A

Problem it solves:

Before catalog:

• Analyst mất 2-3 ngày để tìm "customer churn data"
• Không biết ai là owner để hỏi
• Không biết data quality có OK không
• Duplicate work vì không biết người khác đã làm gì

After catalog:

• Search "customer churn" → 3 relevant datasets hiện ra trong 30 giây
• See owner, description, lineage, quality score
• Request access with one click
• See queries người khác đã viết

Result: 80-90% reduction trong time to find data.

Pre-implementation checklist

Trước khi bắt đầu, đảm bảo bạn có:

1. Executive sponsorship

Critical success factor #1.

Why: Data catalog is organizational change, not just IT project. Cần executive để:

• Allocate budget ($20K-$100K depending on tool)
• Mandate participation (data stewards contribute time)
• Make it part of company culture

Who: Ideally CDO (Chief Data Officer), or CTO, or COO.

Commitment: Monthly steering meeting, champion the initiative publicly.

Red flag: Nếu executive chỉ "approve budget" nhưng không actively involved → high risk of failure.

2. Clear scope definition

Mistake: "Let's catalog everything!" → Overwhelmed, never finish.

Better: Start small, expand gradually.

Decision to make:

Which data sources?

• Start with 3-5 most critical sources (e.g., Snowflake warehouse, Salesforce CRM, Google Analytics)
• Expand sau khi stable

Which datasets within those sources?

• 80/20 rule: Start với 20% most-used datasets that drive 80% queries
• How to identify: Query logs, analytics tool usage stats

Which teams?

• Pilot team: 1-2 teams (e.g., Marketing analytics team, Product team)
• Rollout: Expand to others sau khi proven

Timeline:

• Pilot: 8-12 tuần
• Rollout: 3-6 tháng

3. Tool selection completed

Choose tool trước khi bắt đầu implementation. (More on tool comparison later)

Quick comparison:

4. Dedicated resources

Data catalog is not "side project." Cần dedicated people:

Core team:

• 1 Data Governance Lead (50-100% time): Project manage, define standards
• 2-3 Data Stewards (20-30% time each): Document datasets, review quality
• 1 Data Engineer (30-50% time): Technical setup, integrations

Extended team:

• Domain owners (10% time each): Approve definitions, resolve conflicts

Total effort: Approximately 1.5-2 FTE for 8-12 week pilot.

5. Success metrics defined

What does "success" look like?

Quantitative:

• Adoption: 70%+ users log into catalog monthly
• Coverage: 80%+ critical datasets documented
• Time to find data: < 15 minutes (vs 2-3 days before)
• Self-service rate: 50%+ data questions answered via catalog (không cần Slack/email data team)

Qualitative:

• User satisfaction: NPS score > 50
• Data stewards feel empowered
• Executives reference catalog in meetings

8-week implementation roadmap

Week 1-2: Technical setup

Goal: Get catalog platform up and running.

Activities:

1. Deploy platform

Cloud-hosted (recommended):

• Atlan, Alation: SaaS, deploy in minutes
• Sign contract, create workspace, done

Self-hosted:

• DataHub: Deploy on Kubernetes or AWS ECS
• Requires infrastructure setup (1-2 days)

2. Connect data sources

Integrations to set up:

Data warehouse:

• Snowflake, BigQuery, Redshift
• Automated metadata harvesting: tables, columns, schemas, query logs

BI tools:

• Looker, Tableau, Power BI
• Harvest dashboards, reports, their underlying datasets

Databases:

• PostgreSQL, MySQL, MongoDB
• Optional cho pilot (có thể add sau)

Data pipeline tools:

• dbt: Sync dbt docs → catalog
• Airflow: Harvest DAGs, task descriptions

Example: Snowflake integration in Atlan

# Atlan connector config
connector: snowflake
config:
  account: xy12345.us-east-1
  warehouse: COMPUTE_WH
  database: ANALYTICS
  schema: PUBLIC
  username: atlan_reader
  password: <secret>
  role: READER_ROLE
sync_schedule: "0 2 * * *"  # Daily at 2am
metadata_to_harvest:
  - tables
  - columns
  - schemas
  - lineage
  - query_logs  # For popularity metrics

3. Set up SSO/authentication

• Integrate với company SSO (Okta, Google Workspace, Azure AD)
• Role-based access: Admin, Steward, Viewer

4. Configure basic settings

• Branding (company logo)
• Email notifications
• Slack integration (optional)

Deliverable: Catalog platform operational, basic metadata harvested.

