TL;DR
- Mandiant’s M-Trends 2026 puts global median dwell time at 14 days, up from 11, and attackers sat undetected for 26 days when an outside party had to tell the victim. Waiting for alerts is not the same as finding intruders, and that gap is what threat hunting closes.
- Hunting capability is a ladder, not a hire. David Bianco’s Hunting Maturity Model runs from HMM0 (alerts only) to HMM4 (hunts automated into detections), and its author is blunt that you can’t buy your way to the top, because analyst skill and data quality matter more than tooling.
- For most mid-market teams, the sensible path is decoys first, then data and procedures, then a managed hunting service for the rungs you can’t staff, which is exactly where the skills shortage makes building in-house unrealistic.
Why Hunting Is Not the Same as Alerting
Most security stacks are built around alerts: a tool spots something it recognises and tells a human. Threat hunting starts from the opposite assumption, that something has already got past the tools and someone needs to go and look for it. In David Bianco’s definition, hunting is the collective name for manual or machine-assisted techniques used to detect security incidents, and he is explicit that waiting for alerts, however sophisticated the tool, is not hunting.
The case for looking is in the dwell-time numbers. Mandiant’s M-Trends 2026 puts the global median at 14 days, and how an intrusion is discovered changes the picture sharply:
| How the intrusion was discovered | Median dwell time |
| Adversary announced themselves (usually a ransomware note) | 5 days |
| Organisation detected it internally | 10 days |
| External party notified the victim | 26 days |
| Cyber espionage and DPRK IT worker incidents | 122 days |
Mandiant also points to persistence on edge devices, which often lack standard telemetry, as one reason some intruders stay hidden for so long. The pattern is consistent: the sooner your own team finds something, the less time the attacker has.
The Five Rungs, From Decoys to Managed Hunting
Bianco’s model describes five levels of hunting capability. It is deliberately organisation-agnostic, so we’ve mapped it onto the practical steps we see mid-market teams actually take:
| Rung | What it looks like | Bianco HMM level |
| 1. Decoys and tripwires | Honeytokens, fake credentials and honeypots that alert on any touch, with no hunting skill needed | Sits alongside HMM0, adds a high-fidelity signal |
| 2. Alerting plus routine data collection | SIEM or XDR alerts, with enough logs retained to look backwards when something is reported | HMM0 moving to HMM1 |
| 3. Intel-driven searching | New threat reports are turned into indicator searches across historical data | HMM1 (Minimal) |
| 4. Repeatable hunt procedures | Analysts run documented hunts on a regular cadence, adapting published techniques | HMM2 (Procedural) |
| 5. Managed or automated hunting | A managed SOC or MDR provider hunts continuously; successful hunts become automated detections | HMM3 and HMM4 (Innovative, Leading) |
Our take: most organisations are at rung 2 and believe they are at rung 4. Having a SIEM or XDR platform makes you well instrumented, not a hunting team. The difference is whether a named person, or a managed partner, is proactively asking questions of the data on a schedule.
Rung 1: Start With Decoys
The cheapest, fastest early-warning signal does not need a hunter at all. In September 2026, CISA published guidance on using cyber decoys to strengthen detection and response, aimed at critical infrastructure owners and smaller security teams. It sets out the vocabulary clearly:
- Tripwires: monitored assets or actions that raise an alert on any interaction.
- Honeytokens: data with no legitimate business use, such as fake records, credentials or files, where any access signals unauthorised activity.
- Breadcrumbs: planted clues, such as saved credentials or mapped drives, that lead an intruder towards a decoy.
- Honeypots: full decoy systems built to attract and observe an adversary.
The appeal for a lean team is signal quality. A genuine user has no reason to touch a honeytoken, so a hit is almost always worth investigating, which is the opposite of the alert-fatigue problem we covered in our piece on why more tooling isn’t the fix for noisy SOCs.
Rungs 2 to 4: Data Before Hunters
Bianco identifies three factors behind hunting ability: the quality of the data collected, the tools to analyse it, and the skill of the analysts, with analyst skill the most important. Data comes next. HMM0 organisations collect little beyond what their alerting needs, so even a brilliant hire would struggle to hunt. Practically, that means:
- Collect enough, from enough places: endpoint, identity, network and cloud telemetry retained long enough to search backwards. This is the real argument in our post on whether you still need a SIEM, where the function matters more than the product label.
