Where the bias enters
The discrimination in performance management rarely comes from deliberate intent. It enters through structural choices that appear neutral but produce systematically unfair outcomes. Understanding the main mechanisms is the first step to addressing them.
Forced distribution and rating curves
Requiring a fixed percentage of employees to be rated 'below expectations' - regardless of actual performance - guarantees that someone receives a poor rating whether or not they deserve one. In any diverse workforce, the employees least likely to receive the benefit of the doubt in ambiguous rating decisions are disabled employees, women, people from marginalised ethnic backgrounds, and those with less visible relationships with senior managers. Research by the CIPD and others consistently shows rating gaps that cannot be explained by output measures alone. If your rating distribution requires poor performers to exist, it will find them - in the groups least able to challenge the decision.
A single performance standard without reasonable adjustments
If all employees are assessed against identical standards and one employee has a disability that affects their output in measurable ways, assessing them against that standard without adjustment is almost certainly a failure of the duty to make reasonable adjustments under sections 20–21 of the Equality Act 2010. This is not a theoretical risk. It is one of the most common grounds on which performance-related Tribunal claims succeed: the employer applied a standard, the disabled employee fell short, but the employer had never asked whether adjustments were needed to enable them to meet it.
Recency and relationship bias in calibration
Without structured criteria, calibration sessions reward people who are visible, vocal, and culturally similar to their managers. Employees who are less visible - remote workers, part-time workers, people who have recently returned from parental leave or sick leave - are disproportionately affected. The person who springs to mind as a strong performer is often the person whose face is most familiar, not the person whose outputs are most impressive.
The legal exposure
A performance-related decision - whether a rating, a Performance Improvement Plan (PIP), a demotion, or a dismissal - can give rise to Employment Tribunal claims under multiple heads of discrimination simultaneously. The legal landscape is broad:
- A disabled employee placed on a PIP without first discussing reasonable adjustments has a potential failure-of-adjustment claim under ss.20–21 EqA 2010 - and potentially a direct discrimination claim if the PIP reflects the employer's failure to support them rather than genuine underperformance.
- A woman returning from maternity leave who receives a lower rating because she was absent during a key project has a potential maternity discrimination claim under s.18 EqA 2010. Rating someone against targets set during their absence - without adjustment - is a recurring pattern in Tribunal claims.
- An employee from a Black or other marginalised ethnic background who consistently receives lower ratings than white colleagues at comparable output levels has a potential race discrimination claim under s.9 EqA 2010.
The fact that your process is applied consistently does not protect you. Consistent application of a neutral-seeming standard that produces discriminatory outcomes is indirect discrimination - and it requires objective justification. "We apply the same criteria to everyone" is not justification. It is a description of the problem.
How to audit your process
A discrimination audit of your performance management system has three steps. None of them require sophisticated data infrastructure - most HR teams can run this with what they already have.
Step 1: Run the data
Pull your rating distributions by gender, ethnicity, disability status, part-time working, parental leave history, and whether someone recently returned from extended absence. If the distributions are statistically different - for example, women are disproportionately rated 'meets expectations' rather than 'exceeds expectations' compared to men at equivalent seniority - you have a problem that requires explanation. If the gap is not explicable by role differences or output data, it almost certainly reflects a structural bias in the process.
Step 2: Audit the criteria
Are your performance standards clear, specific, and capable of being applied consistently by different managers? Or are they vague descriptors that allow managers to fill in the blanks with their own judgements about cultural fit, communication style, and relationship quality? A criterion like "demonstrates leadership behaviours" tells different managers different things - and will produce different outcomes for different groups. A criterion like "delivers agreed outputs on time and within agreed parameters, and supports team members in doing the same" is specific enough to apply consistently.
Step 3: Audit your calibration
Are calibration sessions structured with explicit criteria? Are managers required to anchor each rating to specific examples of work and output, rather than general impressions? Is there a mechanism for challenging ratings that seem inconsistent - without that challenge being experienced as an attack on managerial judgement? If the answer to any of these is no, your calibration process is producing outcomes that are determined by the confidence and relationships of the managers in the room rather than the performance of the employees they're rating.
Running calibration well
A well-run calibration session is one of the most powerful inclusion tools available to any organisation. Run badly, it amplifies every existing bias. Before any session, share anonymised data showing rating distributions by team and demographic so facilitators can identify potential inconsistencies before discussion begins. During the session, require each rating to be supported by specific examples of work, output, or behaviour - not general impressions. When a manager says something like "she's just not quite leadership ready," challenge them: what specific behaviours or outputs support that assessment, and how do they compare to the benchmarks for the rating above? After calibration, review the overall outputs. If the distribution has produced wider demographic gaps than the previous year, investigate before publishing ratings.
Legal requirement: Before issuing a Performance Improvement Plan to a disabled employee, you must first ensure all reasonable adjustments are in place. Failure to do so can result in a claim of failure to make reasonable adjustments under ss.20–21 of the Equality Act 2010 - regardless of whether the PIP itself is otherwise justified.
PIPs: the highest-risk moment
Performance Improvement Plans are the point in the process where legal risk is highest and where poor practice is most common. Several patterns create serious exposure:
- A PIP placed on a disabled employee without first asking whether reasonable adjustments are in place is almost certainly a failure of the adjustments duty. The conversation must happen before the PIP is issued, not as part of the PIP process.
- A PIP issued shortly after an employee raises a grievance, makes a whistleblowing disclosure, or takes protected leave can constitute victimisation under s.27 EqA 2010 - even if the performance concerns predate the protected act.
- A PIP that sets targets the employee has no realistic chance of meeting - because the underlying issue is a medical condition, a need for support, or a reasonable adjustment that hasn't been made - is setting up an unfair dismissal claim as well as a discrimination claim.
The rule of thumb: before you issue a PIP, ask whether the performance issue is genuinely about capability, or whether it's about an absence of support, an unaddressed adjustment need, or a management relationship that requires a different intervention. If it's any of the latter, a PIP is the wrong tool.
What good looks like
Good performance management is transparent, frequent, specific, and development-focused. It provides documented feedback on a regular basis - not annual surprises that the employee has no opportunity to respond to. It explicitly accounts for absence, leave, and reasonable adjustments in how targets are set and how outcomes are assessed. It separates performance from visibility - some of the most productive employees are the least visible, and good systems recognise that. It trains managers to give feedback in a way that is anchored in behaviour and output rather than personality and fit. And it reviews its own outputs regularly to check that it is producing fair and consistent results across the workforce.
None of this is unachievable. It requires investment in manager capability, structured criteria, and meaningful data review. But the alternative - a performance management process that is producing discriminatory outcomes, creating Tribunal risk, and driving the exit of capable employees from your organisation - costs far more.