Key Risk Indicators (KRIs) are metrics that give you early warning that a risk is increasing, while there is still time to do something about it. A KRI is not a report on what went wrong last quarter. It is a signal that something is drifting now.
This guide covers what a key risk indicator is, how KRIs differ from KPIs and KCIs, how to define them, how to set thresholds that mean something, and how to wire them into your risk register so they drive action rather than decoration. It includes a library of worked KRI examples by risk category you can adapt directly.
What Is a Key Risk Indicator?
A key risk indicator is a measurable value that tracks how exposed you are to a specific risk. Three words in that sentence carry the weight:
- Measurable. It comes from a data source you can query repeatedly without a project each time.
- Specific. It maps to one named risk or control, not to a vague theme like "cyber".
- Exposure. It tells you about the likelihood or impact of something that has not happened yet.
The last point is what separates a KRI from ordinary management information. "Revenue this month" is a fact about the past. "Number of customers representing more than 10% of revenue" is a statement about how badly a single loss would hurt you — a risk indicator.
A good KRI answers the question: what would I want to know early, so that I could still act?
KRI vs KPI vs KCI
These three get used interchangeably and they are not the same thing. The distinction matters because it determines who owns the number and what happens when it moves.
| Measures | Question it answers | Example | |
|---|---|---|---|
| KPI — Key Performance Indicator | Performance against an objective | Are we achieving what we set out to do? | Customer onboarding time: 4.2 days |
| KRI — Key Risk Indicator | Exposure to a risk | Is a risk getting worse? | Onboardings completed without full KYC: 12 |
| KCI — Key Control Indicator | Whether a control is working | Is the control we rely on actually operating? | KYC checks completed within SLA: 87% |
The same underlying data often feeds all three. What changes is the framing and the response. A KPI miss triggers a performance conversation. A KRI breach triggers a risk conversation. A KCI failure triggers a control remediation.
A practical test: if the number worsens and your reaction is "we need to try harder", it is a KPI. If your reaction is "we may need to change our exposure, our controls, or our appetite", it is a KRI.
What Makes a Good KRI
Most KRI programmes fail not because the metrics are wrong but because they are unusable. Five criteria separate a KRI that gets acted on from one that gets ignored:
- Relevant. Tied to a named risk in your register. If you cannot say which risk it belongs to, it is a statistic, not a KRI.
- Measurable from a real source. Systems, tickets, surveys, audit results. If collecting it needs someone to chase five people by email each month, it will quietly stop being collected by month four.
- Leading or coincident, not purely lagging. Overdue access reviews rise before the breach. Number of breaches is a body count.
- Actionable. There is a defined response at each threshold. A metric with no defined response is a metric nobody reads.
- Owned. One named person is accountable for the number and for acting when it moves.
That last criterion does most of the work. An unowned KRI has no reason to be accurate.
How to Define KRIs: Six Steps
1. Start from the risk, never from the data. The failure mode is starting with what is easy to measure and reverse-engineering a justification. Take your top ten risks and work down.
2. Ask what would change first. For each risk, ask what would move measurably before the risk materialises. For "loss of key supplier", that might be supplier financial health scores, late-delivery rates, or concentration percentage — all of which move before the supplier fails.
3. Pick one or two indicators per risk. Not eight. A risk with eight indicators has no indicator, because nobody knows which one matters.
4. Define the metric precisely. Write down the exact formula, the system it comes from, the population it covers, and the measurement frequency. "Number of incidents" is ambiguous. "Count of P1 and P2 incidents raised in ServiceNow in the calendar month, excluding those later reclassified as non-incidents" is a metric.
5. Set thresholds with the risk owner. See the next section — this is where most of the value and most of the argument sits.
6. Assign an owner and a review cadence. Monthly for most operational KRIs, daily or weekly for fast-moving ones like security or liquidity, quarterly for slow structural ones like supplier concentration.
