Britain has legislated for guaranteed hours, advance notice of shifts and compensation for last-minute cancellations. But almost every number that will determine what workers gain remains undecided. We modelled the choices shift by shift.
A hotel worker is told on Tuesday that Saturday’s eight-hour shift is no longer needed.
A care worker accepts a weekend visit schedule, arranges childcare and then loses six hours when a client cancels.
An agency warehouse worker keeps Thursday free, turns down another assignment and receives a message that morning saying the shift has been moved.
Under the present system, each may lose the expected wages without receiving compensation.
Britain’s new employment legislation is intended to move some of that risk away from workers. The Employment Rights Act 2025 creates three connected rights: an offer of guaranteed hours reflecting work performed over a reference period, reasonable notice of shifts and changes, and payment when shifts are cancelled, shortened or moved at short notice. The measures apply in England, Scotland and Wales, but have not yet taken effect because regulations must still settle the crucial details.
Those details are not administrative footnotes. They will decide:
- whether a worker is covered;
- how many hours an employer must offer;
- how regular the work must have been;
- how much notice is sufficient;
- what percentage of a cancelled shift must be paid;
- which cancellations are exempt;
- and whether accepting guaranteed hours could remove other protections.
The government’s consultation closes at 11.59pm on 25 August 2026. Until the final regulations are made, nobody can responsibly tell an individual worker exactly what they will receive. It is possible, however, to calculate the range of outcomes created by the government’s choices.
Our modelling shows why the final design matters. For workers with steady weekly patterns, guaranteed-hours offers could closely match their normal schedules. For seasonal and highly erratic workers, changing just one calculation—from the mean to the median—can sharply reduce or increase the hours offered. Cancellation compensation could replace only a small fraction of lost wages or most of them, depending on the percentage selected.
The reform’s real value will therefore not be determined by the phrase “guaranteed hours”. It will be determined by the arithmetic underneath it.
The scale of insecure work
Approximately 1.2 million people reported being employed on zero-hours contracts in the final quarter of 2025, equivalent to around 3.6% of UK employment. Hospitality, transport, arts and other services, health and social care, and wholesale and retail are among the sectors most reliant on these arrangements. Zero-hours workers are more likely than other workers to be young, female or in full-time education.
The government’s figures also indicate that 26% of people on zero-hours contracts would like to work more hours, compared with 8% of workers not on such contracts. These figures describe only contracts formally recognised and reported as zero-hours arrangements; low-hours contracts with substantial variable work may expose employees to similar uncertainty.
Shift notice can be exceptionally short. Research cited by the government found that 59% of variable-hours workers received less than one week’s notice of shifts and 13% received less than 24 hours. A separate survey cited in the consultation reported that 64% of agency workers received less than 24 hours’ notice of cancellations or changes.
These arrangements can preserve flexibility for students, carers and others who prefer irregular work. The government is therefore not abolishing zero-hours contracts. Workers will still be able to remain on them rather than accept an offer of guaranteed hours. The policy is intended to restrain one-sided flexibility: the ability of an employer to vary work while transferring most of the resulting financial and logistical risk to the worker.
The first question is not how many hours someone works
It is how many hours they are already guaranteed.
The consultation proposes an “hours threshold”. Workers with no guaranteed hours would be potentially eligible. Workers with some contractual hours would also be covered where those hours are at or below a threshold set in regulations.
The government says its preferred range is between eight and 20 guaranteed hours a week, although the consultation invites views on thresholds from eight to 48 hours and other alternatives.
This produces a counterintuitive result: raising the threshold would bring more, not fewer, low-hours workers into scope.
Consider two employees who both normally work 24 hours:
| Worker | Contracted hours | Actual average | Covered if threshold is 8 hours? | Covered if threshold is 16 hours? |
|---|---|---|---|---|
| A | 0 | 24 | Yes | Yes |
| B | 12 | 24 | No | Yes |
Worker B experiences a 12-hour gap between contracted and usual work. Yet under an eight-hour threshold, that worker could be excluded because the contract already guarantees more than eight hours.
A low threshold concentrates protection on zero-hours and very-low-hours contracts. A higher threshold captures more workers whose contracts understate their ordinary schedules.
It may also influence employer behaviour. A business could respond to a 16-hour threshold by issuing contracts guaranteeing 17 hours, potentially taking workers outside some rights even where schedules continue to fluctuate substantially.
The consultation explicitly recognises this interaction. It notes that if the threshold for notice and cancellation protection were 16 hours, a worker who accepted an 18-hour guaranteed-hours offer might cease to qualify for those protections.
That creates one of the reform’s most important unresolved questions:
Should accepting a more secure contract cause a worker to lose protection against unpredictable scheduling?
