Deployment Options

On this page, you will:

Understand the trade-offs between SaaS and self-hosted BI tools
Learn the true cost of self-hosting (infrastructure + engineering time)
Know when to choose cloud SaaS versus self-hosted deployment
Understand security and compliance considerations for each approach

Overview

BI tools typically offer two deployment models: cloud SaaS (vendor-hosted) and self-hosted (you manage infrastructure). Each has distinct trade-offs in cost, control, maintenance, and complexity.

This page helps you understand which deployment model fits your organisation's capabilities, budget, and requirements.

Deployment Models Explained

Cloud SaaS (Software as a Service)

The BI tool vendor hosts and manages the application. You sign up, connect your data warehouse, and start building dashboards. The vendor handles servers, updates, backups, and scaling.

Examples: - Tableau Online (vs Tableau Server) - Lightdash Cloud (vs Lightdash self-hosted) - Metabase Cloud (vs Metabase self-hosted) - Power BI Service (vs Power BI Report Server) - Looker (cloud-only, no self-hosted option) - Omni (cloud-only)

Self-Hosted (On-Premises or Your Cloud)

You deploy the BI tool to infrastructure you control — AWS, GCP, Azure, or on-premises servers. You're responsible for updates, backups, scaling, and maintenance.

Examples: - Lightdash (Docker, ECS, Kubernetes) - Metabase (Docker, ECS, Kubernetes) - Apache Superset (Docker, Kubernetes) - Tableau Server - Power BI Report Server

Hybrid Models

Some tools (Hex, Mode, dbt Cloud) are cloud-only with no self-hosted option. Others (Snowsight) are embedded in the platform and don't have separate deployment options.

The SaaS vs Self-Hosted Trade-Off

Factor	Cloud SaaS	Self-Hosted
Setup time	Minutes to hours	Hours to days
Infrastructure cost	Included in price	$20-100+/month (your cloud bill)
Maintenance	Vendor manages	You manage (4-8 hours/month)
Updates	Automatic	Manual (test, deploy, rollback plan)
Scalability	Automatic	Manual (resize instances, add nodes)
Data residency	Vendor's cloud	Your cloud/data centre
Customisation	Limited	Full control (within application limits)
Total cost	Higher monthly fee	Lower fee + infrastructure + eng time
Uptime SLA	Vendor SLA (99.9%+)	Your responsibility
Security	Vendor's security + your auth	Your full responsibility

Neither option is universally better — the right choice depends on your organisation's priorities and capabilities.

Cost Comparison: The Full Picture

Cloud SaaS Costs

Cloud SaaS pricing is straightforward: a monthly fee per user or flat rate.

Example: Lightdash Cloud - $2400/month flat rate (unlimited users) - No infrastructure costs - No engineering time for maintenance - Total: $2400/month

Example: Omni - $20-50/user/month - 10 users = $200-500/month - No infrastructure costs - No engineering time - Total: $200-500/month

Example: Tableau Online - $70/user/month (Creator), $42/user/month (Explorer) - 5 Creators + 20 Explorers = $350 + $840 = $1190/month - No infrastructure costs - No engineering time - Total: $1190/month

Self-Hosted Costs

Self-hosted costs are multi-layered: infrastructure + engineering time + potential downtime.

Example: Lightdash Self-Hosted

Infrastructure (AWS ECS + RDS): - ECS Fargate: 1 vCPU, 2GB RAM = ~$15/month - RDS PostgreSQL: db.t3.micro = ~$15/month - Application Load Balancer = ~$20/month - Subtotal: ~$50/month

Engineering Time: - Initial setup: 4-8 hours (one-time) - Monthly maintenance: 2-4 hours (updates, monitoring) - Quarterly upgrade: 4 hours - Average: ~4 hours/month

At $100/hour engineering cost: - Subtotal: ~$400/month in engineering time

Total Self-Hosted Cost: ~$450/month

This is cheaper than Lightdash Cloud ($2400/month) if you have fewer than 10-15 users but more expensive than you might expect from the "$0" open source price tag.

Engineering Time Is Real Cost

Self-hosted tools appear free, but engineering time has opportunity cost. Those 4 hours/month could be spent building pipelines, models, or dashboards instead of updating infrastructure.

