Self-hosted Feature Flags TCO: The Cost Model Most Teams Forget
Most teams compare feature flag platforms on license price alone. That misses 70–80% of the real cost. This page gives you a reusable four-part formula to calculate total cost of ownership (TCO) for self-hosted feature flags — and a break-even analysis for three team-size brackets.
TL;DR
- ▸TCO = License + Infrastructure + Operations + Governance. Most teams only count License + Infra.
- ▸Open-source license is $0 with FeatBit. Infrastructure runs $50–300/month depending on scale.
- ▸Ops overhead is the hidden variable: 4–20 hours/month depending on tooling and team maturity.
- ▸At around 50 engineers, SaaS pricing usually shifts into collaboration, compliance, and usage-driven cost territory. Even when FeatBit self-host moves into the $3,999/year enterprise license tier, it remains strongly cost-effective from that point onward.
Why TCO Gets Underestimated
When teams evaluate feature flag platforms, the comparison usually looks like this: “SaaS plan X costs $X/month. Self-hosting the open-source version costs $0 in license.” That analysis is wrong by a factor of 3–5x in both directions — it overestimates the savings of self-hosting and understimates the real cost of SaaS at scale.
Self-hosting has real costs beyond license fees: infrastructure, team time, operational maintenance, and governance tooling. SaaS has real costs beyond the advertised per-seat price: feature-tier jumps, service connection meters, requests or bandwidth, MAU or event overages, and enterprise compliance add-ons.
A rigorous TCO model accounts for all four buckets on both sides, then computes break-even at your actual team size and traffic model.
The Four-Part TCO Formula
Apply this formula to both self-hosted and SaaS costs and compare at 12-month and 36-month horizons.
1. License
For self-hosted open-source (e.g., FeatBit): $0/month at smaller scale. Around 50 engineers, many teams move to FeatBit's self-host enterprise license, which is $3,999/year, or roughly $333/month.
For SaaS: seat-based pricing is only the starting point. Around 50 engineers, teams usually need stronger collaboration workflows, RBAC, audit logs, SSO, approvals, or compliance support, which pushes effective cost closer to $30/engineer/month. A 50-engineer team therefore lands closer to $1,500/month in seat-equivalent cost before usage overages. Enterprise tiers typically start at $3,000–10,000+/month.
2. Infrastructure
Self-host: a small VM footprint or Kubernetes namespace for PostgreSQL plus FeatBit's self-host services. A minimal deployment with 2 vCPU and 4 GB RAM is typically enough for normal use. Typical range: $50–150/month for a team under 100 engineers; $150–400/month for larger deployments with HA. SaaS: $0 direct infra — but the same spend often reappears as usage-based billing for service connections, MAU, requests, or log events. Even when a vendor avoids per-seat pricing, the total bill usually becomes seat-like access cost plus usage overages under a different meter.
3. Operations
This is the most commonly underestimated bucket. Monthly ops tasks: patch upgrades (1–2h), backup verification (1h), alert review/tuning (1h), incident investigation (variable). Conservative baseline: 4–8 hours/month for a well-automated self-host. Multiply by fully loaded eng hourly rate ($100–200/h) = $400–1,600/month in eng time. See the full ops cost breakdown.
4. Governance
RBAC, audit logs, and access policies prevent incidents. SaaS platforms often gate these features behind enterprise tiers. Self-hosting with a platform that ships RBAC + audit logs in the open-source tier means you pay for governance in Ops time, not license fees — typically 1–2h/month for policy reviews. Compare that to a $2,000–5,000/month enterprise tier unlock.