A guaranteed number of hours and a predictable rota are not the same thing. A contract may guarantee 18 hours while allowing those hours to be placed across almost any day or time.
Twelve weeks could determine years of income
The government prefers an initial 12-week reference period, although it is also consulting on 26- and 52-week alternatives. The stated objective is to make workers eligible reasonably quickly while retaining enough history to establish their normal working pattern.
Twelve weeks is easy to describe. It is much harder to apply fairly.
A reference period running from October to December may capture Christmas overtime. One running from January to March may capture a post-Christmas fall in retail demand. A summer reference period could inflate hours for hospitality and events workers. A school catering worker’s reference period could contain a holiday closure.
A longer period smooths seasonal variation but forces workers to wait longer for an offer. A shorter one delivers rights sooner but makes the result more sensitive to temporary peaks, sickness, unpaid leave and unusual trading conditions.
There will also be subsequent reference periods. The government is consulting on whether these should last 12, 26 or 52 weeks and whether one period should begin immediately after the previous one. Longer or separated periods would reduce employers’ administrative workload, but workers whose schedules rise could wait longer for their guarantee to be revised.
Working frequently may not be enough
Completing a reference period will not automatically trigger an offer. The worker’s hours must also satisfy a regularity test.
The consultation presents two broad approaches.
Under Option A, a worker would have to work during a minimum number of weeks. The weeks would not need to be consecutive. For example, regulations might require work in eight of the preceding 12 weeks.
Under Option B, the worker would have to pass both a weekly-distribution test and a minimum-hours test. The consultation gives the example of a worker contracted for two hours a week who might need to complete at least 96 additional hours, as well as working during at least eight weeks of the period.
Option B would exclude more workers.
A person who works six hours every week could demonstrate an entirely regular relationship but fail a demanding total-hours requirement. Conversely, an employee completing several long weeks during a concentrated peak might pass the total-hours test while having little continuing expectation of work.
The better test depends on what the legislation is trying to identify:
- a continuing working relationship;
- substantial dependence on the employer;
- a regular need for labour;
- or merely enough work to make an offer administratively worthwhile.
Those are not interchangeable concepts.
Mean or median: the calculation that could change everything
Once a worker qualifies, the employer must calculate an offer reflecting the hours worked during the reference period.
The government is consulting on two methods:
Mean average
Add all hours worked and divide by the number of weeks.
Median
Arrange the weekly totals in order and use the middle figure.
The consultation illustrates the difference with a worker completing eight hours for seven weeks and 20 hours for five weeks. The mean produces 13 guaranteed hours. The median produces eight.
Neither method is universally more generous.
A mean is more sensitive to high and low outliers. A median reflects the typical week but can disregard a substantial amount of real work. For a highly seasonal worker, the median may actually be higher than the mean because zero-hour weeks pull down the average while leaving the middle worked week relatively high.
Five illustrative rotas
We constructed five synthetic 12-week rotas representing patterns commonly associated with hospitality, care, retail, agency delivery and events work. They are not presented as survey findings or legal predictions. Their purpose is to demonstrate how the proposed mechanisms behave.
| Illustrative worker | Weekly hours observed | Mean offer | Median offer | Difference |
|---|---|---|---|---|
| Hospitality server | 12–26 hours | 20.0 | 20.5 | +0.5 under median |
| Care worker | 24–40 hours | 32.2 | 32.5 | +0.3 under median |
| Retail assistant | 6–20 hours | 13.1 | 13.0 | −0.1 under median |
| Agency delivery worker | 14–48 hours | 31.0 | 31.0 | No difference |
| Events worker | Four zero-hour weeks; worked weeks of 28–44 hours | 21.2 | 29.0 | +7.8 under median |
For the first four workers, the choice has little effect. For the events worker, it is transformative.
The events rota was:
0, 0, 30, 36, 0, 42, 44, 0, 28, 40, 0, 34 hours.
The mean is 21.2 hours. The median is 29 hours.
This happens because the median falls between the sixth and seventh values after the weeks are ordered. A minority of zero-work weeks lowers the mean substantially but does not necessarily determine the midpoint.
Yet that worker might fail the regularity test before either figure is calculated. If the regulations require work in eight of 12 weeks, the worker qualifies exactly. If they require nine, the worker does not qualify at all.
That exposes a key lesson: the regularity test may matter more than the averaging formula.
A guaranteed-hours offer is not necessarily guaranteed income
The consultation is also asking whether hours must be expressed weekly, monthly or over another period.
A guarantee of 20 hours each week offers much more immediate predictability than 80 hours across a four-week period. Under the latter arrangement, an employer might schedule 35 hours in week one, 35 in week two, ten in week three and none in week four.
The total is the same. The household consequences are not.
