Every operations team has a shelf of lessons that were learned the hard way and told afterwards over coffee. Here are nine of them, collected in the spirit of ResponseOps: security incident response as something a team keeps in practice, not something it improvises. Each item leans on the Ops Atlas, which is open about its own limits, and each comes with the sort of small human detail that makes advice stick.
1. The name is not an official standard
Start with honesty. The Atlas page about ResponseOps says plainly that no response standard uses its label, so a team that quotes it in a contract is asking for trouble. The relevant international standard prefers wording about running responses to ICT incidents. One software company uses the same word for an engineering squad, and emergency agencies use something similar for disasters. Use incident response on paper and ResponseOps in the team’s own conversations, if it helps them.
2. Somebody has to own the moment awareness begins
Among the failure patterns the Atlas catalogues, the missed deadlines are the saddest. Nobody did anything wrong in a technical sense. The facts were known across five desks, but no single person had been given the job of deciding that the company was now officially aware. Reporting clocks, some counted in hours and some in days, started ticking in the gap. Put a name next to that decision before you need it.
3. Evidence has an order, and it is not the order you would guess
An old internet guideline, referenced in the entry, says to collect the most fleeting information first. That means memory ahead of disk, which feels backwards to anyone who grew up copying files. Add a hash at the point of collection, a UTC timestamp in the custody log and a write-once store, and keep your hands off the originals. It is dull work. It is also the work a regulator or insurer may ask to see.
A practical detail that surprises people: each custody entry needs only a few fields, namely what was taken, from where, by whom, at what UTC time and what its hash was. Teams that try to design a grand form usually end up with an empty one. A plain table that anyone can fill in at three in the morning beats an elegant one that nobody opens.
4. Do not trust the channel you are using to plan
If an intruder is reading your email, then your emailed response plan is a gift. The Atlas recommends a separate bridge for response communications, set up in advance, with a few sealed break-glass accounts nearby. One team in a story we often tell kept their emergency numbers on a laminated card in a drawer, and when their chat went strange it was the card that saved them. A card is not a technical control, but it did the job.
5. Eviction is a single move, not a series of nudges
The entry’s most repeated warning is about removing an intruder before you know the full extent. Take out one account and the others keep working, and the visitor now knows you are looking. Account for every persistent route first, then remove them together. There is a directory-specific wrinkle too: the Atlas says the master credential needs two resets with a long overnight-sized gap, since one reset leaves a stale copy trusted for a while.
Containment has its own folklore. People assume that flipping one cloud rule ends the matter, yet the entry notes that a change to a security group does not break connections already being tracked and that new groups allow all outbound traffic by default. Likewise, switching off an identity account can take around a quarter of an hour to bite for some clients and about an hour for others. Verify each cut from the outside before you relax.
6. “Privileged” is a label, not a shield
The Atlas recounts a federal court ruling from June 2020 about a large 2019 bank breach. A forensic firm’s report was ordered released even though lawyers had engaged the firm, because a statement of work that predated the breach suggested the same report would have been produced anyway. The takeaway is not that lawyers are useless but that routing a report through counsel is no automatic protection. Ask a qualified adviser how to structure engagements; this list is general information only.
7. Ransom questions arrive before the ransom note
According to the entry, a US sanctions office strongly discourages paying, while reporting early to law enforcement is treated as a significant mitigating factor. You do not want to be reading that guidance for the first time with an encrypted file server and a countdown timer on the screen. Decide in advance who would be consulted, which advisers are on call and what information the board would need. The decision itself stays with the people accountable for it.
8. Backups are not a finish line
“We have backups” is a comforting sentence and an incomplete one. If the backup is tainted, restoring brings the problem back. The Atlas lists re-encryption after restoring from compromised backups as a known pattern. Even a clean restore cannot undo data theft, so a business facing double extortion still has decisions ahead. Treat the restore as a technical task and the aftermath as a management one.
9. Measure quality, not volume
It is tempting to be proud of a low count of incidents or to worry about a high one. The Atlas says counts describe workload. It prefers quarterly measures that reveal quality: how often an intruder came back after removal, how many notifications met their deadline, how many pieces of evidence kept their integrity. The sample figures in the entry are one return in four, 80 percent on time and 97.4 percent intact. They are illustrations, but they show the shape of a useful scorecard.
Bonus habit worth stealing: rehearse out loud
Several of the items above are easier to do when the team has walked through a scenario together. The Atlas puts the executive tabletop at the end of its first ninety days, after the lead and sponsor are named, a one-page plan exists, a trio of short playbooks are drafted and the evidence store is ready. A tabletop is just a conversation with a script. Run one, write down where people hesitated and fix the top three gaps.
Reading list from our own shelf
If you are weighing how automation might help your operations during an incident, our article on speeding up incident resolution with AIOps covers the working side of the story. And for people shaping the wider operations roadmap, the practical guide to planning AIOps for DevOps, SRE and operations teams shows where response work sits among the other duties.
The lesson underneath all nine items is the same. Good response is mostly decided before the incident, in quiet hours, by ordinary people writing ordinary pages. Pick the item on this list that your team is weakest at, give it an owner and a date, and come back to it next quarter. That is how a shelf of hard lessons becomes a habit.

