Saltar a contenido

Cost Modeling Tco

Análisis de costos del MVP comparado con Camunda 8 SaaS, Camunda 8 Self-Managed, Temporal Cloud, y self-hosted Temporal. Para small/medium/large scale: MVP self-hosted ~5-10x menor que Camunda 8 SaaS, ~2-3x menor que self-hosted comparable. Pricing models sugeridos (per-instance, per-throughput, per-tenant). Break-even analysis: cuándo build vale la pena vs buy. ROI calculations. Cost projections Year 1 / Year 2 / Year 3.

Cost categories

Infrastructure

  • Compute (engine, workers, frontend)
  • Storage (DB, backups, logs)
  • Network (bandwidth, load balancer)
  • Observability (APM tools)

Operational

  • On-call labor
  • Maintenance time
  • Updates / patches
  • Backup management

Development

  • Initial build cost
  • Feature development ongoing
  • Bug fixes
  • Migrations

Compliance / Insurance

  • Security audits
  • Penetration tests
  • Cyber insurance
  • Compliance certifications (SOC2, etc.)

Comparison framework

We compare for 3 scales:

Scale Volume
Small 1-10 TPS, <100K instances/month, 1-3 tenants
Medium 50-200 TPS, 1M-10M instances/month, 10-100 tenants
Large 1000+ TPS, 100M+ instances/month, 1000+ tenants

Camunda 8 SaaS pricing (reference)

Public pricing 2026 (approximate, varies by plan):

Plan Price Includes
Starter $0 Free tier, 250 instances/month
Team ~$2000/month 25,000 instances/month, 1 cluster
Business ~$8000/month 250,000 instances/month, dedicated cluster
Enterprise Custom Volume-based, support, SLA

Per-instance pricing roughly: $0.05-0.10 each.

Overage: ~$0.10 per additional instance.

Camunda 8 Self-Managed

  • License: contact sales
  • Typical enterprise: $50K-$500K/year
  • Plus your infrastructure costs
  • Plus support contracts

Temporal Cloud pricing (reference)

  • Per-action pricing: ~$0.000022 per state transition
  • Storage: $0.025/GB-day
  • Egress: $0.09/GB

Typical workflow with 50 events = $0.0011/instance.

For 1M instances: ~$1100/month + storage + egress = $2-5K typically.

MVP self-hosted costs

Small scale (Phase 0)

Infrastructure:
- 1 engine VM: 4 CPU, 16 GB RAM = ~$100/month (AWS m5.xlarge)
- 1 Postgres VM: 4 CPU, 16 GB RAM, 100 GB storage = ~$150/month
- Load balancer: ~$25/month
- Backups (S3): ~$10/month
- Bandwidth: ~$10/month

Observability:
- Grafana Cloud free tier or Prometheus self-hosted: $0
- Or APM SaaS: ~$50/month

Subtotal: ~$345/month

Medium scale (Phase 2)

Infrastructure:
- 3 engine VMs (HA): 8 CPU each, 32 GB RAM = ~$900/month
- 3 Postgres VMs (Patroni): 8 CPU each, 64 GB RAM, 500 GB SSD = ~$1800/month
- PgBouncer: ~$50/month
- Load balancer: ~$50/month
- Backups (S3): ~$50/month
- Bandwidth: ~$100/month

Observability:
- APM SaaS: ~$500-1000/month (depending on volume)

Subtotal: ~$3500-4000/month

Large scale (Phase 4)

Infrastructure:
- Citus cluster: 1 coord + 6 workers = ~$5000/month
- Patroni HA per shard: ~$3000/month
- Engine instances (10): ~$3000/month
- Network, LB, etc: ~$500/month
- Backups: ~$500/month
- Bandwidth: ~$1000/month

Observability:
- APM: ~$2000-5000/month

Subtotal: ~$15000-18000/month

Side-by-side comparison

Small scale (~25K instances/month)

Solution Monthly cost Yearly
Camunda 8 Team $2000 $24K
Camunda 8 Self-Managed (small license) ~$5000 (incl. license amort) ~$60K
Temporal Cloud $500-1000 $6-12K
Temporal Self-Managed ~$500 ~$6K
MVP Self-Hosted ~$345 ~$4K

MVP wins easily at small scale.