Week 3-4: Content population

Goal: Make catalog useful by adding business context.

Activities:

1. Automated metadata harvesting

Run connectors to crawl:

• All tables/views
• Column names, types
• Sample data (first 100 rows)
• Basic statistics (row counts, update frequency)

Result: Technical metadata populated automatically.

2. Prioritize datasets to document

Use 80/20 rule:

• Top 20% most-queried tables
• Critical business datasets (customer, revenue, product)
• Tables used in executive dashboards

How to identify:

• Query logs từ warehouse
• BI tool usage statistics
• Survey teams: "What data do you use most?"

Example prioritization:

Start với top 20-30 datasets.

3. Add business context

For each prioritized dataset, add:

Table-level:

• Description: What is this table? (plain English)
  • Good: "Contains all customer orders since 2020, updated nightly via Fivetran from Shopify."
  • Bad: "Orders table" (too vague)
• Owner: Who maintains this? (tag actual person)
• Business glossary terms: Link to definitions (e.g., "Monthly Active User")
• Use cases: What is this used for? ("Marketing attribution model", "Executive revenue dashboard")
• Freshness: How often updated? ("Daily at 3am")
• Tags: Categorize (e.g., "Revenue", "PII", "Certified")

Column-level:

• Description: What does customer_ltv mean?
• Data type: Currency (VND), Integer, String
• Constraints: Not null, unique, foreign key
• Sensitive: Is this PII? (email, phone)

Example: dim_customers table documentation

# dim_customers

## Description
Dimension table containing all customers with at least one order.
Updated nightly at 3am UTC via Fivetran from Shopify.
Deduplicated by email address.

## Owner
- Data Owner: @nguyen.van.a (Head of Customer Success)
- Data Steward: @tran.thi.b (Senior Analytics Engineer)

## Business Use Cases
- Customer 360° analytics
- Churn prediction model
- Marketing segmentation

## Key Columns
- `customer_id` (INTEGER, PRIMARY KEY): Unique customer identifier
- `email` (STRING, PII): Customer email address
- `customer_ltv` (DECIMAL): Customer lifetime value in VND, calculated as sum of all order totals
- `segment` (STRING): Customer segment (VIP, Regular, New), updated monthly

## Data Quality
- Completeness: 98% (2% missing email)
- Freshness: Updated daily
- Last quality check: 2025-07-25

## Related Assets
- Dashboard: [Customer Analytics Dashboard](link)
- dbt model: [customer_dimension.sql](link)
- Upstream: Shopify `customers` table

4. Create business glossary

Define business terms used across company:

Example glossary:

Why important: Ensures everyone uses same definition. No more "wait, how do you define MAU?"

Deliverable: Top 20-30 datasets fully documented, business glossary started.

Week 5-6: Governance setup

Goal: Establish processes để maintain quality and trust.

Activities:

1. Define data steward roles

Assign stewards for each domain:

• Customer domain steward: Owns customer-related datasets
• Product domain steward: Owns product catalog, inventory
• Marketing domain steward: Owns campaign, attribution data

Responsibilities:

• Document datasets in their domain
• Review and approve changes
• Respond to questions from users
• Monitor data quality

Time commitment: 10-20% of their time (2-4 hours/week).

2. Create certification workflow

Goal: Distinguish "trusted" datasets từ "unverified" ones.

