Here is a story about a small administrative technology team, invented for this article, and a lesson they learned about the logs behind their scheduling and records systems. There is nothing clinical in it, and no real organisation. It is about the unglamorous back office that keeps appointment systems, staff sign-ins and document portals running, and the practice that the Ops Atlas calls SecDataOps. If you work anywhere near hospital administration, you may recognise a few of the moments.
Monday: the question nobody could answer
The team consisted of four people: a systems administrator, a network engineer, a records coordinator and a newly hired analyst. On a Monday morning a manager asked a simple question. Two weeks earlier, someone had signed in to the staff document portal from an unusual location. Could the team say what that account had opened?
The administrator opened the portal’s log screen. It held the last eleven days. The login was twelve days old. The analyst checked the central archive, which she assumed held more, and found that the portal had never been connected to it. The honest answer to the manager was “we do not know”, and nobody enjoyed saying it.
Tuesday: reading what the experts say
The analyst spent the evening reading. She found the Atlas’s entry on SecDataOps, which frames the goal as evidence that investigators and auditors can rely on without overspending, and which opens by admitting that the term is an emerging, vendor-led label. That caveat made her trust the rest more.
One passage stayed with her. The Atlas recounts a 2023 intrusion into a government email service that was found only because of an access log available on a premium tier, plus a custom rule that someone had bothered to write. A later review recommended that such logs be available to every customer, and the vendor eventually moved them into its standard audit tier with six months of default retention. “So the evidence existed,” she told the team, “only for customers who had paid for it. Which tier are our logs on?” Nobody knew that either.
Wednesday: the register
The team spent a morning building a table of every system that wrote security-relevant records: the portal, the appointment scheduler, the email platform, the VPN, the file server, the wireless controller and the cloud console. For each, they added the owner, where logs went and how long they stayed. The Atlas’s name for it is a coverage register, with a zero-to-five score for five qualities: device reach, field completeness, timeliness, consistency and lifespan of the records.
The table was humbling. Of seven systems, two sent nothing to the archive, one overwrote its own records every week and one was owned by somebody who had left the organisation.
Thursday: contracts, not heroics
The coordinator had spent years writing procedures for paper records, so she took the lead on contracts. For each of the three most important sources she wrote a single page covering owner, required fields, how quickly records should reach the archive, how long they were kept, a periodic test event, the cost and known gaps. The network engineer grumbled about paperwork until the exercise exposed that the wireless controller had been quietly rebooted by a firmware update, which reset its logging level. The Atlas has a similar example: an operating-system upgrade that reset DNS query logging on two hosts, with certificate rotations on two more that bypassed the change board.
The fix was cheap, a line in the change checklist: “Does this change affect logging?” It took one minute to write and probably saved weeks of future confusion.
Friday: retention, argued out loud
The hardest conversation was about how long to keep things. The default on most systems was thirty or ninety days. The analyst shared the Atlas’s dwell-time figures from a Mandiant report on 2025 casework: a median of fourteen days globally, but much longer for some espionage-style cases and as many as about 400 days for one campaign. The team did not set a number that afternoon. They agreed to ask the compliance office, the legal department and their insurer, because requirements vary and nobody on the team was qualified to decide alone.
They did learn about one useful trap. The Atlas mentions that on one cloud plan the first thirty-one days of retention are already included, so reducing retention below that figure saves nothing. A cost conversation that had started with “let us shorten it” turned into “let us find out what we already pay for”.
The following month
The team connected the portal and two other sources to the archive, with raw logs kept in a low-cost tier and tidied summaries in the searchable one. They put a monitor on each source that alerted its owner after a stretch of silence, and they ran their first drill by switching off a test feed and timing the response. It took four hours. Next month it took twelve minutes. The analyst kept both numbers on a slide, which she used to win budget for a small storage upgrade.
What the story teaches
- You cannot investigate what you did not keep, and you cannot keep what nobody owns.
- A short written contract beats a long unwritten assumption.
- Change checklists should mention logging.
- Choose retention with the right people in the room, not alone.
For readers who maintain systems like these and want to build skills, our practice plan for DevOps interview rounds is a good companion read, and the general theme of preparation is also covered in our note on getting ready for a checkup appointment, which applies to paperwork readiness in the same spirit as a good log register.
This article is general information on administrative technology. It does not offer medical, legal or compliance advice, and it makes no claims about any hospital. Check requirements with your own advisers before changing any record-keeping practice.