TL;DR: A defensible redundancy selection matrix needs three things for every criterion: a measurable definition, a fixed maximum score, and a named document the score comes from. This guide gives you a worked matrix and the specific scoring mistakes that UK tribunals have already ruled on.
Where redundancy claims actually come from
The ONS puts the UK redundancy rate at 3.9 per 1,000 employees for May to July 2026, published on 15 September. That is redundancy running as ordinary business activity, which is exactly the condition under which process discipline slips.
The business case is rarely the part that gets attacked. An employee can usually accept that a team of twelve is becoming a team of nine. What they cannot see, and will ask about first, is why the three were them. At that point the scoring matrix stops being an internal working document and becomes the employer's evidence.
So this article does two things. It gives you a matrix you can build, with a maximum score and a named source document for every criterion, then walks through the scoring decisions that have already lost at tribunal, including one almost every published matrix guide still lists as safe. It is written for the HR manager who has to produce the document, so everything below should end up in front of a line manager on Monday.
What counts as a fair selection criterion
Start further back than the matrix. Section 139 of the Employment Rights Act 1996 requires the dismissal to be wholly or mainly attributable to a business or workplace closure, or a reduced requirement for employees to carry out work of a particular kind. If that is not the honest situation, no scoring rigour helps, because there is nothing to select for.
Assuming a genuine redundancy situation, Acas sets out the criteria employers commonly use: standard of work or performance, skills, qualifications and experience, attendance record, and disciplinary record. On attendance, Acas is direct: the record must be accurate, and it must exclude absences relating to disability, pregnancy and maternity.
The procedural expectations come from Williams v Compair Maxam Ltd [1982] ICR 156, and tribunals still work from those principles more than forty years on. Warn people as early as circumstances allow. Consult representatives on the criteria themselves, before they are applied. Choose criteria checkable against something outside one manager's head. Apply them consistently. Look properly for alternative employment. That summary comes from case commentary rather than the primary report.
Guidance often blurs this: no statutory clause says selection criteria must be objective. Objectivity is the mechanism by which an employer discharges the burden of showing the process was fair once someone challenges it. A subjective criterion is lawful and indefensible at the same time.
A second and separate test runs alongside fairness. Section 19 of the Equality Act 2010 covers indirect discrimination: a criterion applied identically to everyone can still be unlawful if it puts people sharing a protected characteristic at a particular disadvantage, unless the employer shows it is a proportionate means of achieving a legitimate aim. Equal treatment is a defence against one claim and no defence at all against the other.
The pool comes before the matrix
A flawless matrix applied to the wrong selection pool still fails. Tribunals examine how the pool was drawn with at least as much interest as how scores were applied inside it, because the boundary is where an employer can quietly decide the outcome before scoring begins.
Capita Hartshead Ltd v Byard is the case most often cited on this. The employer treated the claimant as a pool of one despite there being four actuaries doing materially similar work, reasoning that reassigning the claimant's clients risked losing them. The tribunal found the dismissal unfair, treating that risk as speculative. This is drawn from case commentary, so take it as a description of the outcome rather than of the judgment.
Drawing the redundancy selection pool means asking how interchangeable the work genuinely is. How much retraining would a move require? Have people covered each other's roles during leave or a busy period? Job titles are the weakest signal here and the one most likely to have drifted for reasons unrelated to what anyone does.
Consult on the boundary and write down why it sits where it does. That note is the answer to the first question you will be asked, possibly two years later. Consultation thresholds, notice and appeals are all covered in the stages either side of selection.
A worked scoring matrix: criteria, weighting and evidence
This is the part most guidance skips. Below is a structure to adapt, not a template that guarantees anything, and the criteria have to fit the roles that are staying.
| Criterion | What it measures | Max | Document the score is read from |
|---|---|---|---|
| Performance against role objectives | Delivery against objectives already recorded | 0–15 | The last two completed appraisals |
| Skills and experience for remaining roles | Capabilities the surviving structure needs | 0–10 | Skills matrix or role profile signed off before scoring |
| Breadth of role coverage | Demonstrated work across more than one area | 0–10 | Recorded project history and cover arrangements |
| Attendance | Unauthorised, non-protected absence over 12 months | 0–5 | HR absence records, protected absence removed first |
| Formal disciplinary record | Live warnings on file in the last 12 months | 0–5 | Disciplinary file |
| Continuous service | Length of continuous employment | 0–3 | Employment record |
Weighting is expressed as the maximum points per criterion, and Acas confirms that criteria may carry different weight provided the weighting is decided in advance and applied identically to everyone in the pool. The heaviest maximums sit on the criteria that predict who can do the work that remains. Service gets three points because it is a tie-breaker, for the reason in the next section.
Each criterion also needs written descriptors for what each score means. Attendance scoring 5 should say what 5 is: no unauthorised absence in twelve months once protected absence has been removed. Without descriptors, two managers scoring the same person land in different places.
Score each employee independently, by two people, then hold a moderation meeting. Record where the scorers disagreed and how it was resolved. A score that was argued over and settled reads better than a single clean sheet with no history.
Now the exclusions, because what you leave out is a decision too. Potential. Attitude. Cultural fit. Flexibility, when what it measures is who can work late and therefore who has no caring responsibilities. And anything a manager can only assess from memory.
