Go-Live & Incident Response Guideline

Department

Technology

Summary

Minimum operational playbook for go-live execution and incident response: environment flow (Local → DEV → UAT → Staging → Production), rollback and verification plans, execution steps, and incident triage → contain → communicate → resolve workflow.

Table of Contents


Purpose

This guideline defines the minimum operational playbook for go-live execution and incident response (who does what, what to write in Kiluth Tasks, and what “done” means).

Outcome
Production changes are executed with clear ownership, and incidents are handled with fast containment and clear communication.

Prerequisites

Before proceeding with this document, please review the following documents:

#DocumentPurpose
1Delivery Production & Resource Allocation GuidelineUnderstand delivery execution, sprint planning, and work tracking
2Onboarding GuidelineUnderstand how Kiluth uses Kiluth Tasks, Google Workspace, and other core tools

Scope

This guideline covers:

Included
1Environment definitions and promotion flow
2Go-live execution responsibilities (including rollback)
3Minimum incident response workflow and artifacts

This guideline applies to client project systems — the environments Kiluth builds, ships and operates for a customer.

Not covered
1Kiluth’s own internal systems (the portal, docs site, chat, company website). Incidents there are handled by Technology directly and recorded in the department’s own notes and runbooks, not through this guideline.
2Project-specific deployment runbooks. This guideline does not replace them; if a project needs one, link it from the go-live task.

Definitions

TermDefinition
Go-liveRelease a change into Production (real users / real operations)
RollbackReverting to a known-safe state when go-live introduces unacceptable risk or failure
IncidentAn issue in a client’s production system that materially impacts user experience, data correctness, or business operations

Operating Model (Sprint + Kiluth Tasks + Visibility)

This document follows Kiluth’s standard operating model. See Onboarding Guideline for details on Kiluth Tasks usage, checkpoints, decision points, and handoff processes.

Operating Model
1Kiluth Tasks is the system of record for work tracking, approvals, and handoffs.
2Use checkpoints and decision points: don’t move forward until the previous step is “done”, and branches are explicit.
3Handoff order: upstream defines handoff artifacts/exit criteria; downstream defines execution after handoff.

Environment Overview

EnvironmentDescription
LocalDeveloper machine for coding and testing
DEVInternal team testing server
UAT / DEMOClient-facing trial environment for validating that functions work correctly per agreement and are acceptable
StagingPre-production environment that is as close to Production as possible for final testing
ProductionLive system for real users

Workflow Summary

Development → Testing → Client validation → Production-like validation → Go live

Local → DEV → UAT/DEMO → Staging → Production


Step-by-Step Process

Step
1Confirm go-live owner + execution window (who runs the change, when, who is on standby)
2Confirm rollback plan exists (what rollback means here + how to execute it)
3Confirm verification plan exists (the checklist that declares success)
4Execute go-live and record execution notes in Kiluth Tasks
5Verify, then communicate outcome (success / rollback / follow-up actions)

Decision point: go-live outcome

OutcomeAction
Verified (success)Proceed to hypercare/monitoring plan (if defined) and close the go-live execution task.
Not verified (failure / unacceptable risk)Execute rollback plan, record what happened, and align next steps + new schedule.

Task Template: Go-live execution (Technology)

Task Template
TitleGo-live execution (Technology) – [Client Name] – [Project Title]
AssigneeTechnology owner
DescriptionPlease execute go-live for this project.

References
• Project: [Link]
• Delivery go-live task: [Link]

Environment path
Local → DEV → UAT/DEMO → Staging → Production

Rollback plan
• Rollback approach (high-level): [Notes/link]

Verification plan
• What to verify post-release: [Checklist/link]

Execution
• Go-live datetime: [Datetime]
• Execution notes (minimum): what changed + start time + end time + any deviations + who executed

Result
• Outcome: Verified / Rolled back
• Follow-ups (if any): [List/link]

Incident Response (Minimum)

Step
1Triage and confirm impact (what is broken, who is affected, severity)
2Contain (stop the bleeding: rollback, disable feature, hotfix, mitigation)
3Communicate status (AE/PM aligned; client comms when needed)
4Resolve and verify
5Record a short incident summary and follow-up tasks (prevent recurrence)

Severity Levels

Set severity from impact, not from how hard the fix looks. When two levels seem to fit, take the higher one.

