Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Mean time to detect measures the gap between an attacker starting work and your organisation noticing. It is the most quoted SOC metric and the easiest one to report misleadingly.

What Is Mean Time to Detect?
Mean time to detect is the average interval between the beginning of malicious activity and the point at which the organisation becomes aware of it. The appeal is obvious: a single number, trending in a direction, that a board can read. The difficulty is equally obvious once you try to calculate it, because the moment an attack started is only ever established retrospectively, and only for the incidents you eventually found.
Three definitions are in common use and they produce very different numbers.
Whichever you choose, document it, apply it consistently, and never compare your figure with a published benchmark that used a different definition.
IBM’s Cost of a Data Breach Report for 2024 put the average time to identify a breach at 194 days, with an additional 64 days to contain it, giving a lifecycle of 258 days. Other datasets, particularly those drawn from incident response engagements involving ransomware, show much shorter intervals, because ransomware announces itself. That divergence is the most useful thing in the data: aggregate figures mix incidents that end with an encryption note after four days with espionage that runs quietly for a year.
Break your own figure down the same way. Report it by detection source as well, separating incidents found by your own tooling from those reported by a third party, a client, or the attacker. An organisation whose headline figure looks respectable because most incidents were discovered externally has a detection problem that the average conceals.
Detection time is a function of three things in sequence: whether the telemetry exists, whether a detection covers the behaviour, and whether anyone acts on the alert. Work them in that order.
Telemetry first, because a rule cannot fire on data you do not collect, which is why endpoint telemetry coverage and log source criticality determine the ceiling. Then coverage of the techniques that matter, measured honestly on a coverage map rather than asserted. Then queue health, because a correct alert that waits six hours has not shortened anything, which is the practical cost of alert fatigue.
Pipeline latency deserves separate attention. If events reach your platform 30 minutes after they occur, every detection carries that delay before an analyst has done anything, and no amount of staffing fixes it.
Use the median alongside the mean. A single 300-day incident drags an average of twelve other cases beyond recognition, and the median tells you about the typical case while the mean tells you about the tail. Report both, with the count behind them, and resist quarter-on-quarter comparisons when the sample is five incidents.
Be alert to the perverse incentive. Any metric that rewards a lower number can be improved by detecting less: tune out the rules that find slow intrusions, and your remaining detections will be fast ones. Pair the figure with validated coverage and with the proportion of incidents discovered internally, both of which get worse when someone games the first.
Finally, state what the number does not include. It covers incidents you found. Whatever is currently undetected is excluded by definition, which is the strongest available argument for hunting. At Privia Security, we report detection time alongside coverage evidence in our SOC maturation services, because either figure alone can be made to look good.
There is no universal target, because it depends on the attack type and on how you define the start point. Ransomware precursors should be detected within hours, since the whole intrusion often runs in days. Quiet credential abuse is realistically measured in days to weeks. Rather than chasing an external benchmark, measure your own trend using a consistent definition and segment it by attack type.
They describe the same interval from different directions. Dwell time usually means the period an attacker was present before discovery, measured from initial compromise, and is the term used in incident response reporting. MTTD is the aggregate operational metric a SOC tracks across incidents. The practical difference is that dwell time is almost always measured from first compromise, while MTTD definitions vary.
Easily. Detecting fewer slow intrusions improves the average, as does measuring from first alert rather than from first compromise, and so does excluding incidents reported by third parties. Guard against it by publishing the definition, reporting the median alongside the mean, and pairing the figure with validated detection coverage and the proportion of incidents discovered by your own tooling.
Fix the queue before adding detections. In most organisations alerts already fire for part of the intrusion and wait hours for attention, so reducing noise, grouping alerts into cases, and automating enrichment produce faster gains than new rules. After that, address telemetry gaps, since no rule can detect behaviour on a host that sends you nothing.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.