If your company keeps buying security subscriptions and still feels exposed, the issue is probably not the amount of data coming in. It is what happens to that data once it arrives. This piece diagnoses five common symptoms of a ThreatOps routine that is not working, names the cause behind each, and gives a fix you can start on this week. No theory marathon, just problems and repairs.
A quick definition before the diagnosis begins. ThreatOps means running outside threat intelligence and active searching inside your own environment as a single operation with owners, a cadence and outputs. The Ops Atlas from xOps School treats it that way, while conceding that the term has no official standard behind it, which is why your team should record its own working meaning.
ThreatOps symptom one: the inbox is full of warnings nobody acts on
Cause. Information is arriving without a question attached. Someone subscribed to a feed because it sounded useful, and now alerts pile up with no decision waiting for them.
Fix. Write between five and eight standing questions, which the Atlas calls priority intelligence requirements. Each should ask one thing, need one fact and support one decision. “Are we exposed to attacks on our payment pages this quarter?” works. “Keep us safe from fraud” does not. Review the list every three months and retire questions nobody used. If you want a model for planning around outcomes rather than activity, our earlier article on turning digital growth planning into a working plan follows a similar logic of picking goals before tools.
Symptom two: a feed blocked something important
Cause. Someone switched a subscription from “alert” to “block” because it felt safer. Shared infrastructure, such as a content delivery network that many legitimate sites use, ended up on a bad list and customers lost access.
Fix. Treat blocking as a privilege that a source has to earn. The Atlas sketches a policy in which blocking is allowed only when the source is rated in its top two reliability grades, the credibility score is 1 or 2, the feed has proved at least ninety percent precision over ninety days, and the item is not on an allowlist. Anyone on call can undo a block immediately, and the intelligence owner reviews each reversal within an hour. Track how many blocks were reversed each month; if the number climbs, the feed drops back to alert-only.
The same source mentions a worked quarter where one feed turned out to be right on only eighteen of a hundred and twenty indicators, roughly fifteen percent. The verdict was to keep it alert-only and flag it as a candidate to cancel. That is how a subscription should be judged: on evidence, not on how impressive the brochure was.
Symptom three: reports pile up but nothing changes
Cause. The team measures output. Counting indicators ingested, reports published or hunts started feels like progress, and each number is easy to raise without improving anything.
Fix. Switch to ratios that end in an action. The Atlas gives examples from its worked quarter: the share of questions that had coverage, the share of intelligence items that led to some action, the share of hunts that produced a detection request or a gap ticket, and how many gaps actually got closed. You can calculate every one of these on a spreadsheet. Start with a single number, such as the fraction of items that led to a change, and publish it internally every month.
Symptom four: “we checked, we are fine”
Cause. A sweep against a list of known bad addresses came back empty and somebody declared the company unaffected. A list tells you about yesterday’s attacks on other people. It cannot rule out a visitor who uses ordinary accounts and remote-access tools.
Fix. Add behavioural hunts. The Atlas points to a US government advisory from February 2024 on a state-linked group, where traditional indicators were described as generally lacking while access had been maintained for years. Its advice by implication is to ask “what would normal-looking misuse of our accounts look like here?” and go search for that.
Practical hunt design from the same entry is refreshingly small. Pick a hypothesis, frame it as actor, behaviour, location and evidence, limit yourself to one or two behaviours, and repeat about every two weeks. End each hunt with one of four outputs: a request for a new alert rule, a ticket for missing data, an escalation, or a record that you would have noticed it had it happened. Then apply the acceptance test: could someone who did not take part repeat the hunt from your notes alone?
Symptom five: the work stalls when one person is away
Cause. Everything lives in one person’s head and one person’s browser tabs.
Fix. Keep the register of sources, the list of questions, hunt notes and expiry rules in a shared, version-controlled place. The Atlas says small teams can begin with a repository of plain files and delay buying a full platform. Pair that with a named owner and a short escalation rule so there is always somebody who can say yes or no.
A note on old data
Stale information deserves its own mention because it quietly inflates every other problem. A warning from last spring about an address that has since been reassigned to an innocent hosting customer is worse than no warning, since it sends people chasing ghosts and erodes trust in the whole routine.
One cause hides behind several symptoms: indicators that never expire. In the common sharing format, the “valid until” field is optional, and when it is missing the item has no time limit at all. The Atlas suggests setting your own lifetimes, for example two weeks for network addresses, a month for links, two months for domain names and a year for file fingerprints. Those numbers are an example policy, so adjust them to your own tolerance. What matters is that everything has an end date.
Where people actually get in
It is also worth keeping the questions realistic. The Atlas cites a 2026 industry report saying that phone-based social engineering, often called voice phishing, was the second most common initial route at eleven percent, and that the median time before an intruder was found rose to fourteen days from eleven. Read those as the entry reports them. For a general business audience the lesson is plain: your people and your accounts are the front door, so some of your five questions should be about them.
Know what belongs elsewhere
Scope creep is the quiet killer of small programmes. When every odd request lands on the same two people, the hunting slot gets eaten by ticket queues, and what the Atlas calls hunting as overflow for the alert desk takes over. Protect the fortnightly slot in the calendar the way you would protect a payroll deadline.
A final diagnosis tip: if you are tempted to use ThreatOps to cover everything, stop. The Atlas keeps round-the-clock alert watching, production alert rules, live incident handling and software-weakness ranking outside its scope. The handoff is deliberate: a hunt result that stands up turns into a request to whoever owns alert rules. Your small business may have one person doing all four jobs, but keep the hats separate in the paperwork so each output has an owner.
A thirty-day repair plan
- Week one: name an owner and write five questions.
- Week two: list every source you pay for or read, grade each for reliability, and put an end date on every indicator you keep.
- Week three: run a first hunt on a single behaviour and write down the result, even if the result is “nothing found, data was complete”.
- Week four: pick one ratio, publish it, and decide whether any subscription should be cancelled or kept alert-only.
If you are working through the business-technology side of this at the same time, our guide on choosing and sequencing business technology solutions is a useful companion for deciding what to buy and what to skip.
For the full background on terms and frameworks mentioned above, see the Ops Atlas page on ThreatOps. Read it as a reference and then shape the routine to fit your own size and risk.

