Home › Topics › Post-Incident Reviews › Mini-Reviews

The Fifteen-Minute Mini-Review for Small Incidents


The big outage gets a proper post-incident review: meeting, timeline, action items, the works. The twenty-minute blip gets a shrug, a "weird, it's back," and everyone returns to their queue. The reasoning feels sound, because a 45-minute review for a 20-minute incident is overkill. The conclusion drawn from it, therefore no review, quietly throws away most of what your infrastructure is trying to teach you. Small incidents outnumber big ones many times over, and their lessons evaporate within days if nobody writes them down.


I ran a team like that for a while. We reviewed the outages with great care and shrugged off the blips. It took an embarrassingly long time to notice that the outages were made of the blips.


The fix is not more meetings. It is a right-sized ritual: the mini-review. That means fifteen minutes, solo or in a pair, four questions on one card, done within a day of the incident. This guide makes the case for reviewing the small stuff at all (it is stronger than it looks). It walks the four-question format. It gives you a severity-to-ritual ladder so nobody has to debate which treatment an incident gets. It shows three filled-in example cards and finishes with the quarterly habit that turns a stack of minis into strategy. Nothing here needs a process person, a template committee, or anyone's afternoon.


In this guide

  • Why small incidents deserve review at all
  • The format: four questions, fifteen minutes
  • The review-depth ladder
  • Keeping it fifteen
  • Three worked minis
  • The quarterly harvest
  • Starting the habit
  • The card

Why small incidents deserve review at all


Three arguments, each independently sufficient.


Small incidents are the big ones' rehearsals

Most major outages send a smaller version of themselves ahead, weeks or months early. Suppose the failover worked this time — on a Tuesday at 2 PM with everyone at their desks. That is a dress rehearsal for the night it runs with nobody watching. The disk that a threshold alert caught at 94% is the full-disk outage that did not happen, this once. These are near-misses: incidents where the difference between "blip" and "disaster" was luck, timing, or one working safety net. A near-miss reviewed is a disaster prevented at almost zero cost; a near-miss shrugged off is a disaster on layaway. On a worse day — the standby also down, the alert unseen overnight — the same event plays out with a very different ending. The mini-review is where you notice that before the worse day auditions.


Patterns live in aggregates

Picture five brief, apparently unrelated blips across two months. There is a dropped VPN, two slow mornings, a flapping port (one that keeps going down and coming back), and a printer subnet tantrum. They can share one underlying cause: the same aging switch, the same overloaded uplink, the same vendor's gear misbehaving after power events. No single blip justifies investigation, so investigated singly they never reveal it. The pattern only becomes visible when the blips are recorded in a comparable format and read together. That is precisely what a stack of mini-review cards is. (Hunting shared causes across seemingly separate symptoms is the same skill the Alert Correlation & Root-Cause Thinking guides teach for live alert storms; the mini-review stack is its slow-motion version.)


At Tollgate Foods the warehouse scanners dropped off the wireless for a few minutes most Monday mornings. For two months the response was "weird, they're back." On the ninth Monday somebody filled in a card. It took eleven minutes. The WHY line was that the weekend patch window rebooted the server that hands out network addresses. It also said the scanners gave up waiting before the server came back. The action was one line: move that reboot to Sunday morning. There was no tenth Monday, and nobody had to sit through a meeting about it.


Practice keeps the review muscle warm

A team that only reviews catastrophes reviews maybe twice a year. And it shows: the timeline skills are rusty, the why-questions come out accusatory, and the actions come out vague. All of this happens under the emotional weight of a big failure. A team that mini-reviews weekly has run the same mental motions dozens of times on low-stakes material. When the big one arrives, the full review — the blameless meeting, the timeline walk, the cause analysis — is just the familiar ritual at a larger size. Small weights, kept warm, are how the heavy lift goes smoothly.


The format: four questions, fifteen minutes


The mini-review is done by whoever handled the incident, alone or with one other person, within a day. Memory of small events decays even faster than memory of big ones. (By Thursday, a Monday blip is a rumor.) The mini-review is written, not just thought about; the writing is what makes the aggregate readable later. Four questions, one card.


Solo works fine and is the honest default in a small shop. But the pair variant earns its second person whenever it is available. The one who fixed it narrates, and the other asks the four questions and writes. The asker's ignorance is a feature. "Wait, why would the switch do that?" from someone who was not there routinely surfaces the assumption the fixer stopped noticing they were making. Ten of the fifteen minutes are the same either way; the last five are better with someone who does not already know the answer.


Question 1: WHAT happened?

Write two sentences plus the key timestamps — a micro-timeline with exactly four facts. Record how it was detected (alert? user report? luck?), when it started and ended (duration), what the impact was, and how it was fixed. Pull the stamps from your monitoring history rather than memory — thirty seconds of looking beats a guess. These four facts are not bureaucratic filler. Detected-how and duration are exactly what feeds your incident log and your detection/response measurements. The Measuring Detection & Response guides run on this data. And "detected by a user, not an alert" written down five times in a quarter is a monitoring-coverage finding no single incident would have surfaced.


