Guide · Job evaluation

Job evaluation bias: Where it hides

Most of the bias in a points plan is not in how jobs are scored. It is in how much each factor is allowed to count.

Short answer: Gender bias in job evaluation hides in the factor weights more often than in the scores. A panel can score every job fairly yet get a biased order if the plan counts the demands of mostly male jobs twice, gives them more levels or gives the demands of mostly female jobs too little weight to move a grade. Test the weights as well as the scores.

Every score agreed, two grades apart

Picture a 260-person company in Sacramento that services and repairs medical imaging equipment for hospitals. It worries about gender bias in its job evaluation, so it does what careful companies do. It sets up a panel of six people from different functions, trains them, writes up every job and scores each one against a points plan. The plan came from the operations director's previous employer, a manufacturer, where it has been used for years.

The panel works through 70 jobs over three weeks. The discussions are good. Scores are challenged, evidence is checked and every score has a written reason. Nobody can point to a job that was scored unfairly.

Then the results come back. Field service engineers, almost all men, land two grades above clinical support specialists, almost all women. The specialists take calls from radiographers when a scanner fails mid-scan. They talk them through recovery steps, decide whether a patient session can continue and manage the hospital's expectations until an engineer arrives. The head of people reads both job descriptions again. She cannot see two grades of difference in the work.

She goes back to the scores and finds nothing wrong with them. What she finds is in the plan. Responsibility for equipment, budget authority and value of assets handled are three separate factors. The engineers score high on all three. Those factors have eight levels each. Communication has four. Handling distressed callers sits inside a "working conditions" factor worth a small fraction of the total. Nobody on the panel ever discussed any of this. The weights came with the plan.

This is a composite of a common situation at companies this size, not a client. The details are illustrative.

Why job evaluation bias is hard to see

Two terms first. A factor plan is the set of factors, such as knowledge, responsibility, effort and working conditions, that jobs are scored on, each with its own levels. A weight is how much a factor can add to a job's total compared with the other factors.

Weights are meant to reflect values. The EU's 2026 toolkit on gender-neutral job evaluation says plainly that factors are weighted according to their importance to the organization [1]. That is legitimate. A company that relies on deep technical knowledge can weight knowledge heavily. The problem is that values carry history. What a company has always paid more for tends to feel more important. A plan built to reflect "what matters here" can quietly encode which jobs have always been paid more and who has done them.

The same toolkit warns that gender bias can enter through how factors are defined, how points are assigned and how weights are applied. It notes that this can overvalue skills more often found in jobs typically held by men and undervalue those common in jobs done mostly by women [1]. The UK equality regulator's guidance lists five design risks in a points plan. Only one of them is the weights themselves. Two more, the number of levels and the points per level, act like weights in practice. A third, the choice of factors, can count one demand twice [2].

There is also an attention problem. Scoring gets all the scrutiny because it is where people argue. Every job has an owner who cares about the result, so every score is challenged. Weights are set once, early, by a few people, often before anyone knows which jobs they will lift. They look like technical settings rather than choices. The scoring meetings feel rigorous because they are. But rigor applied to the scores cannot fix a tilt built into the plan.

A related problem is adjusting weights after scoring until the results match today's pay order. Our guide on which job evaluation method to use covers that one. This guide is about the bias that is already there before any score is given. Many employers know this is the hard part. In a September 2026 poll of more than 200 HR professionals in Ireland, 48% named job evaluation and grading as their biggest challenge in preparing for the EU pay transparency rules [8].

Where weight hides in a factor plan

This table shows the places in a points plan that decide how much a factor counts. It shows where to look, not how to set a plan.

WhereWhat it decidesWhy it gets missed
The stated weightThe share of the total each factor is meant to carryIt is the only number people check, so it feels like the whole answer
Overlapping factorsWhether one demand is counted once or several timesEach factor sounds distinct on its own
Number of levelsHow far a factor can separate jobsMore levels looks like more precision, not more weight
Points per levelHow much each step up is worthIt sits in a scoring table few people read
How jobs spread on a factorWhether the factor ever changes the orderIt only shows once real jobs are scored

Five weighting traps in job evaluation

"The weights came with the plan"

What it looks like: The company adopts a plan from a former employer, a consultant or a published toolkit and keeps its weights as they are. Nobody can say why knowledge carries the weight it does, only that it always has.

