Marguerite ran operations for a six-person online shop that sold handmade lamps, and she remembered the Monday the group chat caught fire. A security notice had landed overnight about a flaw in a logging component, and everyone had the same question at once: do we use that thing? Nobody knew. What followed was four days of guessing, spreadsheets and cold coffee. The people in this story are invented, a composite of situations small teams describe, but the lesson at the end is the practical heart of VulnerabilityOps, the habit of handling software flaws as a steady flow with owners and proof instead of an occasional panic.
Monday: the question nobody could answer
The shop’s developer, a quiet contractor named Teodor in this retelling, started by searching the code repository. That found the obvious uses. It did not find the library hiding inside a third-party plug-in, or the one bundled in a reporting tool the bookkeeper ran on her laptop. By lunchtime the team had three different lists of “where we might be exposed”, each written by a different person, and none matched.
The real-world version of this scramble was enormous. A later public review of the December 2021 logging-library crisis noted that one large government department spent about 33,000 hours responding, and that organisations holding software bills of materials reported none of them had actually used those documents to find the affected library. The bar suggested afterwards was blunt: you should be able to say where a given component lives within minutes, not days.
Tuesday: ordering the pile
Marguerite did what many owners do and asked for a ranking. Teodor opened the scoring pages and found that nearly everything on the list carried a high severity label. Reading further, he found a line in the scoring body’s own guidance that base scores should not be used alone to assess risk. The label told him how damaging a flaw might be in the worst case. It said nothing about whether anyone was exploiting it today.
So they added two questions. Is this flaw on the public list of vulnerabilities known to be exploited in the wild? And is the affected system reachable from the internet at all? Those two answers cut a list of dozens down to three items for the week. The rest were not ignored; they were given later, written deadlines.
Wednesday: the ticket that said “done”
On Wednesday afternoon Teodor marked the main fix as resolved and the team exhaled. Thursday morning, Marguerite asked a question that changed their process: how do we know? Teodor had updated the package and closed the ticket. Nobody had re-scanned the live shop with a logged-in check, and nobody had confirmed that the old build was gone from the server that actually took customer orders.
That small question captures one of the discipline’s firmest rules. A flaw is finished when evidence says so: a fresh authenticated scan, a configuration test, a new container image fingerprint, or a formally upheld dispute. A ticket status is a note about intentions. When they ran the scan, one staging copy of the shop still carried the old component, and it had a public address nobody remembered creating.
What the Atlas says about this pattern
When Marguerite went looking for a name for what she wanted, she found a short Ops Atlas page defining VulnerabilityOps. She liked that it described a loop rather than a product: sort, assign, verify, measure. It also admitted, honestly, that no standards body owns the term. She printed the one-line scope list and taped it above her monitor.
Friday: writing down the rules
The team spent Friday turning the week’s chaos into a small policy, and it fit on one page.
- Every system gets a tag: public, internal, or “touches money”. Anything untagged counts as public until someone proves otherwise.
- Flaws that are known to be exploited and sit on a public system get a very short clock, measured in days, and a named person.
- Everything else gets a longer clock, and the clock does not restart when a scan is re-run.
- A finding leaves the list only with proof. The person who did the fix never approves their own closure.
- Once a month, the team looks at how many new items arrived versus how many were verified closed.
That last line took them longest to appreciate. The Atlas material describes a ratio of arrivals to verified closures per tier, with anything over one meaning the queue is filling. Marguerite’s team had never counted arrivals at all; they had only counted what they cleared, which made them feel steadily productive while the shelf of old problems quietly grew.
The following month: the quiet win
Four weeks later a different notice arrived, this time about a popular edge appliance that one of the shop’s suppliers used. The shop did not run the appliance, but the new inventory made that answer take nine minutes instead of four days. Teodor wrote to the supplier, asked what they had done beyond installing the update, and learned that the supplier had also ended all existing logged-in sessions, because the real-world advice for the similar 2023 appliance incident said exactly that: a patch alone leaves hijacked sessions alive.
The shop’s own staging box showed the other lesson. A scheduled scan reported a huge drop in findings after a password change, and for one pleasant afternoon the report glowed green. Marguerite, newly suspicious, noticed that the scanner’s login had failed. Fewer findings did not mean a safer shop; it meant a blindfolded scanner. They added an alert for credentialed coverage and never celebrated a sudden drop again without checking it first.
An argument that taught them something
Not everything in the new routine went smoothly. In the second month the bookkeeper objected to a deadline on her reporting tool, pointing out that the flaw in question could only be triggered from the office network and that the vendor had no fix yet. Marguerite’s old instinct would have been to wave her through or to shout. Instead the team used the dispute path they had written: the bookkeeper stated her reasoning on a single form, Teodor tested whether the claim was true, and a third person approved or rejected it. The claim held up, the finding was marked as not exploitable in that setting, and the decision was recorded with a review date.
The detail that mattered was who approved. Allowing the owners of a system to certify their own exemption invites wishful thinking, so the team made a simple rule that the approver is never the person whose workload gets lighter. It is a tiny piece of governance, and it kept the exceptions list short enough that everyone could still read it in one sitting.
Reading list from our own archive
If you run a small digital business and want broader context, our essay on choosing and organising business technology covers the purchasing side, while a companion look at finding spending leaks and protecting digital payments pairs naturally with a security routine like this one. Both rest on the same instinct: know what you own, know where the leaks are, and check that your fix actually fixed something.
What the story is meant to show
None of Marguerite’s steps needed a large budget. She needed an inventory that a person could query quickly, a ranking that looked at evidence of real attacks rather than labels alone, owners with dates attached, and closure that required proof. Large organisations with dedicated security staff learned the same things the expensive way, and the numbers published about them are sobering. If you run something small, you get to learn from their bills rather than your own.