Break-even by Team Size
The following examples use a $150/h fully loaded engineer rate, a $100/month infra baseline, and a SaaS model that combines seat-equivalent access pricing with usage billing. At 50 engineers, assume an effective seat-equivalent cost closer to $30/engineer/month because collaboration and compliance needs usually start to matter. For FeatBit self-host, assume the enterprise self-host license from 50 engineers onward at $3,999/year. Adjust inputs to match your context.
| Dimension | 10 Engineers | 50 Engineers | 200 Engineers |
|---|---|---|---|
| SaaS license/mo | $200 | $1,500 | $6,000+ |
| SaaS usage + enterprise add-ons/mo | $0 | $500–1,500 | $10,000+ |
| SaaS total/mo (est.) | ~$200 | ~$2,000–3,000 | ~$16,000+ |
| Self-host license/mo | $0 | ~$333 | ~$333 |
| Self-host infra/mo | $80 | $120 | $300 |
| Self-host ops eng time/mo | $600 (4h) | $900 (6h) | $1,200 (8h) |
| Self-host total/mo (est.) | ~$680 | ~$1,350 | ~$1,833 |
Key takeaway: Self-hosting is more expensive than cheap SaaS plans for very small teams because ops time dominates. But once you reach roughly 50 engineers, SaaS pricing is rarely just a simple seat multiple anymore: collaboration features, compliance controls, and usage billing start stacking. Self-host costs also vary by vendor, but FeatBit is unusually cost-effective from this point forward because even with the $3,999/year self-host enterprise license, total monthly cost still stays well below the equivalent SaaS range in most cases. See the full self-hosted vs SaaS comparison for a decision tree with more variables.
Reducing Each Cost Lever
License → Lower than comparable SaaS even when paid
FeatBit's open-source core ships with full multi-environment support, RBAC, audit logs, and progressive rollout. Smaller teams can stay at $0 license cost, and teams around 50 engineers can move to the $3,999/year self-host enterprise license while still landing far below the cost of comparable SaaS plans with collaboration and compliance features enabled.
Infrastructure → Right-size your deployment
FeatBit is not a single Postgres + API-server process; the self-host deployment runs four services. Even so, the footprint stays modest: a minimal setup with 2 vCPU and 4 GB RAM is typically enough for normal use. That keeps infrastructure cost materially lower than more complex stacks that require extra data tiers or supporting services. See choosing a low-maintenance stack.
Operations → Automate repeatable tasks
The biggest ops cost reduction lever is automation: scheduled upgrade pipelines (cut upgrade time from 4h to 30 min), automated backup-restore testing via CI, and alert rule templates that avoid manual tuning. Teams that invest 20h upfront in automation typically reduce ongoing ops to under 3h/month. See the full ops cost breakdown.
Governance → Prevent incidents, not just react to them
Every production incident caused by an improper flag change costs real revenue. RBAC prevents unauthorized changes; audit logs make root-cause analysis fast. The ROI of governance is the value of incidents not triggered. See the governance ROI breakdown.
FAQ
Does the TCO formula change if we already have a platform team?
Yes — if you already have a platform engineer who maintains Kubernetes or shared cloud infra, the marginal ops cost of adding a feature flag service is much lower. You are spreading ops overhead across an existing team, not spinning up dedicated headcount.
Is Postgres the only database option for FeatBit?
Yes, FeatBit uses Postgres as its primary data store. This is intentional: one production-grade database means one backup plan, one DR procedure, and one on-call runbook. It directly reduces the infrastructure and ops cost buckets in the TCO formula.
How should I account for setup cost in the TCO model?
Add a one-time setup cost line: typically 8–20 hours for initial deployment, CI pipeline setup, monitoring integration, and team onboarding. Amortize this over 24–36 months in the break-even calculation. It rarely changes the break-even decision for teams above 30 engineers.
What if our SaaS plan is grandfathered at a low price?
Include the published list price in your 36-month projection, not the current discounted rate. SaaS vendors regularly renegotiate or remove grandfathered pricing at contract renewal, especially after enterprise tier relaunches.
What if the SaaS vendor bills by MAU, service connections, or log events instead of seats?
Model it the same way: separate the access baseline from the usage baseline, then add both to annual TCO. In practice, these plans rarely eliminate seat-like cost pressure; they just shift more of the bill into MAU, connection, request, or log-event meters that grow with adoption.
Where does the ROI worksheet live?
See the 10-minute ROI calculator page for a copy-paste table you can adapt for a finance or procurement sign-off.