Skip to main content
Service Management

Changes governed, not just tracked.

Every change carries a recorded purpose and rollback plan, a risk score that explains itself, a decision taken by the right people under a frozen rule, and closing evidence of what actually happened. Separation of duties is enforced in the service code, so it holds the same way whether a person clicks, a pipeline calls the API, or an approver votes from an email link.

What you can do

The controls that stop a change reaching production the wrong way.

Raised from where the work starts

A four-step form, or straight from an incident, problem, vulnerability finding or missing patch, with the affected CIs copied across. Employees can propose a change from the portal for a change manager to accept or decline, and deployment pipelines raise changes idempotently through the REST API.

A risk score that shows its working

Risk is scored from a fixed catalogue of facts about the change, weighted by your policy, plus your own risk questions, and every point is explained under Why this score. The snapshot is frozen at submission, so a later edit to a CI, a freeze window or a weight never changes the risk an approval was granted against.

Collisions, freezes and blackouts

The collision check walks the dependency graph: same CI, dependency path, shared dependency or maintenance window, with the hop count. Linking a CI that another in-flight change uses in an overlapping window is refused outright. Scheduling inside a freeze needs a written acknowledgement; a blackout blocks approval, scheduling and start, Emergency included, until someone authorised records an override.

A decision docket, not a vote count

Each change shows its decision rule, stage, frozen seats, votes and the version of the description being decided on. Choose from single approver, all approvers, majority with quorum, or staged technical-then-business approval. A tenant minimum can raise the number of approvers, never lower it. A reject or veto ends the decision.

Segregation of duties in code

The requester can never approve their own change, and the implementer cannot be its only approver. Override and blackout-override rights are separate from approval rights. Approvers can hand their seat to a delegate for up to 90 days, and each delegated vote shows both names.

Approved stays approved

Once a change leaves Draft, the fields approvers decided on are locked: title, description, justification, rollback plan, window, risk, service and linked CIs. Changing one means withdrawing to Draft, and earlier approvals no longer count. Moving the window after approval is re-checked against blackouts, freezes and overlaps, and audited.

Test evidence that actually gates

A change model can require test evidence before implementation starts: every item passed or waived, and a waiver needs a reason. Required checklist tasks must be done before the work starts, and pipelines ask a deployment gate that answers allow, hold or deny with a reason code.

Outcome, review and an evidence pack

Every implemented change records its outcome and whether it caused service impact. Every Emergency, failed or rolled-back change gets a structured review, and that floor cannot be switched off. The closure panel lists exactly what is still missing, and the evidence pack exports with its SHA-256 fingerprint in the audit log.

Compliance relevance

How this connects to regulation

DORA
Art. 9(4)(e) — ICT changes recorded, tested, assessed, approved, implemented and verified in a controlled manner, with one lifecycle and an evidence pack per change.
MaRisk
AT 4.3.1 — Internal control: the requester is never the approver, enforced in code, with every decision in the hash-chained audit log.
DORA
Art. 11 — Rollback plan required for Normal and Emergency changes, with a Rolled back state and outcome on the record.
Common questions

What change managers ask.

See TENSOR running on your own estate.

30 minutes, screen-shared, no slides. Bring the process you are least sure we handle and we will start there.