LevelWhat it means
HighProduction is down, data is at risk, or payments/checkout are broken. Most or all users affected. No workaround.
MediumA feature is broken but the system still works, and there is a workaround. A subset of users affected.
LowCosmetic, affects one user, or affects a non-production environment only.

Severity drives what happens next:

LevelWhat changes
HighStop other work. Technology owner responds immediately and names an AE/PM as client comms owner before doing anything else.
MediumFix within the current sprint. Raise it at the weekly cadence rather than interrupting the sprint.
LowQueue as a normal task. No interruption.

Telling the client

LevelDo we tell the client?
HighYes, straight away — before the fix, not after. The AE makes the contact; the Technology owner keeps working the incident.
MediumAt the next normal update. No separate message needed.
LowNo, unless the client reported it — then reply to them directly.

On a High incident the client will see the problem themselves. Speaking first is a different conversation from being caught after they noticed. Say what is broken, that it is being worked on, and when the next update comes — an early message without a fix in it is still worth sending.

This sets when to speak, not how fast Kiluth must fix. Kiluth has not committed a response or resolution time to any client, and none is stated here.

Internal aim (not a client commitment): a High incident is picked up the same working day. Miss it and nothing is breached — it is a bar for the team to judge against, not a promise made to anyone outside Kiluth.

When personal data may have been exposed

If an incident may have exposed personal data — names, email addresses, phone numbers, ID numbers, anything identifying a real person — a legal clock starts, and it is not Kiluth’s clock.

RoleWhoDuty
Data controllerNormally the client. It is their users’ data and they decide what it is used for.Notifies Thailand’s PDPC, and notifies affected people when the risk to them is high.
Data processorNormally Kiluth. We build and operate the system to the client’s instruction.Notifies the client, without delay.

What Kiluth does:

Action
1Tell the client immediately, in writing. Do not wait for the fix, and do not wait for certainty about the full extent.
2State what you know: what data may be affected, roughly how many people, when it started, and whether it is contained.
3Say plainly what is still unknown. A partial picture sent now is more useful to the client than a complete one sent late.
4Support the client’s investigation and give them what they need for their own filing.
5Record it all on the incident task.

What Kiluth does not do: contact the PDPC, or contact the affected people. Those are the controller’s calls to make. Making them for the client would take a decision that is legally theirs.

The controller’s deadline is short, so Kiluth being slow to report removes time the client needs. Speed here is the whole obligation.

#Exceptions to check
1Processor is the normal case, not a guarantee. The role follows who decides what the data is used for. If Kiluth chose what to collect or how to use it rather than building to instruction, Kiluth may be a controller, and the regulator deadline becomes Kiluth’s. Raise it rather than assuming.
2The contract wins if it is stricter. If a client agreement sets a tighter notification deadline than “without delay”, follow the contract.
3Kiluth’s own systems are out of scope here. This guideline covers client project systems (see Scope).

Hypercare

A short, agreed watch period straight after go-live — typically the first days of live traffic — where the team monitors more closely than normal and responds faster. The window is agreed with the AE at go-live and written on the go-live task. Outside a hypercare window, incidents follow the normal severity rules above.

Task Template: Incident

Task Template
TitleIncident – [System/Feature] – [Short description]
AssigneeTechnology owner
DescriptionImpact: [Who is affected / what is broken]
Severity: [Low/Medium/High] (see Severity Levels)
Start time: [Datetime]

Containment: [Actions taken]
Current status: [Short]

Client comms owner (AE/PM): [Name]

Resolution: [What fixed it]
Verification: [What checks passed]

Follow-ups: [Tasks/links]

Go-Live Checklist (Technology Use)

Preconditions

Checklist
1☐ Go-live owner assigned
2☐ Rollback plan exists (high-level)
3☐ Verification plan exists
4☐ Relevant stakeholders are on standby (as agreed)

Execution

Checklist
1☐ Execution notes recorded in Kiluth Tasks
2☐ Verification completed
3☐ Outcome communicated (success / rollback / next steps)