Five scoring mistakes that turn selection into a claim
1. Scoring protected absence as ordinary attendance. Absence relating to disability, pregnancy and maternity has to come out before the score is calculated. Acas says this directly. Footnoting it afterwards leaves the contaminated number sitting in the total.
2. Treating last in, first out (LIFO) as more than a tie-breaker. In Rolls-Royce plc v Unite the Union [2009] EWCA Civ 387, a scheme awarding one point per year of continuous service was held proportionate because service sat alongside five other substantive criteria and could not determine the outcome by itself. A LIFO-only scheme is a much harder argument after the Equality Act, because younger employees have had less opportunity to accrue service. Again, commentary rather than the primary report.
3. Awarding points for qualifications the role does not require. This is the one that should change how you read your own matrix. In Norman v Lidl Great Britain Ltd (ET 1804509/2023), a knowledge criterion gave additional points for holding a degree. The claimant was 63 and scored against colleagues in their 30s, and produced statistics showing that employees in their sixties are less likely to hold a degree. The tribunal accepted indirect age discrimination, and compensation exceeded £50,000. Almost every matrix guide still lists qualifications among the safely objective criteria. The case is reported through commentary from Bates Wells, Irwin Mitchell and HR Magazine rather than a primary transcript, and the lesson survives that caveat: score a qualification only where the remaining role needs it.
4. Scores with nothing behind them. A manager's recollection of how someone performed last year is not evidence. Where the appraisal was never completed, the honest fix is to drop the criterion for the whole pool rather than reconstruct a score from memory.
5. Adjusting the weighting after the scores are in. Deciding that skills should count double once you have seen who came bottom is the clearest evidence available that the outcome was chosen first. Fix the weighting, circulate it, then score.
One further boundary. If someone is selected because of a protected reason rather than despite one, this leaves ordinary unfair dismissal territory altogether and becomes automatically unfair dismissal, where the matrix is beside the point.
What 'evidenced' actually looks like on paper
Acas says scores must be evidenced, without showing what evidence looks like sitting in a folder. So here is the pack.
For each scored employee: the criteria and weightings exactly as circulated before scoring began, the scoring descriptors, both independent score sheets, the moderation note, and a copy or clear reference for every source document behind every score.
Absence scoring earns its own audit line, separate from the matrix. Record the raw absence total, the number of days removed, the reason category for the removal, and who made that call. Do not write medical detail into the matrix. A category label is enough.
Where appraisals are missing, or inconsistent because two managers kept up with them and one did not, write that down and say what you did about it. An acknowledged gap handled the same way for everyone reads far better than a score that simply appeared.
At the individual consultation meeting, give the employee their own scores and the descriptors, take their challenge, then record whether any score changed and why. A score tested on the record and held is stronger evidence than one nobody questioned.
Keep the pack for at least the tribunal limitation window plus any appeal period, and assume you may have to explain one cell years later.
Is there a 70/30 rule for redundancy?
No. There is no 70/30 rule in UK redundancy law, and nothing in statute, Acas guidance or professional body material establishes one.
People usually mean one of three unrelated things. The most likely is role mapping in a restructure, where a new role can be treated as a direct match for an old one and filled by assimilation without competitive interview, on the basis that roughly 70% or more of the old role carries over at the same grade. That is an organisational convention with no legal threshold behind it.
The second use is informal shorthand for splitting scoring weight 70% objective and 30% subjective. It traces to no named authority, and budgeting for subjectivity is a strange thing to do deliberately. Make every criterion measurable instead.
The third belongs to settlement negotiation, where people talk about offering around 70% of likely tribunal value. Nothing to do with selection.
Underneath the question sits a real principle: what matters is whether each criterion has a measurable definition and a document behind it.
What changed in 2026, and what is coming
One change is already in force. From 6 April 2026, the maximum protective award for failing to consult collectively doubled from 90 to 180 days' pay per affected employee, and a 25% uplift for unreasonable failure takes maximum exposure to 225 days' pay per employee.
The thresholds a scoring exercise sits inside are unchanged. GOV.UK sets out that 20 or more redundancies at one establishment within 90 days triggers collective consultation, with a minimum of 30 days before the first dismissal takes effect for 20 to 99 redundancies, and 45 days for 100 or more.
Three further changes are proposed for 2027, and should be read as proposals rather than settled law, since no firm commencement date has been located: counting proposed redundancies across the whole organisation rather than establishment by establishment, cutting the unfair dismissal qualifying period from two years to six months, and removing the compensation cap.
That middle one lands directly on selection. If the two-year qualifying period drops to six months, a far larger share of any pool can bring an ordinary unfair dismissal claim. If you are running a contested exercise, have the criteria and the pool reviewed by an employment lawyer before scoring opens rather than after someone appeals.
Build the matrix before you need it
The short version. Draw the pool first and write down why it sits where it does. Give every criterion a maximum score and a written descriptor. Name the document each score is read from. Fix the weighting before anyone is scored. A matrix built that way is one you can defend, which is a different claim from one that makes the process immune.
The real constraint sits upstream of matrix design, in whether the appraisal and absence records exist and are consistent across everyone in the pool. That cannot be repaired in the four weeks before a consultation opens.
So here is one thing to do this week. Pick any ten people in a team, no redundancy in prospect, and check whether you could score them today from documents alone. Whatever you cannot evidence is a criterion you do not yet have. The step-by-step UK redundancy process guide linked above covers consultation, notice and appeals.