KRI Examples by Risk Category
The fastest way to start is to adapt indicators that already work elsewhere. Below are worked examples across the categories most organisations carry. Treat them as candidates, not a checklist.
Cybersecurity and Information Security
| Key risk indicator | Why it leads |
|---|---|
| Critical vulnerabilities open beyond SLA | Exposure window before exploitation |
| Average patch lag on internet-facing systems | Time attackers have to use a known exploit |
| Overdue user access reviews | Accumulating unnecessary privilege |
| Privileged accounts without MFA | Single credential compromise becomes total |
| Phishing simulation click rate | Human susceptibility ahead of a real campaign |
| Endpoints missing current EDR agent | Blind spots in detection coverage |
Operational Resilience
| Key risk indicator | Why it leads |
|---|---|
| Critical processes without a tested recovery plan | Untested plans fail on first use — see our BIA guide |
| Recovery tests overdue against schedule | Confidence decays as systems change |
| Single points of failure identified but unmitigated | Known fragility carried forward |
| Unplanned downtime minutes on tier-1 systems | Trend toward instability |
| Backlog of unresolved incidents older than 30 days | Systemic capacity problem |
Third-Party and Supplier Risk
| Key risk indicator | Why it leads |
|---|---|
| Revenue or service concentration with one supplier | Magnitude of a single failure |
| Critical suppliers without current due diligence | Unknown exposure — see TPRM |
| Suppliers with deteriorating financial scores | Distress precedes failure |
| Contracts expiring within 90 days without renewal plan | Forced acceptance of bad terms |
| Fourth-party dependencies unmapped | Concentration you cannot see |
Compliance and Regulatory
| Key risk indicator | Why it leads |
|---|---|
| Overdue control tests | Assurance gap widening |
| Policy attestation completion rate | Awareness before behaviour |
| Open audit findings past remediation date | Known issues left unfixed |
| Regulatory changes unassessed for impact | Surprise obligations |
| Breaches of reporting deadlines | Process strain before a reportable failure |
Financial and Credit
| Key risk indicator | Why it leads |
|---|---|
| Exposure to a single counterparty as % of capital | Concentration ahead of default |
| Limit breaches per month | Discipline erosion |
| Debtor days beyond terms | Collection deterioration |
| Liquidity coverage against 30-day stress | Runway before a squeeze |
People and Conduct
| Key risk indicator | Why it leads |
|---|---|
| Voluntary attrition in critical roles | Key-person exposure |
| Vacancies open beyond 90 days in control functions | Control capacity gap |
| Mandatory training completion rate | Competence before error |
| Whistleblowing reports by theme and trend | Culture signal ahead of conduct failure |
Setting Thresholds That Mean Something
A KRI without thresholds is a chart. Thresholds turn it into a decision.
The convention is three bands:
- Green. Within appetite. Routine monitoring only.
- Amber. Approaching the limit of appetite. Review, investigate, and decide whether to act.
- Red. Beyond appetite. Escalate and act; doing nothing must be an explicit, recorded decision.
There are three defensible ways to set the numbers, and one indefensible one.
From risk appetite (best). Work backwards from the loss you have said you will tolerate. If the board has agreed you will not accept more than four hours of downtime on a tier-1 system, your amber sits at the point where that becomes likely.
From historical distribution. Use your own data: set amber at, say, the 90th percentile of the last two years and red at the 98th. Honest about normal variation, but it anchors on how bad things have been rather than how bad you are willing to let them get.
From external benchmark or regulation. Where a regulator or standard specifies a limit, use it — and set amber inside it so you get warning before you breach.
By picking a round number that feels about right. This is the indefensible one, and it is what most organisations actually do. A threshold nobody can justify is a threshold nobody defends when it is breached; the conversation becomes about whether the threshold was fair rather than about the risk.
Whichever you use, write down the rationale next to the threshold. When someone challenges an amber six months from now, that sentence is what saves the discussion.