Certification levels:

Certified (Gold):

• Fully documented
• Data quality tests passing
• Owner assigned và responsive
• Approved by governance lead

Verified (Silver):

• Basic documentation
• Some quality checks
• Owner assigned

Uncertified (Bronze):

• Auto-harvested, no manual review
• Use with caution

Workflow:

1. Data steward documents dataset
2. Submits for certification review
3. Governance lead reviews (checklist)
4. If pass → Badge "Certified"
5. If fail → Feedback, steward fixes, resubmit

Benefits:

• Users know which datasets to trust
• Incentivizes stewards to complete documentation
• Creates quality bar

3. Set up data quality scores

Integrate với data quality tools:

• Great Expectations
• dbt tests
• Soda
• Monte Carlo

Display quality metrics trong catalog:

• Freshness: "Updated 2 hours ago" (Green) vs "Last updated 5 days ago" (Red)
• Completeness: "98% of required fields populated"
• Accuracy: "0 schema drift detected"
• Test results: "15/15 tests passing"

4. Access request workflows

Enable self-service access:

User flow:

1. User searches, finds dataset
2. Clicks "Request Access"
3. Fills form: "Why do you need this?" + "How long?"
4. Request routed to data owner
5. Owner approves/denies (with notification)
6. If approved: Access granted automatically (via API to Snowflake/BigQuery)

Benefits:

• Faster than Slack/email (approval in hours vs days)
• Audit trail (who has access to what, why)
• Data owners maintain control

Deliverable: Governance processes defined, certification started, access workflows operational.

Week 7-8: Adoption & training

Goal: Get users to actually use the catalog.

Activities:

1. Train data stewards (content owners)

2-hour workshop:

• How to document datasets
• Best practices (good vs bad descriptions)
• How to use catalog features (lineage, queries, tags)
• Certification process
• Q&A

Materials:

• Step-by-step guide (PDF)
• Video tutorials (5-10 min each)
• Slack channel for support

2. Train end users (data consumers)

1-hour workshop per team:

• How to search for data
• Understanding catalog entries (what do these fields mean?)
• How to request access
• How to provide feedback (comments, ratings)

Format:

• Live demo (15 min)
• Hands-on exercise (30 min): "Find dataset for X use case"
• Q&A (15 min)

3. Create documentation

User guide:

• Search tips: "Use quotes for exact match", "Filter by owner/tag"
• FAQ: "How do I know if data is trustworthy?" → Look for "Certified" badge
• Video library: Screen recordings of common tasks

Quick reference card:

• 1-page cheat sheet
• Print and put next to monitors

4. Onboarding for new employees

Make catalog part of onboarding:

• Day 1: Create catalog account
• Week 1: Complete "Data Catalog 101" training
• Assign buddy (data champion) to answer questions

5. Launch communication

Announcement email from executive sponsor:

Subject: Introducing Our Data Catalog - Find Data in Seconds, Not Days

Team,

I'm excited to announce the launch of our Data Catalog (link).

Finding data has been a pain point for too long. With the catalog, you can:
- Search and discover datasets in seconds
- See who owns data and how to access it
- Trust data with quality scores and certifications

This is a game-changer for becoming more data-driven.

I encourage everyone to:
1. Attend training sessions (schedule)
2. Explore the catalog (start with "Most Popular" datasets)
3. Provide feedback to help us improve

Thank you to the data team for making this happen!

[Executive Name]

6. Gamification & incentives

Encourage participation:

Leaderboard:

• Most documentation contributed
• Most helpful comments
• Most active users

Recognition:

• "Data Champion of the Month" award
• Shoutout in all-hands meeting

Prizes:

• Small rewards (gift cards, swag) for top contributors

Deliverable: Teams trained, documentation available, catalog officially launched.

Adoption strategies

Training is not enough. These strategies drive sustained adoption:

Strategy 1: Make it mandatory

Policy:

• All new datasets must be cataloged before use
• Include in code review checklist: "Is this dataset in catalog?"
• dbt models: Auto-sync to catalog (docs → catalog integration)

Enforcement:

• Data platform team gates access: "Need access to X? Check catalog first."