Question 2: WHY?

One why-chain, one branch, five minutes — the pocket edition of the full cause analysis in contributing causes and the five whys. Ask why from the symptom and follow the single most important thread. And stop at the first actionable cause — the first answer naming something you could realistically change. The full analysis would branch and sweep; the mini deliberately does not. You are not writing the definitive account, you are bagging the one lesson this event offers. Same grammar rules apply at this scale, by the way: the why-chain asks about systems, not people — even when the only person involved is you.


Question 3: WHAT'S OWED?

Zero or one action. Not "a few" — zero or one. Zero is a perfectly respectable answer that many minis should reach. "Provider maintenance, communicated, nothing to fix" is a finding, not a failure. When there is an action, it gets written properly even at this scale: verb, specific deliverable, owner, date, done-looks-like. That is the same anatomy from action items that actually happen. The action goes into the same work queue as every other incident action. One well-written small action per week, actually landing, compounds into a startling amount of hardening per year. A pile of half-written ones compounds into nothing.


Question 4: WORTH A FULL REVIEW?

The last question is the triage function. And it is what makes the mini safe to use as the default: does what we just found justify the full treatment? Usually no. But sometimes the five-minute why-chain hits something with a floor under it. You may find "this is the third one this month." Or you may find "the fix was a restart and nobody knows why it works." Or you may find "on a weekend this would have been an eight-hour outage." In those cases, the honest answer is yes. Then the mini's output is a booking: full review, within the week, with the mini card as its starting brief. The mini did not replace the full review; it detected the need for one. That detection is worth the fifteen minutes even when everything else on the card turns out boring.


The review-depth ladder


Which incidents get which ritual should never require a judgment call at 5 PM on a tired day. That is how everything quietly gets "we'll skip it just this once." Map review depth onto the severity levels you already use for response. If you have not formalized those, the Incident Response guides cover building a severity model. Let the ladder decide:


SeverityLooks likeReview ritualWhen
Sev 1 — majorReal outage, many users, real durationFull review meeting: timeline, causes, actionsWithin a week
Sev 2 — significantDegraded service or brief outage, limited scopeMini-review, upgradeable by Question 4Within a day
Sev 3 — minorBlip, single system, little or no user impactMini-review — or a one-line log entry if truly routineSame day
Near-missNo impact — because a safety net or luck heldAlways at least a mini. No exceptionsWithin a day

Two clarifications make the ladder run smoothly. "Log-only" on the Sev 3 row means one dated line in your incident log — what, when, how long, fixed how. This is reserved for the genuinely routine event you have seen before and understood before. The moment a blip is new, odd, or user-visible, it earns the card. And "upgradeable" means Question 4 can promote any incident upward, but nothing ever gets demoted on the day. A tired 5 PM is exactly when a Sev 2 gets talked down to "basically fine." That is the failure mode the ladder exists to remove.


The near-miss row is the most valuable line on the ladder, and the one teams resist — "nothing even happened, why write it up?" Because nothing happened for a reason, and you need to know whether that reason was a defense you can rely on or a coin flip. The failover that engaged, the alert that caught the disk in time: reviewing those takes fifteen minutes and tells you which safety nets are real. It is the cheapest incident analysis you will ever do — all of the lesson, none of the downtime.


Keeping it fifteen


The timebox is load-bearing. Fifteen minutes is short enough to do the day it happens, short enough to never need scheduling. It is short enough that "we always mini-review" survives busy weeks. The ritual works because it is small, and every minute it grows makes skipping it more defensible. So defend the box. A mini that wants to balloon is always one of two things. Either the incident is genuinely deeper than it looked — in which case that is Question 4 answering itself: stop, book the full review, done. Or you are polishing — hunting the perfect why-chain, wordsmithing the summary — in which case stop for the opposite reason: the card is a field note, not a publication. The card format itself is the container that enforces this; there is simply nowhere on it for a fourth paragraph.


Tip: anchor the habit to an existing moment rather than willpower: the mini-review happens before the incident's ticket is allowed to close. Ticket closure is a gate you already pass. Hanging the card on it means the ritual gets remembered by the process instead of by a human at their most relieved and forgetful.


Three worked minis


Here is what the card looks like in real use — a blip with nothing owed, a near-miss that produced one action, and a mini that escalated.


MINI-REVIEW: WAN latency spike            (pair: me + Jo)
WHAT  Latency to the internet tripled 10:04-10:19; calls
      choppy, nothing down. Detected by latency warning
      at 10:06. Cleared on its own. ~15 min, minor.
WHY   Provider maintenance on their edge (their status
      page confirms; announced on a list nobody reads).
OWED  Nothing to fix on our side. Noted: subscribe ops
      inbox to provider maintenance notices? — folded
      into existing "tidy notifications" task, not new.
FULL REVIEW?  No.