Why smart teams fall into it: A plan with a track record feels safer than one written in-house. Changing the weights feels like tampering with a method that has been proven elsewhere.

What it costs: Weights are tuned to the workforce they were built for. A plan made for a factory weights the demands of factory work. Applied to a company of support specialists, engineers and coordinators, it can tilt the order without anyone choosing a tilt. Even the EU toolkit says its own default weights are illustrative only and are not a model it recommends or endorses [1]. The UK regulator's guidance says that in a legal challenge, an employer will need to justify its weights by showing how important the factors are to the organization as a whole [2]. "It came with the plan" is not a justification. The weights have to be yours, with your reasons.

"Budget, headcount and impact are three different things"

What it looks like: The plan has separate factors for budget held, number of people managed, value of equipment handled and scope of impact. A team lead with a budget and a large installed base scores high on all of them for what is, at root, one demand: Scale of control.

Why smart teams fall into it: Each factor has its own definition and sounds like a different thing. A plan with many responsibility factors also looks thorough. Splitting a demand into parts feels like measuring it more carefully.

What it costs: When factors overlap, a demand is counted more than once and the jobs that have it gain points several times over. The UK regulator gives this exact example: In one biased plan, a factor for strength partly double-counts one for sustained physical effort [2]. The EU toolkit names double counting in ways that favor jobs dominated by men as one of the ways bias comes into play [1]. It also warns that adding a factor for something already measured elsewhere overstates the value of certain jobs. The stated weight of each factor can look modest while the combined weight on one kind of work is large. The cost is invisible until someone adds the overlapping factors together.

"Each factor has the weight we gave it"

What it looks like: The plan document shows a percentage for every factor and the leadership team signs it off. In the scoring table, though, financial responsibility has eight levels and caring responsibility has four. Or both have five levels, but each step in one is worth more points than each step in the other.

Why smart teams fall into it: The percentages are easy to read and easy to approve. Level counts and point tables look like detail for the people running the project. More levels in a factor feels like finer measurement, not extra weight.

What it costs: The number of levels in a factor is itself a form of weighting [2]. A factor with eight levels can separate jobs far more than a factor with four, whatever the stated percentages say. The same is true of points per level. The UK guidance compares a financial responsibility factor at five points a level with a caring responsibility factor at three. It says that where factors typical of mostly male jobs carry higher scores, the employer has to be able to justify the difference objectively [2]. There is a quieter version too. If almost every job sits at the same one or two levels on a factor, that factor barely moves the order, however high its stated weight. The weight a plan really gives a factor only shows once real jobs are scored.

"Emotional demands are in the plan"

What it looks like: The company has read the guidance and added what was missing: Communication, emotional effort and dealing with distressed people now appear in the plan. But they sit in one small factor, or inside a working conditions factor with few levels. That factor is lightly weighted and was added late.

Why smart teams fall into it: Including the factor feels like solving the problem. The company can show it covers the demands the law worries about. The weight looks like a second-order question.

What it costs: A factor that is present but can never move a job across a grade boundary changes nothing that matters. The UK guidance says a plan should not give very high or very low weights to factors found only in jobs done mainly by one sex [2]. The EU toolkit says its default weights were designed so that elements often found in jobs mostly held by women, such as soft skills and specific communication skills, are properly recognized and not overlooked [1]. The EU directive asks that relevant soft skills are not undervalued [3]. Our guides on whether the EU rules force job evaluation and proving equal value cover what these factors should measure. This trap is about what happens after they are in the plan: They are measured, then barely counted.

"We checked every score, so the plan is fair"

What it looks like: The panel challenges every score and records every reason. The company is proud of the process, rightly. But nobody looks at what the plan does as a whole. Nobody asks how the mostly female jobs and the mostly male jobs moved, or what would change if a weight were set differently.

