Incident Postmortems for MAS: The 1-Hour and 14-Day Rules

How a blameless postmortem becomes the root cause and impact analysis report MAS expects within 14 days, with SGT timelines, the 4-hour downtime cap and PDPA.

By · · 6 min read

Illustration for the Incident Postmortem Generator guide

Singapore edition. Written for Singapore teams: the MAS Notice on Technology Risk Management (1-hour notification, 14-day root cause report, 4 hours of downtime a year), MAS-Tx reporting from February 2026, and PDPC's three-day breach deadline. US edition

In a Singapore bank, the postmortem is not an internal nicety. MAS Notice FSM-N05 on Technology Risk Management requires a bank to notify MAS within 1 hour of discovering a relevant incident and to submit a root cause and impact analysis report within 14 days. That report asks for almost exactly what a good blameless postmortem already contains, so the practical question is how to run your review so that it produces the regulatory report as a by-product, on time, with a timeline MAS can trust.

This edition walks through that process with an outage at a Singapore bank, built in the Incident Postmortem Generator. It covers the MAS clocks, the 4-hour annual downtime cap, how the February 2026 change to MAS-Tx reporting fits in, and the separate three-day PDPA clock if personal data is involved. It describes the rules as published by MAS and PDPC; your compliance team and the instruments that apply to your licence have the final word.

The MAS clocks a postmortem has to satisfy

Notice FSM-N05, issued on 9 May 2024 and in force from 10 May 2024, applies to banks in Singapore. Its page lists only banks under "Applies to", so if you hold another kind of MAS licence, check which technology risk notice applies to it. For banks, the notice says:

RequirementFSM-N05 rule
What countsA "relevant incident": a system malfunction or IT security incident with a severe and widespread impact on operations, or that materially affects service to customers
First callNotify MAS as soon as possible, and not later than 1 hour after discovery (para 7)
Written reportA root cause and impact analysis report within 14 days of discovery, or longer if MAS allows (para 8)
Downtime capUnscheduled downtime of each critical system no more than 4 hours in total in any 12 months (para 5)
RecoveryA recovery time objective of no more than 4 hours for each critical system (para 6)

MAS's Circular on Financial Institution Incident Reporting (MAS/TCRS/2025/08, 16 December 2025) sets out how the reports travel. From 1 February 2026, reportable incidents go through the MAS-FI Transactions Platform (MAS-Tx) on a revised template with two parts: an Initial Incident Report no later than 24 hours from discovery, and a Final Incident Report within 14 days of discovery. The first notification is still a phone call or message: to your MAS Review Officer during office hours (Monday to Friday, 8.30 am to 6 pm, excluding public holidays), or to the 24-hour MAS Duty Officer line outside them. Critical Information Infrastructure owners also file the national cyber incident form with the Cyber Security Agency, submitted through MAS.

Every one of these clocks starts at discovery, which makes the timeline the most important part of the postmortem. If the review cannot show when the bank knew, it cannot show it reported on time.

What the root cause and impact analysis report must contain

Paragraph 8 of FSM-N05 lists four contents. Mapped to a normal postmortem:

FSM-N05 para 8Postmortem section that feeds it
(a) an executive summary of the incidentSummary and impact
(b) an analysis of the root cause that triggered itRoot cause, separated from the trigger
(c) the impact on the bank's compliance with laws and regulations, its operations, and its service to customersImpact, plus a line from compliance on which rules were engaged
(d) the remedial measures taken to address the root cause and its consequencesAction items with owners and dates

Two things in that list are easy to miss. Item (c) asks for the impact on compliance, which an engineering postmortem rarely covers: whether the downtime cap or the recovery objective was breached, and whether other notifications (PDPC, CSA) were triggered. Item (b) asks for the root cause, not the trigger. A failed certificate rotation is a trigger; the root cause is whatever allowed an untested rotation onto a critical system during business hours.

Worked example: a 2 h 35 min login outage

A retail bank's internet banking login service, classified as a critical system, starts rejecting logins after a certificate rotation on a load balancer. All times are SGT (UTC+8, no daylight saving), written in the 24-hour format the logs use.

Open the Incident Postmortem Generator, and in the timeline box enter one event per line:

14:05 SGT - Login error rate rises after certificate rotation on the login load balancer
14:11 SGT - Synthetic login check alerts the on-call engineer
14:20 SGT - Duty manager confirms customers cannot log in; classed as a relevant incident (MAS 1-hour clock starts)
14:52 SGT - MAS Review Officer notified by phone
15:30 SGT - Rotation rolled back on one node; errors continue on the second
16:40 SGT - Both nodes on the previous certificate; logins succeed

The generator splits each line into its time (for example "14:20 SGT") and the event. Enter 6 for time to detect, 155 for time to mitigate and, if the permanent fix shipped at 11:00 SGT the next day, 1255 for time to resolve. The report shows 6 min, 2 h 35 min and 20 h 55 min.