MINI-REVIEW: file-server disk near-miss        (solo)
WHAT  Disk on FS1 hit the 94% warning 07:40; cleaned up
      by 08:15. Zero user impact — caught by threshold
      alert. Would have filled overnight Friday.
WHY   App debug logging left on after last week's
      troubleshooting session; nothing reminds anyone
      to turn it off. First actionable cause — stop.
OWED  Add "revert temporary logging/debug settings" as
      a closing step in the troubleshooting runbook
      — me — Thursday — done when the step is in the
      runbook and this week's two open tickets used it.
FULL REVIEW?  No — but filed as a near-miss: the alert
      is a proven safety net (worth knowing).

MINI-REVIEW: app server outage, 18 min        (solo)
WHAT  App down 14:02-14:20, all users. Detected by
      service check 14:03. Fixed by service restart.
WHY   Service ate its memory until it stalled. Why?
      Unknown — and restart "fixed" it last month too.
      Second occurrence, cause not understood, fix is
      a restart. That is not a why-chain; that is a
      trend.
OWED  No quick action — the fix IS the problem.
FULL REVIEW?  YES — booked Thursday. This card is the
      opening brief; pulling memory graphs for it now
      while the fine-grained data still exists.

Note the range: the first card cost ten minutes and cleanly established "nothing owed." The second converted luck into a permanent runbook improvement. The third did the mini's highest-value trick — it refused to let a repeat offender pass as a blip. A restart that "fixes" the same fault twice is a ritual sacrifice. And the thing it keeps fixing is a trend with a costume on. The note about pulling graphs immediately is timeline-reconstruction instinct. Fine-grained monitoring data ages away fast. The guide to reconstructing the timeline explains why capturing it early matters.


The quarterly harvest


Individually, mini cards are small change. The compounding happens once a quarter, in a single sitting. Read the whole quarter's stack in one pass — twenty or thirty cards, maybe an hour. Hunt for what no single card could show. The same vendor appearing four times. The same subnet under three different symptoms. The cluster of Monday-morning cards that all trace to the weekend patch window. The count of "detected by user" versus "detected by alert" — your monitoring coverage grading itself. Recurring near-miss saves by the same safety net (that net deserves investment) or repeated luck (that gap deserves an action).


Write the harvest up in half a page: patterns found, and one or two properly written actions or project candidates they justify. This is where fifteen-minute field notes become strategy input. The fragile subnet becomes a replacement case with twelve receipts attached. The case is filed in the same uptime evidence file that holds your availability numbers and full review documents. Budget conversations go differently when "that switch is flaky" arrives as a stack of dated cards instead of a feeling.


Starting the habit


Do not launch this with an announcement and a policy page; launch it with three cards. Sit down once, pull up your monitoring history, and mini-review the last three small incidents you can still partially remember. Yes, retroactively, and yes, the WHY sections will be thin because the details have already evaporated. That thinness is the sales pitch: feeling the memory gap on a two-week-old blip makes the within-a-day rule self-evident in a way no argument does. From there, run the ticket-closure gate for a month and expect the early cards to be clumsy — too long, then too vague, then about right. The habit has stuck when a card takes ten minutes without ceremony and someone reaches for the stack mid-discussion. They check whether "haven't we seen this before?" is a memory or a fact. That reach is the whole system working.


The card


The centerpiece, ready to copy into your wiki or print and stick to the wall. One card per incident; fifteen minutes; within a day.


MINI-REVIEW — <incident name> — <date> — <who>
Timebox: 15 minutes. Written, not just thought about.

1. WHAT   Two sentences + four facts, stamps from
          monitoring history, not memory:
          detected how/when · duration · impact · fixed how
2. WHY    One why-chain, one branch, five minutes.
          Ask about systems, not people.
          Stop at the first actionable cause.
3. OWED   Zero or ONE action. If one: verb + deliverable
          + owner + date + done-looks-like → work queue.
4. FULL REVIEW?   If this revealed depth, repetition, or
          a lucky escape — book the full meeting within
          a week. This card is its opening brief.

File the card with the incident. Read the stack quarterly.

Where to go next


The mini-review is the small end of one continuous practice. When Question 4 says yes, you graduate to the full machinery. That means the blameless review meeting, the timeline reconstructed from monitoring data, the full cause analysis, and actions that land. Start with the minis, though; they are the habit that makes the rest feel normal. The whole practice lives in Post-Incident Reviews. Fifteen minutes is not much longer than a shrug, and it leaves a card behind.


Put this into practice with SysPeep Monitor

Question 1 of every mini-review — detected how, when, for how long — answers itself if your monitoring history is good. SysPeep Monitor records alert times and performance data for auto-discovered devices across your network. So the micro-timeline is a thirty-second lookup in one web console instead of an act of memory. And its threshold alerts are how near-misses like the 94% disk get caught at all. The free trial installs on a Windows machine in minutes.

Download Free TrialLearn About SysPeep