The Break-Even Point

For most tools, self-hosted becomes cheaper than SaaS once you cross a certain team size:

Lightdash: - Self-hosted: ~$450/month (infrastructure + eng time) - Cloud: $2400/month (flat rate, unlimited users) - Break-even: Always self-host if cost-conscious (unless you value zero-maintenance)

Metabase: - Self-hosted: ~$400/month (infrastructure + eng time) - Cloud: $85/user/month - 5 users: $425/month (SaaS) vs $400/month (self-hosted) → roughly equivalent - 10 users: $850/month (SaaS) vs $400/month (self-hosted) → self-hosted saves $450/month - Break-even: ~5 users

Omni vs Lightdash Self-Hosted: - Omni Cloud: $60-150/month (3 users) - Lightdash Self-Hosted: ~$450/month - Break-even: Omni is cheaper until ~5-7 users (if you value low maintenance)

Infrastructure Requirements for Self-Hosted

Minimum Viable Setup

Most self-hosted BI tools require:

Application server (Docker container, ECS task, Kubernetes pod)
1-2 vCPUs, 2-4GB RAM
Cost: ~$15-30/month (AWS ECS Fargate, t3.small EC2)
Database (PostgreSQL or MySQL for metadata)
Stores dashboards, users, queries, cached results
db.t3.micro (1 vCPU, 1GB RAM) sufficient for small teams
Cost: ~$15-25/month (RDS)
Load balancer (for HTTPS termination)
Application Load Balancer (AWS ALB)
Cost: ~$20/month + data transfer
Object storage (optional, for cached query results)
S3 bucket
Cost: ~$1-5/month (minimal for BI tool caching)

Total: $50-80/month for basic self-hosted BI infrastructure.

Production-Grade Setup (Larger Teams)

For teams with 50+ users or high query volumes:

Auto-scaling application servers
2-4 ECS tasks or Kubernetes pods
Cost: ~$60-120/month
Database with read replicas
Primary + 1 read replica
Cost: ~$40-80/month
Redis for caching
Elasticache Redis (cache.t3.micro)
Cost: ~$15/month
CloudWatch monitoring and logs
Cost: ~$10-20/month

Total: $125-235/month for production-grade infrastructure.

Add 6-10 hours/month engineering time for a larger deployment.

Maintenance Overhead

Cloud SaaS: Vendor-Managed

With SaaS, the vendor handles: - ✅ Software updates and security patches - ✅ Database backups and disaster recovery - ✅ Scaling for increased usage - ✅ Monitoring and alerting - ✅ Uptime SLAs (99.9%+)

Your responsibilities: - User management (SSO, roles, permissions) - Dashboard creation and maintenance - Query performance optimisation (warehouse-side)

Time investment: ~1-2 hours/month

Self-Hosted: You Manage Everything

With self-hosted, you're responsible for:

Monthly (2-4 hours): - Monitoring application health and logs - Reviewing performance metrics - User support (login issues, access problems)

Quarterly (4-6 hours): - Upgrading to new releases (test, deploy, verify) - Database maintenance (vacuum, reindex if needed) - Reviewing and rotating secrets/credentials

Annually (8-12 hours): - Infrastructure review and right-sizing - Disaster recovery testing - Security audit and updates

Ad-hoc (varies): - Scaling for increased load - Debugging issues and downtime - Restoring from backups

Total: ~6-10 hours/month average

Reduce Maintenance with Automation

Use Terraform for infrastructure-as-code, automated backups (RDS), and container health checks (ECS). This reduces manual work but still requires initial setup and monitoring.

Security and Compliance

Data Residency and Access

Cloud SaaS: - BI tool hosted in vendor's cloud (typically AWS, GCP, or Azure) - Vendor has access to metadata (dashboard definitions, user queries, cached results) - Your data warehouse data is not copied to the BI tool (it's queried on-demand) - Most tools cache query results (aggregated, not raw data)

Self-Hosted: - BI tool hosted in your cloud or data centre - You control all data and metadata - No third-party access (unless you grant it) - Full control over data residency and compliance

How BI Tools Access Your Data

Both SaaS and self-hosted tools query your warehouse on-demand. They don't copy your raw data. Some tools cache query results (e.g., "sum of sales by region") to improve performance, but this is aggregated data, not raw customer records.

Compliance Considerations

Requirement	Cloud SaaS	Self-Hosted
GDPR	Vendor must be GDPR-compliant	Your responsibility
SOC 2	Check vendor's SOC 2 report	Your audit scope
HIPAA	Vendor must sign BAA	Your responsibility
Data residency (EU-only)	Vendor must offer EU hosting	Deploy in EU region
Customer data isolation	Multi-tenant (shared infra)	Single-tenant (your infra)

For regulated industries (healthcare, finance): Self-hosted is often required or strongly preferred for compliance and audit reasons.

For most teams: SaaS is compliant enough. Check vendor's security page and SOC 2 report.