Leading, Coincident and Lagging Indicators
Most KRI sets are lagging and their owners have not noticed. A quick way to audit yours:
- Leading. Moves before the risk event. Patch lag, overdue reviews, supplier financial scores. These are the valuable ones and the hardest to find.
- Coincident. Moves as the risk materialises. System availability, error rates. Useful for detection, too late for prevention.
- Lagging. Moves after. Losses incurred, incidents closed, fines paid. Necessary for validating that your risk assessment was right, useless as a warning.
Count yours. If more than half your KRIs are lagging, you have a reporting pack, not an early-warning system.
Wiring KRIs into Reporting and Escalation
An indicator only matters if breaching it causes something to happen. Three connections make that real:
- To the risk register. Each KRI links to one or more risks. When the indicator moves into amber, the risk's likelihood rating and treatment plan get revisited — not at the next annual review, but then.
- To escalation. Amber and red trigger a defined route: who is told, within what time, and who can accept the position. Automate this or it will depend on somebody remembering.
- To the board pack. Trends, not point-in-time values. A KRI that has been amber for five consecutive months is a different conversation from one that went amber last week.
Common Mistakes
- Too many indicators. Forty KRIs means nobody reads any of them. Ten that drive decisions beat forty that fill a slide.
- Measuring what is easy. The data you already have is rarely the data that gives warning.
- All lagging. Covered above, and the most common failure by far.
- No owner. Unowned metrics rot.
- Thresholds nobody can justify. The breach conversation turns into a threshold conversation.
- No link to the risk register. The KRI becomes a parallel universe of numbers with no consequence.
- Never revisited. Risks change, businesses change, and a KRI defined three years ago may now be measuring something that no longer matters. Review the set annually.
Frequently Asked Questions
What is a key risk indicator?
A key risk indicator is a measurable metric that provides early warning that exposure to a specific risk is increasing. It is tied to a named risk in the risk register, comes from a repeatable data source, and has defined thresholds that trigger review or escalation when breached.
What is the difference between a KRI and a KPI?
A KPI measures performance against an objective and answers "are we achieving what we set out to do". A KRI measures exposure to a risk and answers "is a risk getting worse". The same data can feed both; the difference is the response. A KPI miss prompts a performance conversation, a KRI breach prompts a risk conversation about exposure, controls or appetite.
How many KRIs should we have?
One or two per significant risk, which for most organisations means ten to twenty in total at board level. Programmes with forty or more indicators typically find none of them get acted on, because attention is spread too thin to notice which numbers matter.
How do you set KRI thresholds?
The strongest method is to derive them from stated risk appetite, working backwards from the loss level the board has agreed to tolerate. Alternatives are the historical distribution of your own data, or an external regulatory or benchmark limit with amber set inside it. Always record the rationale alongside the threshold.
What is a leading key risk indicator?
A leading KRI moves before the risk event occurs, giving time to intervene — for example patch lag rising before a breach, or supplier financial scores deteriorating before a supplier fails. Lagging indicators such as losses incurred or fines paid only confirm what has already happened.
How often should KRIs be reviewed?
Measurement frequency depends on how fast the risk moves: daily or weekly for security and liquidity, monthly for most operational risks, quarterly for structural ones like supplier concentration. Separately, the whole KRI set should be reviewed annually to retire indicators that no longer reflect the risks the business actually carries.
Making KRIs Operational
The reason KRI programmes drift is almost always mechanical: the indicators live in a spreadsheet, the risk register lives somewhere else, and keeping them aligned is manual work that loses to more urgent tasks.
When KRIs sit in the same platform as the risk register and controls, the link maintains itself. Thresholds are configured once, breaches raise actions automatically against the right owner, and the board sees trends rather than a snapshot assembled the week before the meeting.
See how ActiveERM handles indicators, thresholds and escalation on our Risk OS and GRC Cloud pages, or read our guides to risk matrices and appetite and business impact analysis.