Infrastructure as code is often sold as a tooling decision. Governance is the harder, more expensive part: deciding who may change production, how a change is reviewed, what evidence is kept and how drift is detected. The tool is a small line in the budget. The operating model is the rest of it.
This page builds a cost model for that operating model. The comparison of four IaC governance models explains the choices; here the question is what each choice costs to build and run in a Bangladeshi team. No fixed prices are quoted because engineering rates, cloud fees and audit requirements vary too widely.
What governance adds to ordinary infrastructure as code
A team can write Terraform or OpenTofu without any governance at all: one engineer, one state file, direct applies from a laptop. Governance adds four things. Changes become reviewable before they reach production. State and secrets are protected and recoverable. Environments are separated so a test change cannot touch live data. And evidence exists to show who approved what, which is increasingly requested by enterprise customers and auditors.
Each addition has a build cost and a run cost. The run cost is the one most often left out.
The four cost pools
- Ownership and engineering time. Someone must own the module library, review changes and answer questions. Partial ownership is a false economy; a module nobody maintains becomes a risk.
- Pipeline and tooling. Version control, CI runners, state backend, locking, secrets management and drift detection. Some are subscriptions, some are cloud consumption, and some are engineering days.
- Policy and evidence. Writing policy checks, defining exceptions, exporting reports and keeping records for audit. This is usually the smallest build line and the most underestimated run line.
- Environments and migration. Separating development, staging and production, importing existing resources, and cleaning up what the import exposes.
Line items to price
- Module library design: a small set of reusable modules for networks, compute, storage and identity, each with an owner and a version.
- State backend: remote storage, encryption, locking and backup, including a restore test.
- Pipeline setup: plan on pull request, apply on merge, approvals for production and a rollback path.
- Policy as code: checks for public storage, open security groups, unencrypted volumes, missing tags and prohibited regions.
- Secrets management: a vault or managed secret store, rotation and access review.
- Drift detection: scheduled plans that compare live infrastructure with the repository, with an owner for the resulting queue.
- Environment segmentation: separate accounts or subscriptions, separate state and separate permissions.
- Import of existing resources: often the largest hidden line in a brownfield estate.
- Training: engineers, reviewers and the person on call, measured in sessions and written runbooks.
- Audit evidence: a monthly export of changes, approvals and policy exceptions.
- Incident rehearsals: a failed apply, a corrupted state file and a rollback, tested before they happen for real.
A formula you can fill in
Annual cost = (engineering days × internal day cost) + tooling subscriptions + CI and cloud usage + training + evidence preparation + contingency.
Split engineering days into three buckets: build, migration and run. Build covers modules, pipeline and policy. Migration covers importing existing resources and fixing what the import reveals. Run covers reviews, drift triage, module updates and on-call. A common mistake is to budget only build and migration, then discover that reviews and drift consume a fixed number of hours every month.
Calculator worksheet
| Line | Unit | Your quantity | Unit cost | Notes |
|---|---|---|---|---|
| Module library build | Module | Include tests and documentation | ||
| State backend | Environment | Encryption, locking, backup | ||
| Pipeline configuration | Pipeline | Plan, apply, approval | ||
| Policy checks | Rule | Start with the highest-risk rules | ||
| Secrets management | Store or team | Rotation and access review | ||
| Resource import | Resource group | Brownfield effort | ||
| Environment separation | Account | Permissions and billing | ||
| Training | Session | Engineers and reviewers | ||
| Monthly reviews | Hours per month | Drift and pull requests | ||
| Audit evidence | Report | Monthly or quarterly | ||
| Contingency | Percentage | Applied to build and import |
Modelling the ongoing share
Ask for the run cost as a monthly figure, not a percentage of the build. A percentage hides whether the vendor expects one review a week or ten. If a percentage is the only number offered, treat it as an assumption to test: multiply it by the build, divide by twelve, and compare the result with the hours your team would need to do the same work. The lower number is not automatically better; a low run cost may simply exclude module updates and drift triage.
What moves the number most
- Estate size and age. Importing years of manually created resources costs more than starting clean.
- Number of cloud accounts. Each account adds permissions, state and reporting work.
- Compliance demands. Customer audits and regulated data increase evidence requirements.
- Team skill. A team new to infrastructure as code needs training and review support; skipping it shifts cost into incidents.
- Change frequency. A stable estate needs fewer reviews than a product team deploying daily.
Managed service or in-house: the cost shape
In-house keeps knowledge close and gives direct control, but requires a named owner with time reserved every month. A managed service converts much of the cost into a predictable fee and brings process experience, but you still own the cloud account, the code and the decisions. Whichever route you choose, keep the repository and documentation in your own account so the relationship remains exit-able.
Questions to ask before approving the budget
Who reviews pull requests when the owner is on leave? What happens when a drift check fails at 11pm? Which policy exceptions exist, and who approved them? How is state backed up, and when was it last restored? What is excluded from the monthly fee? The answers will tell you whether the proposal is a governance model or a pipeline with a governance label.
A 90-day plan
Days 1–30: inventory the estate, choose the state backend and define the first five modules. Days 31–60: put one non-critical service through the pipeline, add three policy checks and document the rollback. Days 61–90: import a production service, run a drift exercise and produce the first evidence pack. Keep the legacy process available until the new one has survived a real change window.
Where this fits in our work
Cloud and pipeline work sits inside Managed Technology & Cybersecurity. The pricing page explains how an assessment is separated from a monthly retainer. For the model choice itself, read the IaC governance comparison.
Sources
- OpenTofu Documentation, accessed 17 November 2026.
- AWS, Organizing Your AWS Environment Using Multiple Accounts, accessed 17 November 2026.
- FinOps Foundation, FinOps Framework, accessed 17 November 2026.
- NIST SP 800-190, Application Container Security Guide, accessed 17 November 2026.
Responses (0)
No comments yet. Add the first one below.
Leave a response
Your email address will not be published. Required fields are marked *.