The team behind a fictional online directory of advocates had a problem they could not see. Four people looked after thousands of profile pages, a search feature, a contact-request form and a handful of servers. Each Monday someone pasted a long scan report into a shared sheet, and each Friday someone else said it was too long to read. Nothing had gone wrong yet, which was exactly the danger. This is an invented story, assembled from the kinds of situations small web teams describe, and it follows how they learned VulnerabilityOps the practical way: by replacing a pile of findings with a flow that had owners, dates and proof.
The sheet nobody trusted
Priya, who handled operations in this telling, opened the shared sheet one morning and counted 1,900 rows. The oldest were fourteen months old. Roughly a third were marked “in progress” by people who had since changed roles. She asked a simple question in the team chat: which of these could hurt us this week? Nobody could answer, because the sheet sorted by a severity number and the top of the list was packed with items that needed no attention from anyone outside the building.
The scale of the problem was not unusual. Public CVE counts nearly tripled over five years, and an often-quoted benchmark of about 300 companies found the average one fixing roughly one open weakness in ten each month. A team of four does not out-work arithmetic like that. They need to choose.
Choosing: three questions per finding
Priya’s first change was to stop sorting by severity alone. Even the group that stewards the severity score cautions that a base figure is one ingredient in a risk judgement, never the whole recipe. She asked for three answers per item instead. Is the flaw on the public catalogue of vulnerabilities known to be exploited? How likely is exploitation, according to the free probability score? And does the affected system face the public internet, or sit behind several layers?
Using those questions, the 1,900 rows collapsed. Eleven items were on the exploited catalogue. Four of those touched systems reachable by anyone, and one of the four was in the contact-request form, which handled enquiries about people seeking representation. That single finding became the week’s work. The rest got dates, longer ones, written next to a named owner.
What the reference page added
The team wanted a shared vocabulary, so Priya shared the Atlas page about VulnerabilityOps in the chat and asked everyone to read just the scope section. It made two points that stuck. First, it treats the work as a loop: tiers, an owner queue, closure that is verified, and metrics describing movement. Second, it deliberately excludes the mechanics of rolling out patches, which they already did well. Their weakness lay upstream and downstream of that step, in deciding what to do and in proving it had worked.
The ticket that said “done”
On Thursday, Imran, the developer, updated the form’s library and closed the item. Priya asked for the scan. Imran had not run one. When they did, using a logged-in scan against the live server, the old library version still answered on one older host that had been forgotten during a migration. A ticket status would have let that sit for months.
That episode turned into a rule: a finding closes only when one of four things exists. Either a credentialed rescan no longer shows it, a configuration test passes, a rebuilt image carries a new digest, or a challenge was upheld by someone other than the owner. They added a field to the ticket for the evidence link. Closing without it was simply not possible.
The quiet dashboard
A month later the weekly report showed 300 fewer findings, and the team nearly celebrated. Priya, now suspicious of good news, checked how many hosts had been scanned with working credentials. The scan account’s password had expired. Without a login, the scanner saw only what an outsider sees, so fewer problems were reported. They wired up an alert on how many hosts had been scanned with a working login in the preceding week, and agreed that any abrupt fall in findings gets investigated before anyone announces it.
A related worry, raised by the hosting provider, was that the scan results themselves were a map of the directory’s soft spots, and that the scanner’s login was a powerful credential. The team moved both into restricted storage with access logging.
Deadlines that came from facts
The next improvement was to replace “high means 30 days” with a short table. They borrowed from a US federal directive of June 2026. It applies to federal civilian agencies alone, yet its layout is tidy: a deadline worked out from reachability, catalogue listing, scope for scripting the attack, and the extent of technical damage. For their own directory they chose three days for exposed, listed, automatable cases and a few weeks for exposed, unlisted ones, with internal systems waiting for scheduled upgrades. Where information was missing they defaulted cautiously: unknown exposure meant public.
Counting what mattered
At the monthly review, Priya showed two numbers per tier: arrivals and verified closures. For the lowest tier, 600 items arrived against 520 closures, so that tier was losing ground at a ratio of roughly 1.15, whatever the sheet implied. She showed how old the stale items were too, but only beside the arrival numbers, having read the Atlas caution on the subject. In that example the oldest slice of the queue appeared to improve, falling from 30 weeks to 12, even as 25 additional items accumulated weekly.
The team set a modest goal for the next quarter: bring that tier’s ratio below one. They did not set a goal of “zero criticals”.
The argument about the old host
The forgotten host from the migration caused the team’s only real argument. Imran wanted to switch it off immediately; Priya wanted to know what it was still doing. They settled it with the dispute route they had just invented. Imran wrote down the case for shutdown, a colleague from outside the team reviewed the claim, and the decision was recorded: the host was a leftover and was removed within the day. What mattered was less the outcome than the rule behind it, that the person proposing a change does not also approve it.
Not every case was as simple. A week later a separate finding on an internal reporting tool could not be fixed because the vendor had not released an update. The team filed an exception with a stated reason, a compensating control that blocked outside access, and a date on which it would be reviewed. Because the exceptions were dated and counted, they could see in the monthly review that the number was small and stable, which is the healthy pattern. A growing pile of exceptions would have been a warning that deadlines were being moved rather than met.
The team also learned to ask a better question of suppliers. Rather than “have you patched?”, they asked what evidence the supplier could show that the fix was effective, and whether any extra steps, such as ending existing sessions after the 2023 gateway-appliance incident, applied. Suppliers who could answer quickly turned out to be the ones with the calmest incidents.
What the story is not
It is not a claim that a four-person team can handle any amount of work. It is a claim that a small team can choose better, prove more and measure honestly. Nor is it legal advice or a statement about any real directory. The pages linked below are part of this site’s own directory of advocates by city, shown here only as examples of the sort of large, repetitive content a team like this one must look after, for instance the intellectual property advocates listing for Guwahati and the matching page for Madurai.
Lessons the team wrote on the wall
- Sort by evidence, then probability, then severity.
- Every finding gets a named owner and a date.
- Proof closes a finding; a status does not.
- A sudden improvement is a question, not an answer.
- Compare arrivals with verified closures every month.
Related posts:

Memory, Full Notes or Three Lines? Recording a Legal Meeting

State of Play: Preparing a Consultation Folder in Seven Days

Six SRE Myths Worth Dropping if You Run a Legal Directory

A Three-Line Note Habit After Legal Appointments, Step by Step

A Ten-Point Audit of Your Client-Facing Practice Habits