Medium scale (~1M instances/month)

Solution Monthly cost Yearly
Camunda 8 Business $8000 $96K
Camunda 8 Self-Managed $12K-20K $144-240K
Temporal Cloud $3-5K $36-60K
Temporal Self-Managed $2-4K $24-48K
MVP Self-Hosted (Phase 2) $3500-4000 $42-48K

MVP competitive with Temporal, half of Camunda.

Large scale (~100M instances/month)

Solution Monthly cost Yearly
Camunda 8 Enterprise $30K-50K $360-600K
Camunda 8 Self-Managed $30K-60K $360-720K
Temporal Cloud $20-40K $240-480K
Temporal Self-Managed $15-25K $180-300K
MVP Self-Hosted (Phase 4) $15-18K $180-216K

MVP wins at scale.

TCO including development

Add development costs to operational:

Year 1 (MVP build)

Development (per [analysis/implementation-roadmap-concrete](<../../analysis/implementation-roadmap-concrete.md>)):
- 3 FTE × $150K/year (loaded cost) = $450K
- 6 months to MVP = $225K (M4)
- 8 months to Phase 1 HA = $300K (M5)

Year 1 total dev: $300K
Year 1 ops (medium scale): $48K
Year 1 TCO: $348K

Year 2-3 (operational + features)

Development (1.5 FTE maintenance + features): $225K/year
Operational (medium scale): $48K/year
Year 2 TCO: $273K
Year 3 TCO: $273K

Cumulative 3-year TCO comparison

Solution Year 1 Year 2 Year 3 3-year TCO
Camunda 8 SaaS $96K $96K $96K $288K
Camunda 8 Self-Managed $244K $144K $144K $532K
Temporal Cloud $60K $60K $60K $180K
MVP Self-Hosted $348K $273K $273K $894K

Wait, MVP is MORE expensive?

Yes, if you only need basic features.

Reasons MVP costs more in 3-year TCO: 1. You're paying full development cost 2. Camunda/Temporal amortize dev across thousands of customers 3. Their per-instance cost is marginal

But MVP wins when: 1. Year 4+: development cost amortizes 2. Many tenants: you can multi-tenant the MVP cost 3. Custom requirements: features Camunda/Temporal don't have 4. Compliance: airgap, data residency Camunda can't provide 5. Strategic asset: own your platform, not vendor-dependent

Year 5+ TCO

Cumulative 5-year:
- MVP: $348K + $273K × 4 = $1,440K
- Camunda 8 SaaS: $96K × 5 = $480K

But:
- Cost per instance scales differently
- At 10M instances/month sustained: 
  - Camunda 8: $5K × 12 × 5 = $300K plus overage charges = $500K-$1M
  - MVP: $48K × 5 = $240K
- MVP wins at this scale

When does MVP break even?

Camunda 8 marginal cost = $0.10/instance
MVP marginal cost = ~$0.001/instance (mostly fixed)

Fixed development cost recoup:
$300K dev cost / ($0.10 - $0.001) per instance = ~3M instances
At 1M instances/month: break-even ~3 months operational
But add maintenance: ~$225K/year = ~2.5M instances/year overhead

Net break-even:
~10M instances total OR
~$1M/year Camunda cost equivalent volume

When MVP is justified

Build (MVP) makes sense when

  1. Very high volume: > 50M instances/month sustained
  2. Many small customers: $0.10/instance margin matters
  3. Compliance constraints: airgap, on-prem required
  4. Strategic platform: building product on workflow engine
  5. Cost-sensitive enterprise: $500K/year+ to save justifies build
  6. Unique requirements: features Camunda doesn't have
  7. Long-term commitment: 5+ year horizon

Buy (Camunda/Temporal) makes sense when

  1. Small volume: < 1M instances/month
  2. Limited team: < 3 engineers available
  3. Fast time-to-market: weeks not months
  4. Standard features: BPMN + workflows core
  5. Vendor support valued: dedicated support team needed
  6. Compliance certified: Camunda has SOC2 etc.

Pricing model for MVP (if commercializing)

If selling MVP as service:

Option 1: Per-instance

Pricing:
- $0.005 per process instance (5x cheaper than Camunda)
- Include 10K instances/month free

Cost (medium scale):
- Customer with 1M instances = $5000/month revenue
- Your cost to serve = ~$400/month (mostly infra divided)
- Margin: 92%

Option 2: Per-tenant (flat)

Pricing:
- $500/month per tenant (regardless of usage)
- Includes 100K instances
- Overage at $0.001/instance

Simpler billing, predictable revenue.