Why smart teams fall into it: Process is visible and easy to defend. A clean record of scoring meetings feels like proof of fairness. Testing the plan's results by gender can feel like working backward from an answer.

What it costs: The courts look at the system, not only the meetings. In the Rummler case, the EU Court of Justice held that a job classification system may use a criterion such as muscular effort where the work requires it, but only so long as the system as a whole precludes discrimination on grounds of sex. It must, where the work permits, take into account criteria for which workers of each sex may show particular aptitude [4]. The EU toolkit asks employers who change its weights to check for unintended bias by testing the weights on women-dominated and men-dominated jobs [1]. In the UK, a tribunal can set aside a job evaluation study it has reasonable grounds to suspect discriminates because of sex or is otherwise unreliable [6]. A clean scoring record does not answer that question. A test of the weights does.

What a factor plan you can defend looks like

A sound plan is not one with no weights or equal weights. Unequal weights are normal [2]. It is one where every weight was chosen on purpose and its effect is known. A defensible plan at 50 to 500 people has these qualities.

Every weight has a reason in your own words. The reason refers to the work your company does and what it needs from its jobs. It is not "the plan came this way" or "that factor has always mattered most". Someone outside the project could read the reasons and understand them.

Each demand is counted once. Factors are distinct. Where two factors could both capture the same thing, such as scale of control, the plan says which one does. A reader can see where each kind of work earns its points.

The weight on paper is the weight in practice. Someone has looked at how many levels each factor has, what each step is worth and how real jobs spread across the levels. Then they have confirmed that the share each factor really carries is close to the share the company intended. Where it is not, the difference is a decision, not an accident.

Demands typical of mostly female jobs have room to count. Factors that capture communication, emotional effort, care for people and coordination have enough levels and weight to move a job across a grade when the demand is high. The same is true in the other direction for factors typical of mostly male jobs.

The plan has been tested as a whole. Before grades are final, the company has looked at where its mostly female and mostly male jobs landed. It has also looked at whether the order holds when weights move within a reasonable range. A result that changes sharply with a small change in one weight is a signal to look harder at that weight.

It is written down and can be shared. Weights, reasons, level counts and the results of the test are recorded. Under the EU rules, the criteria used to set pay must be objective, gender-neutral and accessible to workers [3]. Where worker representatives exist, they have seen the weights and the reasons. The EU toolkit suggests unions ask for justification of weights that might indirectly favor traditionally male jobs [1].

None of this needs a heavier method. A short points plan can meet every one of these tests. A long one can fail all of them.

The calls only you can make

What your company values most. Weights are a statement about what the company pays for. Only the leadership team can say whether deep technical knowledge, responsibility for people, judgment under pressure or something else matters most. We can test the effect of a choice. We cannot make it for you.

How far to depart from a reference plan. A published or borrowed plan is a fair starting point. Deciding which of its weights fit your work and which belong to someone else's is a judgment about your business.

What to do when the test moves jobs. If testing shows that mostly female jobs fall when the weights are set one way and rise when they are set another, someone has to decide which setting the company can justify. That is a values call with legal consequences. It should not be left to the project team.

Who sets the weights. A plan weighted by the leadership team alone reflects the jobs the leadership team knows. Whether to involve a wider group and how far to involve worker representatives where they exist is the company's decision.

How much to explain. The EU rules require pay criteria to be accessible to workers [3]. How much of the weighting logic to share beyond that is a choice about trust.