- Turn intelligence into searches: when a new threat report lands, someone should be able to check your historical data for those indicators within hours, not weeks.
- Document the hunts that work: repeatable procedures survive staff turnover, which matters given how fragile small security teams are.
- Start at HMM2: Bianco’s own recommendation for organisations getting into hunting is to aim for the procedural level first, running well-known hunts regularly rather than trying to invent new ones.
The catch is people. Every rung from three upwards needs analyst time, and that is precisely the resource the market is short of, as we set out in our look at the cyber skills shortage.
Rung 5: When to Stop Building and Start Buying
Bianco warns that although a good hunting platform can give your team a boost, you can’t buy your way to HMM4, because the inventive, curious human work isn’t something a product delivers on its own. That is an argument for buying hunting as a service, not for buying a tool and hoping:
| Question | Build in-house | Managed hunting |
| Analyst cover | Needs experienced hunters, ideally more than one, and out-of-hours cover | Provider’s analysts and global SOC hours |
| Data and tooling | You collect, store and tune the telemetry | Provider ingests and correlates it, often via an XDR or AI platform |
| Speed to value | Slower: hiring and maturing a team takes time | Faster: the provider’s analysts and platform already exist |
| Where it fits | Large teams with mature data and a real hunting skill base | Mid-market teams that can’t staff round-the-clock hunting |
Providers package this differently. Darktrace, for example, describes a managed tier with a proactive threat hunting component, backed by a 24/7 global SOC. ConnectProtect, which we introduced recently, combines managed XDR with a 24/7 SOC across three global locations. Either way, the test is the same: ask what hunting actually means in the contract, how often it happens, and what you receive when it does.
What We’re Seeing in Practice
What we see most often when this comes up with clients:
- Teams buy an XDR or SIEM platform and assume hunting comes with it, when in practice nobody has the time or mandate to proactively search the data.
- Decoys are almost always the quickest win: a handful of honeytokens in file shares and identity systems gives a small team a low-noise signal inside a week.
- The conversation usually turns on staffing rather than technology: the organisations that move fastest are the ones that decide early which rungs they will not staff themselves.
UK & New Zealand Perspective
Hunting maturity looks similar in both markets, but the constraints differ:
- UK: larger talent pool, but strong competition for experienced hunters and a growing expectation, from boards and insurers, that detection goes beyond alert response.
- NZ: a smaller candidate pool makes an in-house hunting capability harder to sustain, which is one reason managed, follow-the-sun models suit many Kiwi mid-market teams.
ITogether’s Independent Verdict
Threat hunting is not a product and not a single hire, it is a capability that grows in stages. The honest first question is not which hunting tool to buy, it is which rung you are really on, and whether you have the data, the skills and the time to climb the next one.
For most organisations between 50 and 5,000 staff, the practical route is to deploy decoys now, get your data collection in order, document the hunts you can run, and use a managed partner for the rungs where analyst depth and round-the-clock cover are the constraint. That is a more reliable path than trying to build a hunt team in a market that can’t supply one.
👉 Curious which rung your team is really on, and what the next one would take? Get in touch.
📞 UK +44 (0) 113 341 0123
📞 NZ +64 (0)9 802 2444
📧 hello@itogether.com
FAQs
What Is the Hunting Maturity Model?
A five-level model created by David Bianco, running from HMM0 (relies on automated alerts) to HMM4 (successful hunts are automated into detections). It gives organisations a way to measure where they are and a roadmap for improvement.
Do We Need a Dedicated Threat Hunting Team?
Not necessarily. Many mid-market organisations get better value from decoys, good data collection and a managed hunting service than from trying to recruit and retain dedicated hunters, particularly given the current skills shortage.
Are Honeypots and Honeytokens Worth Deploying?
For most teams, yes. They are low-cost and generate high-fidelity alerts, because legitimate users have no reason to touch them. CISA’s September 2026 guidance specifically recommends them for critical infrastructure and smaller security teams.
How Is Managed Threat Hunting Different From Managed Detection and Response?
MDR typically focuses on monitoring, triaging and responding to detections. Managed hunting adds analysts proactively searching your data for activity your tools haven’t flagged. Always check which of these a provider actually includes.

0 Comments