Strategy 2: Embed in workflows

Meet users where they are:

Looker integration:

• In Looker dashboard, show "View in Catalog" link next to table name
• Bi-directional: Catalog shows which Looker dashboards use this table

dbt integration:

• dbt docs → Catalog (automated sync)
• Developers document in dbt YAML → automatically in catalog

Slack integration:

• /catalog search customer_churn → Results in Slack
• Notifications: "Dataset you're watching was updated"

BI tool tooltips:

• When hovering over field in dashboard, show description from catalog

Strategy 3: Show quick wins

Celebrate and share success stories:

Example:

Before catalog: Analyst spent 3 days trying to find "customer segmentation data", asking 5 people, still didn't find it.

After catalog: Searched "customer segment", found dataset in 30 seconds, read description, requested access, got approved in 2 hours. Started analysis same day.

Time saved: 3 days → 30 seconds.

Share these stories:

• In all-hands meetings
• In Slack #data channel
• In monthly data team newsletter

Quantify impact:

• "Catalog has saved 200 hours of data discovery time this quarter"
• "50% reduction in ad-hoc data requests to data team"

Strategy 4: Executive endorsement

CEO/CTO uses catalog publicly:

• In board meeting: "Let me check the catalog for customer retention metrics..."
• In Slack: "Found this dataset in catalog: [link]"
• In all-hands: "Great example of using catalog to self-serve"

Message: "If the CEO uses it, it must be important."

Strategy 5: Continuous improvement

Listen to feedback:

Monthly survey:

• "Did you use catalog this month?" Yes/No
• "What would make it more useful?" (open text)
• NPS: "How likely would you recommend catalog to colleague?"

Office hours:

• Data team holds weekly 30-min "Catalog Office Hours"
• Users can drop in with questions

Iterate:

• Address top pain points monthly
• Add requested features
• Improve documentation based on FAQ

Strategy 6: Deprecate old ways

Don't let users fall back to old habits:

Old way: Slack data team

• User: "Where is customer churn data?"
• Data team response: "Please check catalog first: [link]. If not found, let me know."

Old way: Shared spreadsheet of "data dictionary"

• Archive it
• Redirect to catalog

Message: "Catalog is the new way. We won't maintain old ways."

Common pitfalls & solutions

Pitfall 1: Data stewards too busy

Problem: Assigned stewards "in addition to day job" → they don't have time → documentation doesn't happen.

Solution:

• Allocate dedicated time: 10-20% formal time allocation, in their OKRs
• Reduce other work: Can't just add more work without removing something
• Recognition: Make stewardship visible (part of performance review)

Pitfall 2: Stale metadata

Problem: Catalog documented at launch, but 6 months later, it's outdated.

Solution:

• Automated sync: Re-harvest metadata daily/weekly
• Quarterly review: Stewards review their domains every quarter
• Alerts: Notify stewards when datasets haven't been updated in 6+ months
• Metrics: Track "% datasets updated in last 90 days" → dashboard for governance lead

Pitfall 3: Low adoption

Problem: Catalog exists, but users still Slack data team or ask around.

Solution:

• Make it mandatory: Access requests must go through catalog
• Remove alternatives: Deprecate old wikis/spreadsheets
• Embed in tools: Integrate với Looker, dbt, Slack
• Show value: Track and share time savings

Pitfall 4: Too much governance

Problem: Certification workflow so strict that nothing gets certified → users don't trust catalog.

Solution:

• Start lightweight: Simple certification criteria (has description + owner)
• Iterate: Add more criteria gradually
• Self-service: Let stewards certify their own domains (with audit)

Pitfall 5: No executive buy-in

Problem: Executive "approves budget" but doesn't champion → middle management doesn't prioritize.

Solution:

• Re-engage executive: Show business case (time savings, ROI)
• Quick wins: Demonstrate value with pilot team success
• Escalate blockers: If stewards not allocated time, executive must intervene

Pitfall 6: Wrong tool choice

Problem: Bought expensive enterprise tool for small team → overkill, hard to use.