Now apply the MAS rules to those numbers:

  • 1-hour notification. Discovery at 14:20 SGT gives a deadline of 15:20 SGT. The call at 14:52 SGT was 32 minutes after discovery. Write both times into the report; a reviewer at MAS will look for them.
  • Initial Incident Report. No later than 24 hours from discovery: by 14:20 SGT the next day, on MAS-Tx.
  • Final report. Within 14 days of discovery. If discovery was on a Tuesday, day 14 is the Tuesday two weeks later. Schedule the blameless review in the first week so the draft has time for compliance and management sign-off.
  • Downtime cap. The outage used 2 h 35 min of the 4 h a year allowed for this critical system, leaving 1 h 25 min for the rest of the rolling 12 months. That belongs in item (c), and it should change the priority of the action items: one more outage of the same length would breach the cap.
  • Recovery objective. Service was restored 2 h 35 min after disruption, inside the 4-hour objective. The report should still say how long the failed first rollback at 15:30 SGT cost, because that is the margin the next incident will need.

Good action items here are specific and owned: "Run certificate rotations on the login tier through the pre-production synthetic check, owner: platform lead, due in two weeks", and "Alert on login success rate below 95% for 2 minutes, owner: SRE lead". The generator groups them into detection, mitigation, prevention and solution, which makes a list with nothing under mitigation obvious before MAS sees it.

If customer data was involved: the PDPA clock

A pure availability outage may involve no personal data at all, but many incidents do, and then the Personal Data Protection Act 2012 adds its own clock. PDPC's guidance says a breach is notifiable if it is likely to result in significant harm to individuals, or affects 500 or more individuals. Once an organisation determines that a breach is notifiable, it must notify PDPC no later than three calendar days later, and PDPC expects containment and assessment to start immediately after discovery. Missing the deadline can lead to enforcement, including financial penalties.

PDPC's C.A.R.E. framework (Contain, Assess, Report, Evaluate) ends with a post-breach evaluation: an incident review, a prevention plan and a review of policies. That is the same postmortem, so keep one document with one timeline, and add the extra timestamps PDPC will ask about: when the breach was discovered, and when it was assessed as notifiable.

Running the review in a regulated team

  • Separate the blameless review from any disciplinary process. The postmortem asks what in the system and process let the change through. If people expect the document to be used against them, the timeline will have gaps exactly where MAS will look.
  • Fix the clock source first. Mixed time zones in logs from regional data centres, SaaS tools in UTC and a chat channel in SGT are the commonest cause of a wrong discovery time. State at the top of the timeline that everything is SGT.
  • Write "not recorded" rather than guess. If nobody noted when the duty manager classed the incident, say so and add an action item. A guessed time that turns out wrong is worse in a regulatory report than an admitted gap. The generator prints "Not recorded" for any blank duration for this reason.
  • Track the downtime budget across incidents. The 4-hour cap is per critical system over any 12 months, so each postmortem should record the minutes used and the running total.
  • Use a tool for the deeper analysis. For a long causal chain, the Root Cause Analysis tool walks through 5 Whys and a fishbone diagram, and the Time Zone Converter helps when vendor logs arrive in another zone.

Tools for this job

For the general method behind the review, see the blameless postmortem guide.

Frequently asked questions

How soon must a bank notify MAS of an IT incident?

Under Notice FSM-N05, a bank must notify MAS as soon as possible and not later than 1 hour after discovering a relevant incident. The MAS circular then asks for an Initial Incident Report within 24 hours and a Final Incident Report within 14 days, both on MAS-Tx.

What is a root cause and impact analysis report?

It is the written report FSM-N05 requires within 14 days of discovery. It must contain an executive summary, an analysis of the root cause, the impact on compliance, operations and customer service, and the remedial measures taken. A blameless postmortem with a compliance section covers all four.

Does the 4-hour downtime limit reset every January?

No. FSM-N05 caps unscheduled downtime of each critical system at 4 hours in total within any period of 12 months, a rolling window, so an outage in November still counts the following October.

When does the PDPA three-day deadline start?

PDPC says the three calendar days start when the organisation determines that the breach is notifiable, not when it is discovered. Assessment is expected to start immediately after discovery, so the postmortem timeline should show both times.

Sources and further reading

Open the Incident Postmortem Generator

More Documentation & Review

Every tool in the Documentation & Review section, free and in your browser.