Business management and professional decisions · Lesson 2

Assessing the risk of an HR change and choosing a control

8 min read

Lesson objectives

  • Describe a risk by cause, event and consequence rather than a vague topic.
  • Compare likelihood reduction, impact reduction and acceptance using a clear escalation threshold.
  • Test whether a control remains effective after a workflow change.
Reading tools

Suppose an organization lets line managers change employee contract end dates directly in the HR system to speed up renewals. This removes a central queue, but it also changes approval of a fact that affects payroll, access rights and staffing plans. The phrase 'system risk' does not support a decision. State the risk as a chain: broader edit rights may allow a manager to enter an incorrect date or change it without review; as a result, a payment may continue incorrectly or an employee's access may end on the wrong day. Cause, event and consequence are distinct. Confusing them can lead to a control that misses the cause.

HR makes an initial estimate of likelihood and impact without pretending it is a precise measurement. A simple register may rate likelihood low to high using edit volume and past corrections, and impact limited to severe using money, privacy and continuity. Multiplying two ordinal scores does not turn the risk into an exact monetary quantity. The score helps prioritize discussion when accompanied by confidence and missing evidence. A single wrong date that disables access to a critical system can have high impact even if it is uncommon.

There are three practical responses. Reducing likelihood means limiting who can edit, requiring a reason, checking that a date is plausible and sending a change to an independent reviewer before it takes effect. Reducing impact means detecting errors quickly through reconciliation between contract records and payroll, with an alert before access is withdrawn. Informed acceptance means a small residual risk remains within a threshold accepted by the decision owner; it is not permission to ignore the risk. If suitable controls are unavailable, pause the launch or escalate to the person with risk authority. A button in a system does not assume the organization's accountability.

How can the team tell whether the control works? Count changes, the share rejected or corrected before approval, and errors discovered after they affect pay or access. More early corrections could mean better detection; do not label every rise as worse manager input. Also compare renewal processing times so that control friction does not erase the intended benefit. If the same person enters and reviews a change, two clicks do not constitute independent review. Where a small team cannot separate duties, narrow edit rights and add frequent independent after-the-fact review while acknowledging the greater residual risk.

A preventive control can itself obstruct work. In the contract-date example, second approval may take three days although an employee needs renewal confirmation today. Neither deleting approval nor accepting unlimited delay is satisfactory. Classify edits by impact: a spelling correction with no pay or access effect differs from an effective-date change that alters payment or removes rights. With policy-owner approval, the first class might receive logged later review; the second needs a delegated reviewer and an absence backup. Trial ten cases of each kind and measure completion time, errors before and after approval, and bypasses. Faster average processing is not enough if consequential payroll errors rise. A second reviewer who sees only the same field, without the approved source and effective date, may add delay without detection. Record the remaining risk, its authorized owner and the condition that would stop wider rollout.

Add a test of the control itself to the risk register. A procedure may require review of every change while the reviewer approves everything at day's end without reading supporting records. The existence of a step differs from its ability to detect errors. Select earlier changes and ask the reviewer to repeat the check using authorized documentation. Record which errors can be detected and the time needed. If an impossible date is found but an incorrect approving authority is missed, the control covers less than expected. Clarify the comparison and identify who resolves conflicting evidence. A sample tests an operating procedure; extrapolating its error rate across all months requires checking representativeness. The sample can establish a specific defect without precisely measuring how widespread that defect is.

Lesson summary

Describe a risk by cause, event and consequence rather than a vague topic. Test whether a control remains effective after a workflow change.