Solution:

• Reassess: Can you downgrade or switch?
• Open-source alternative: DataHub might be better fit
• Focus on adoption, not features: A simple catalog that's used > fancy catalog that's not

Tool comparison: Atlan vs Alation vs DataHub vs Collibra

Quick decision matrix

Use this to choose:

Detailed comparison

DataHub (Open-source)

Pros:

• ✅ Free (open-source)
• ✅ Modern architecture (Python/Java)
• ✅ Active community (LinkedIn, Airbnb use it)
• ✅ Extensible (APIs, custom integrations)
• ✅ Good lineage visualization

Cons:

• ❌ Requires infrastructure management (Kubernetes)
• ❌ Less polished UI than commercial options
• ❌ Limited support (community-based)
• ❌ Steeper learning curve

Best for:

• Technical teams comfortable với self-hosting
• Startups / small companies với limited budget
• Companies wanting full control and customization

Pricing: Free (but infra costs: ~$500-2K/month AWS/GCP)

Atlan

Pros:

• ✅ Modern, intuitive UI (Figma-like collaboration)
• ✅ Easy setup (SaaS, minutes to deploy)
• ✅ Good integrations (Snowflake, dbt, Looker, Tableau)
• ✅ Collaborative features (comments, @mentions, Slack notifications)
• ✅ Fast adoption (users love the UX)
• ✅ Good support

Cons:

• ❌ Younger product (less mature than Alation/Collibra)
• ❌ Limited governance workflows (improving)
• ❌ Smaller ecosystem

Best for:

• Mid-market companies (50-500 employees)
• Teams prioritizing user adoption
• Modern data stacks (Snowflake, dbt, Looker)

Pricing: $20K-$50K/year (depends on users and data sources)

Alation

Pros:

• ✅ Established, mature product (since 2012)
• ✅ Powerful search (Google-like relevance)
• ✅ Strong lineage capabilities
• ✅ Good query analytics (shows popular queries)
• ✅ Proven tại enterprise (Cisco, eBay, Nasdaq)
• ✅ Comprehensive connectors

Cons:

• ❌ UI less modern (feels dated)
• ❌ Expensive
• ❌ Setup can be complex
• ❌ Steeper learning curve

Best for:

• Large enterprises (500-5000 employees)
• Companies with diverse data landscape (many sources)
• Organizations where search discoverability is top priority

Pricing: $50K-$150K/year

Collibra

Pros:

• ✅ Comprehensive governance platform (not just catalog)
• ✅ Workflow engine (complex approval processes)
• ✅ Data quality management integrated
• ✅ Regulatory compliance features (GDPR, CCPA)
• ✅ Enterprise-grade security
• ✅ Strong professional services

Cons:

• ❌ Very expensive
• ❌ Complex (months to implement)
• ❌ Overkill for most companies
• ❌ Heavy, slower to adopt

Best for:

• Large enterprises (5000+ employees)
• Highly regulated industries (banking, healthcare, insurance)
• Organizations needing full data governance suite

Pricing: $100K-$300K+/year

Carptech recommendation:

• Starting out, budget-conscious: DataHub
• Most companies: Atlan (best balance of usability and features)
• Enterprise with complex needs: Alation
• Highly regulated, governance-heavy: Collibra

Measuring success

Track these metrics to measure catalog effectiveness:

Adoption metrics

1. Active users

• Metric: % employees who logged into catalog in last 30 days
• Target: 70%+ (after 6 months)

2. Search activity

• Metric: Number of searches per week
• Target: Growing trend (plateau is OK once mature)

3. Documentation coverage

• Metric: % critical datasets with description + owner
• Target: 80%+

4. Certification rate

• Metric: % datasets certified (gold/silver)
• Target: 50%+ for critical datasets

Impact metrics

5. Time to find data

• Metric: Average time from "need data" to "found data"
• Measurement: Survey users monthly
• Target: < 15 minutes (vs 1-3 days before)

6. Self-service rate

• Metric: % data questions answered via catalog (không cần Slack/email)
• Target: 50%+

7. Ad-hoc requests reduction

• Metric: Số ad-hoc data requests to data team
• Target: 30-50% reduction

8. Data quality improvement

• Metric: % datasets with quality scores available
• Target: 60%+

User satisfaction

9. NPS (Net Promoter Score)

• Question: "How likely would you recommend catalog to colleague?" (0-10)
• Target: NPS > 50 (good), > 70 (excellent)

10. Qualitative feedback

• Monthly survey: "What do you like?" "What needs improvement?"
• Theme analysis: Common requests → prioritize features

Dashboard example

Monthly Governance Dashboard:

Case study: Enterprise với 1000+ datasets

Company: E-commerce marketplace, 800 employees, 1,200 datasets in Snowflake.

