Risk Matrix and Heat Map: How They Are Built, Step by Step
The two tables your supervisor asks to see, explained with a worked example.
A risk matrix crosses two scales and gives a level in each cell. The first crosses the probability of a risk event with its impact and gives the inherent risk. The second crosses that inherent risk with the rating of the controls and gives the residual risk. Painted by colours, with each event in its cell, it is the heat map.
Two Matrices, Two Questions
The law asks for both, and neither replaces the other.
The client risk matrix
How risky is this client? Each factor (country, activity, product, channel, PEP status) has a weight, the weights add up to 100 %, and the result places the client in a level that sets its due diligence and how often it is reviewed.
The entity risk matrix
How could someone use my business? It measures risk events by probability and impact, rates the controls that mitigate them, and consolidates the residual risk by factor and for the entity as a whole.
The heat map
The matrix of the entity painted by colours, with each event placed in its cell. At a glance, the board sees which events sit in red and need a plan.
The Scales
Everything starts with scales that your methodology defines. Four levels is the most common choice; this guide uses four.
Swipe the table sideways to see every column.
| Level | Probability | Impact | Control rating |
|---|---|---|---|
| 1 | Low: rare, few clients or operations involved | Low: minor and quickly corrected | Good |
| 2 | Medium-low: possible | Medium-low: noticeable, without a sanction | Adequate |
| 3 | Medium-high: likely | Medium-high: a sanction, a claim or a loss of clients | Improvable |
| 4 | High: frequent, many clients or operations | High: serious harm to the entity | Deficient |
In the control scale a higher number is a worse control. Each scale says how it is measured: probability with a count or a percentage of clients, alerts or operations in a period, impact on each associated risk.
The Two Heat Maps
Read a cell where a row and a column meet. In these examples each cell is the average of the two levels, rounded up; your methodology can fill each cell its own way.
Inherent risk: probability by impact
The risk of an event before any control. An event of medium-high probability (3) and high impact (4) is a high inherent risk.
| Probability | 1 Low | 2 Medium-low | 3 Medium-high | 4 High |
|---|---|---|---|---|
| 4 High likelihood | 3 Medium-high | 3 Medium-high | 4 High | 4 High |
| 3 Medium-high likelihood | 2 Medium-low | 3 Medium-high | 3 Medium-high | 4 High |
| 2 Medium-low likelihood | 2 Medium-low | 2 Medium-low | 3 Medium-high | 3 Medium-high |
| 1 Low likelihood | 1 Low | 2 Medium-low | 2 Medium-low | 3 Medium-high |
Columns: impact. Rows: probability, from high at the top to low at the bottom.
Residual risk: inherent risk by control rating
What remains after the controls. A high inherent risk (4) with an improvable control (3) stays high; the same risk with a good control (1) falls to medium-high.
| Inherent risk | 1 Good | 2 Adequate | 3 Improvable | 4 Deficient |
|---|---|---|---|---|
| 4 High | 3 Medium-high | 3 Medium-high | 4 High | 4 High |
| 3 Medium-high | 2 Medium-low | 3 Medium-high | 3 Medium-high | 4 High |
| 2 Medium-low | 2 Medium-low | 2 Medium-low | 3 Medium-high | 3 Medium-high |
| 1 Low | 1 Low | 2 Medium-low | 2 Medium-low | 3 Medium-high |
Columns: control rating. Rows: inherent risk. A better control (to the left) lowers the residual risk.
Discrete, Continuous or Arithmetic
Three common ways to obtain the residual risk. The methodology chooses one, and it is approved with it.
Discrete matrix
Whole levels only: the cell where the inherent level and the control rating meet gives the residual level. It is the one shown above, and the easiest to read for a board.
Continuous matrix
Each level divided into quarters (3.25, 3.50, 3.75…). It shows movement inside a level: a residual of 3.25 is still high, but close to medium-high.
Arithmetic
An average or a weighted average of the inherent risk and the control rating, without a table. The result is turned into a level with the same scale.
A Worked Example
One event, from its probability to the plan, with the conservative rounding: a figure with decimals rises to the next level.
The event
Clients who pay in cash amounts that do not match their declared activity.
Probability: 3
18 % of the clients paid in cash in the period. The methodology places 15 % to 25 % in level 3, medium-high.
Impact: 3
Reputational 4, legal or compliance 3, operational 2, contagion 2. The average is 2.75, which rises to 3.
Inherent risk: 3
Probability 3 and impact 3 meet in the medium-high cell of the inherent risk matrix.
Controls: 3
Two controls: a cash limit rated 2 and a review of the source of funds rated 3. The average is 2.5, which rises to 3, improvable.
Residual risk: 3, and the plan
Inherent 3 and control 3 give a medium-high residual. If the methodology asks for additional actions from medium-high, the event goes to the strengthening plan.
How COX Builds Them
Your methodology sets the scales and the cells; COX computes and places each event.
Computed, never typed
At a cut-off date COX computes the probability, the impact, the inherent risk, the rating of the controls and the residual risk of each event. The same data always gives the same map.
Probability counted from your files
PEPs, foreign nationals, legal entities, high-risk clients, countries on the FATF lists and non-face-to-face channels are counted from the files you already have in COX.
Each event in its cell
The heat map places each event by its inherent risk and its control level. Events without controls go apart, and those that could not be computed are named, never dropped.
Rounded up, the conservative reading
A figure with decimals rises to the next level: a 2.1 is already level 3. For the combined rating of several controls, the methodology chooses up or nearest.
Controls rated with their tests
A control is rated by whether it is documented, preventive or detective, how often it runs, and the compliance and effectiveness its registered tests measured.
In the report and in Excel
The same heat map goes in the sealed report to the board, and the matrices are exported to Excel. An empty figure stays empty: the map never shows a zero that nobody measured.
Frequently Asked Questions
Build Your Heat Map From Your Own Files
Open your account: the methodology starts from a reference, and the assessment counts from the clients you already have.