Option 3: Tiered

Free: 10K instances/month
Starter: $100/month, 100K instances, 5 users
Pro: $500/month, 1M instances, 25 users
Business: $2000/month, 10M instances, unlimited users
Enterprise: Custom

Cost optimization strategies

Infrastructure

  • Reserved instances: 30-60% off on-demand
  • Spot/preemptible for non-critical (workers, dev)
  • ARM instances (Graviton): 20-40% cheaper
  • Auto-scaling: scale down at night
  • Right-sizing: profile before sizing

Storage

  • Compression: Postgres TOAST + JSONB compression
  • Tiered storage: Hot (SSD) → Cold (HDD/S3)
  • Archival: per concepts/command-log-compaction
  • Retention policies: only what compliance requires

Network

  • CloudFront/CDN: cache static assets
  • Same-region: avoid cross-AZ/region transfer
  • VPN: avoid public internet bandwidth charges

Observability

  • Sampling: traces 1%, logs 1% INFO
  • Retention: hot < cold appropriately
  • Self-hosted: Prometheus + Grafana saves $500-2000/month vs Datadog

Hidden costs to watch

Compliance certification

  • SOC2 Type 1: ~$10K-20K initial
  • SOC2 Type 2: ~$30K-50K (annual)
  • ISO 27001: $30K-100K initial + ongoing
  • HIPAA BAAs: legal review costs

Penetration testing

  • Annual: $20K-50K
  • Continuous (bug bounty): $50K-200K/year

Insurance

  • Cyber liability: $5K-20K/year
  • E&O (errors and omissions): $5K-15K/year
  • Privacy policy / Terms: $5K-15K
  • DPA (Data Processing Agreement) templates: $5K-10K
  • Customer contract review: ongoing

Sales / Support (if SaaS)

  • Customer success: 1 person per ~50 customers
  • Documentation: ongoing cost
  • Marketing/sales: variable

Cost vs benefit decision matrix

                    Low volume    High volume
                    -----------   ------------
Standard features:  Buy (Camunda) Build (MVP)
Custom needs:       Build (MVP)   Build (MVP)
Limited team:       Buy           Buy or partner
Strategic asset:    Build         Build
Cost-sensitive:     Buy first     Build
Time-pressed:       Buy           Buy or hybrid

ROI calculator

For decision-making:

def mvp_roi_breakeven_years(
    annual_volume_instances,
    camunda_cost_per_instance=0.10,
    mvp_dev_cost=300_000,
    mvp_yearly_ops=48_000,
    mvp_yearly_maintenance=225_000,
):
    camunda_yearly = annual_volume_instances * camunda_cost_per_instance
    mvp_yearly = mvp_yearly_ops + mvp_yearly_maintenance

    annual_savings = camunda_yearly - mvp_yearly
    if annual_savings <= 0:
        return float('inf')  # Never breaks even

    return mvp_dev_cost / annual_savings

# Examples
mvp_roi_breakeven_years(1_000_000)    # 1M/year → 1.1 years
mvp_roi_breakeven_years(10_000_000)   # 10M/year → 0.32 years
mvp_roi_breakeven_years(100_000_000)  # 100M/year → 0.03 years

At 10M+ instances/year, build pays back in months. At <1M instances/year, may never break even — buy.

Strategic considerations beyond cost

Reasons to build despite higher cost:

  1. Vendor lock-in avoidance: own your destiny
  2. IP/competitive advantage: integrated platform
  3. Compliance flexibility: change auditors, regions
  4. Customer trust: own data fully
  5. Custom roadmap: prioritize what you need
  6. Talent attraction: engineers want to build cool stuff
  7. Exit value: tech IP increases company valuation

These can be worth millions despite higher operational cost.

Conclusion

For most companies: buy Camunda or Temporal initially. Build MVP only when: - Volume justifies it (>10M instances/year) - Custom needs requires it - Strategic differentiator (workflow IS the product) - Team has expertise + capacity

MVP economics improve dramatically at scale. Below 1M instances/year, buying makes more sense unless other factors dominate.