The law and the guidance

  • European Union, pay structures. The pay transparency directive requires pay structures that let employers assess whether workers are in a comparable situation regarding the value of work. The assessment must use objective, gender-neutral criteria that include skills, effort, responsibility and working conditions. Relevant soft skills must not be undervalued (Article 4(4)). The criteria used to set pay, pay levels and pay progression must be objective, gender-neutral and easily accessible to workers (Article 6(1)) [3].
  • European Union, the system as a whole. In Rummler v Dato-Druck (1986), the Court of Justice held that a job classification system may use muscular effort as a criterion where the work requires it, so long as the system as a whole precludes sex discrimination by taking other criteria into account [4].
  • European Union, the 2026 toolkit. The EU-wide guidelines on gender-neutral job evaluation, published on 26 March 2026, warn that bias can enter through how weights are applied and through double counting. They say any change to the default weights needs a strong, objective, gender-neutral reason and should be tested on women-dominated and men-dominated jobs [1].
  • United Kingdom, rated as equivalent. Work is rated as equivalent if a job evaluation study gives the jobs equal value, or would do so if the evaluation were not made on a sex-specific system. A system is sex-specific if it sets different values for men and women for any demand (Equality Act 2010, section 65(4) and (5)) [5]. Where a study has given two jobs different values, a tribunal must find they are not of equal value unless it has reasonable grounds to suspect the study discriminates because of sex or is otherwise unreliable (section 131(5) and (6)) [6].
  • United Kingdom, the regulator's guidance. The Equality and Human Rights Commission's guide on gender-neutral job evaluation schemes names factor choice, definitions, levels, scoring and weighting as design risks. It says the number of levels in a factor is a form of weighting [2].
  • United States. Under the EEOC's compliance guidance, a job classification system offered to explain a pay difference has to reflect the duties of the jobs accurately and be applied the same way to men and women [7].
  • Readiness. In a September 2026 poll of more than 200 HR professionals in Ireland, 48% named job evaluation and grading as their biggest challenge in preparing for the directive [8]. The poll was run by Mason Hayes & Curran, a law firm.

This is general information, not legal advice. Rules and their national versions differ. Check the rules where you employ people.

Our view

We think most of the bias in a points plan comes from weighting, not scoring. Scoring gets the attention because people argue about it. Weighting gets set early and then taken as given. A company can run a careful, well-documented scoring process and still end up with an order that undervalues its mostly female jobs, because the plan was tilted before the first meeting.

The fix is not to stop weighting or to weight everything equally. It is to treat every weight as a choice: Give it a reason, check that it means in practice what it says on paper and test what it does to the jobs before the grades are set. A plan that passes that test can be explained to a panel, an employee or a tribunal. A plan that only passed its scoring meetings cannot.

Sources

  1. European Institute for Gender Equality, EU-wide guidelines on gender-neutral job evaluation and classification: Step-by-step toolkit (26 March 2026), pages 13, 17, 19, 36, 59, 60 and 78, checked 2026-10-29
  2. Equality and Human Rights Commission, Gender-neutral job evaluation schemes: An introduction guide (March 2014), risks of gender bias in design, checked 2026-10-29
  3. Directive (EU) 2023/970 on pay transparency, Articles 4(4) and 6(1) (EUR-Lex), checked 2026-10-28
  4. Court of Justice, Rummler v Dato-Druck GmbH, Case 237/85 (1 July 1986), operative part (EUR-Lex), checked 2026-10-29
  5. Equality Act 2010, section 65: Equal work (legislation.gov.uk), checked 2026-10-29
  6. Equality Act 2010, section 131: Assessment of whether work is of equal value (legislation.gov.uk), checked 2026-10-29
  7. US EEOC, Compliance Manual Section 10: Compensation Discrimination, 10-IV.F.2.d (job classification systems), checked 2026-10-26
  8. Mason Hayes & Curran, Only one in five Irish employers have made real progress on EU pay transparency rules (8 September 2026; poll of more than 200 HR professionals), checked 2026-10-26

Market pricing or job evaluation?

Market data sets the level of pay. Job evaluation sets the order. Five traps in picking one over the other and what using both well looks like.

Buy a leveling framework or build one?

A bought leveling framework brings shared words, not decisions. Five traps in buying or building one and what a sound company level guide looks like.

Pay equity regression or cohort review?

Below about 30 people a group, a pay equity regression misleads. Five method traps for small teams and when a cohort review is the sounder test.

See Resources for every guide and the self-check.