Problem:

• Data analysts spend 30-40% time just finding data
• Same datasets documented trong wikis, Confluence, Slack, Google Docs (scattered)
• No one knows who owns what
• Duplicate analysis vì không biết người khác đã làm gì

Solution: Atlan implementation

Month 1-2: Setup + pilot

• Deployed Atlan SaaS
• Integrated Snowflake, dbt, Looker
• Pilot với Marketing Analytics team (50 datasets)
• Documented top 50 với stewards

Month 3-6: Rollout

• Trained 8 domain stewards (Customer, Product, Marketing, Finance, Ops)
• Documented 400 critical datasets (top 35%)
• Certification workflow established
• Training sessions cho 200 users

Month 7-12: Maturity

• Expanded to remaining datasets
• Quarterly review cycles
• Integrated access requests with Snowflake RBAC (automated)
• Embedded catalog links trong Looker

Results sau 12 tháng:

Adoption:

• 85% active users (680 / 800 employees logged in monthly)
• 600+ searches per week
• 80% critical datasets documented
• 55% datasets certified (gold/silver)

Impact:

• Time to find data: 2 days → 10 minutes (95% reduction)
• Ad-hoc requests giảm 60%: Data team từ 80 requests/month → 30 requests/month
• Analyst productivity: Surveys show analysts feel 40% more productive

User feedback:

• NPS: 72 (excellent)
• Top comment: "Game changer. Can't imagine working without catalog now."

ROI:

• Investment: $40K Atlan + $80K labor (stewards time) = $120K
• Benefit: 200 employees × 2 hours saved/week × 50 weeks × $50/hour = $1M/year productivity gain
• ROI: 733% in year 1

Success factors:

1. Executive sponsorship: CTO championed, monthly steering meetings
2. Dedicated stewards: 8 stewards với formal 20% time allocation
3. Phased rollout: Pilot → expand, không big bang
4. Embedded in workflows: Looker integration, dbt sync
5. Continuous improvement: Monthly surveys, quarterly reviews

Kết luận

Data catalog triển khai thành công không phải vì tool đắt tiền hay fancy features. Nó thành công vì:

1. Executive sponsorship và organizational commitment
2. Dedicated people (stewards) với time allocation rõ ràng
3. Phased approach: Start small, expand gradually
4. Focus on adoption: Training, embedding in workflows, making it mandatory
5. Continuous improvement: Listen to users, iterate

8-week roadmap recap:

• Week 1-2: Technical setup
• Week 3-4: Content population
• Week 5-6: Governance setup
• Week 7-8: Adoption & training
• Month 3-6: Rollout
• Month 6+: Maturity & expansion

Key metrics to track:

• Active users (target: 70%+)
• Documentation coverage (target: 80%+)
• Time to find data (target: < 15 min)
• User satisfaction (NPS > 50)

Nếu bạn đang consider triển khai data catalog, đừng chỉ treat nó như "mua tool". Treat it như organizational change initiative với proper planning, resources, và ongoing commitment.

Muốn được tư vấn về data catalog implementation?

Tại Carptech, chúng tôi đã giúp 8 doanh nghiệp triển khai data catalog thành công với adoption rate 70%+. Đặt lịch tư vấn miễn phí 60 phút để chúng tôi đánh giá hiện trạng và đề xuất roadmap phù hợp.

Để tìm hiểu thêm về data governance và metadata management, đọc:

• Data catalog: democratizing data discovery với metadata management
• Data governance 101: framework cho doanh nghiệp Việt Nam
• Data lineage: traceability từ source đến dashboard