Authentication and Access Control

Cloud SaaS: - Managed SSO integrations (Okta, Azure AD, Google Workspace) - Vendor-managed user authentication - You control authorisation (who sees which dashboards)

Self-Hosted: - You configure SSO (requires SAML setup, testing) - You manage authentication infrastructure - Full control over access policies

When to Choose Cloud SaaS

Choose cloud SaaS if:

✅ Small engineering team — no dedicated DevOps/infrastructure resources
✅ Limited time — you'd rather build dashboards than manage infrastructure
✅ Budget allows — $200-2400/month for BI tool is acceptable
✅ Compliance permits — no strict data residency requirements
✅ Prefer simplicity — value vendor-managed updates and zero maintenance
✅ Fast time-to-value — need BI running in hours, not days

Best SaaS options: - Omni ($60-300/month for small teams, dbt-native) - Power BI ($100-500/month for medium teams, Microsoft ecosystem) - Lightdash Cloud ($2400/month for larger teams, unlimited users)

When to Choose Self-Hosted

Choose self-hosted if:

✅ Budget-constrained — willing to invest engineering time to save money
✅ DevOps expertise — team already manages self-hosted services (Prefect, Airflow)
✅ Data residency required — compliance mandates data stays in specific regions
✅ Customisation needed — want full control over deployment and configuration
✅ Large team — 20+ users where SaaS per-user costs are prohibitive
✅ Infrastructure-as-code culture — everything Terraform-managed already

Best self-hosted options: - Lightdash (dbt-native, Terraform-friendly, ~$50/month infra) - Metabase (user-friendly, mature, ~$50/month infra) - Superset (advanced viz, Apache project, ~$60/month infra)

Hybrid Approach: Start SaaS, Move to Self-Hosted Later

Some teams start with SaaS for speed, then migrate to self-hosted as they scale:

Phase 1: SaaS (Months 0-6) - Fast setup (hours, not days) - Learn what you actually need - Get stakeholder buy-in with quick wins - Evaluate usage patterns

Phase 2: Migrate to Self-Hosted (Month 6+) - Once you have 10+ users or predictable usage - When engineering team has capacity - When budget requires cost optimisation - Export dashboards and redeploy to self-hosted

Works well for: - Omni → Lightdash self-hosted (both dbt-native, similar features) - Metabase Cloud → Metabase self-hosted (same product, different hosting)

Harder migrations: - Tableau → Lightdash (different paradigms, rebuild dashboards) - Power BI → Superset (different data models, significant rework)

Decision Matrix

Use this matrix to choose your deployment approach:

	Cloud SaaS	Self-Hosted
Team size	< 10 people	5+ with DevOps skills
Budget	$200-2400/month	$50-100/month (infra only)
Engineering time available	Limited (< 5 hours/month)	6-10 hours/month
Compliance	Standard (GDPR, SOC 2)	Strict (HIPAA, PCI-DSS, data residency)
Infrastructure expertise	Limited	Strong (Docker, ECS, Terraform)
Time to value	Immediate (hours)	Days to weeks
Vendor lock-in concern	Low (can export dashboards)	None (you control everything)

Cost Projection Example

Let's model Lightdash costs for a growing team over 2 years:

Team Size	Cloud SaaS (Lightdash Cloud)	Self-Hosted (Lightdash)
5 users	$2400/month	$450/month (infra + 4 hrs eng time)
10 users	$2400/month	$450/month (same)
20 users	$2400/month	$550/month (upgraded infra, 6 hrs eng time)
50 users	$2400/month	$700/month (production setup, 8 hrs eng time)

Break-even: Never (self-hosted always cheaper for Lightdash due to flat-rate SaaS pricing)

But if engineering time is $150/hour (senior engineer): - Self-hosted at 5 users: $450 + ($150 × 4 hrs) = $1050/month - Break-even: ~15-20 users when factoring expensive engineering time

Summary

You've learned about deployment options:

Cloud SaaS costs more but requires zero infrastructure management (ideal for small teams)
Self-hosted costs less in subscription fees but requires infrastructure ($50-100/month) and engineering time (6-10 hours/month)
True cost of self-hosting includes infrastructure, engineering time (opportunity cost), and maintenance overhead
Security and compliance may require self-hosted for regulated industries or data residency requirements
Hybrid approach (start SaaS, migrate to self-hosted) works well for cost optimisation as you scale

For most teams following this documentation (small, budget-conscious, DevOps-capable), self-hosted Lightdash offers the best value. For teams wanting zero maintenance, Omni (cloud SaaS, dbt-native) is worth the premium.

What's Next

Now that you understand deployment trade-offs, start with Snowflake's built-in dashboarding for a quick win before deploying a dedicated BI tool.

Continue to Snowflake Snowsight →