The government acknowledges that allowing employers to select the allocation period would increase operational flexibility but could reduce security and predictability for workers.
The regulations should therefore distinguish between:
- volume predictability — how many hours will be provided; and
- timing predictability — when those hours will be worked.
A reform that guarantees only volume may improve average earnings without solving childcare, travel, second-job and household-planning problems.
What might a cancelled shift be worth?
The Act permits ministers to define “short notice”, but the period cannot exceed seven days. The consultation offers possible thresholds of one, two, three, five or seven days. It also considers a second “very short notice” band carrying higher compensation.
For ordinary short notice, the possible compensation rates listed are:
- 10%;
- 30%;
- 50%;
- 65%;
- 80%;
- or another percentage.
For very short notice, the listed options begin at 30% and rise to 80%. Payment cannot exceed the amount the worker would have earned from the affected hours.
The government is also deciding what wage should form the base:
- the worker’s expected contractual rate; or
- the applicable National Living Wage or National Minimum Wage rate.
For workers earning above the statutory minimum, that choice affects compensation even when the percentage is identical.
Our five modelled workers
For illustration, we assumed each recorded cancellation was inside whatever notice threshold ultimately applies. We then applied three of the possible compensation percentages to the worker’s ordinary hourly rate.
| Worker | Cancelled hours | Expected wages lost | 30% payment | 50% payment | 80% payment |
|---|---|---|---|---|---|
| Hospitality server | 15 | £192.00 | £57.60 | £96.00 | £153.60 |
| Care worker | 18 | £241.20 | £72.36 | £120.60 | £192.96 |
| Retail assistant | 9 | £109.89 | £32.97 | £54.95 | £87.91 |
| Agency delivery worker | 24 | £348.00 | £104.40 | £174.00 | £278.40 |
| Events worker | 30 | £412.50 | £123.75 | £206.25 | £330.00 |
The percentages do not describe minor variations.
For the agency delivery worker, the difference between a 30% and 80% formula is £174 across three cancelled shifts. For the events worker, it is more than £206.
At 30%, workers still absorb 70% of the direct wage loss. The payment may not cover childcare, travel tickets, preparation time or alternative work declined.
At 80%, most expected wages are preserved, but the employer bears far more of the cost of demand uncertainty.
That is the central economic choice. Compensation is not only redress after cancellation. It changes the incentive to cancel.
A token payment may be treated as another scheduling cost. A payment approaching the full shift value encourages employers to improve forecasting, retain backup work or cancel earlier.
A seven-day rule and a one-day rule protect different workers
Suppose a five-hour shift worth £70 is cancelled:
| Notice given | Covered by a one-day rule? | Covered by a three-day rule? | Covered by a seven-day rule? |
|---|---|---|---|
| 10 hours | Yes | Yes | Yes |
| 36 hours | No | Yes | Yes |
| Four days | No | No | Yes |
| Eight days | No | No | No |
A one-day threshold targets the most disruptive cancellations but leaves employers free to withdraw work two or three days ahead without payment.
For some workers, 48 hours is enough to find replacement work. For others, particularly those arranging care or coordinating several jobs, it is not.
A tiered system could recognise this difference:
- cancellation within seven days: 30% compensation;
- cancellation within 72 hours: 50%;
- cancellation within 24 hours: 80% or 100%.
That is more complicated than one rate, but it better matches payment to the inconvenience and preventability of the cancellation.
The consultation’s proposed structure allows a short-notice and very-short-notice distinction, although it has not fixed the periods or percentages.
Moving a shift can be as costly as cancelling it
The protection is not limited to cancellations. It also covers shifts moved or curtailed at short notice.
That distinction is important.
Moving a Saturday afternoon shift to Saturday evening may preserve the worker’s gross hours while making the work impossible because of childcare or transport. Moving Thursday’s assignment to Friday may conflict with a second job. Cutting an eight-hour shift to three hours may leave the worker paying the same travel cost for less than half the expected income.
Under the proposed approach, compensation for moved or curtailed shifts would relate to the affected hours rather than automatically to the entire original shift.
This creates difficult implementation questions:
- Is a shift “moved” if its start time changes by 15 minutes?
- What happens when the worker accepts the change under pressure?
- Does payment remain due when the replacement shift is longer?
- How are split shifts treated?
- Is an unpaid extension of a break a curtailment?
- Who records when the worker was first informed?
The regulations will need an auditable definition of the original shift, the change, the notification time and the worker’s response. Otherwise disputes will turn on screenshots, edited rota apps and conflicting recollections.
Agency workers face a three-party problem
Agency workers are included, but their position is structurally more complicated.
There may be three participants:
- the worker;
- the employment agency;
- the hirer directing the work.
The hirer may generate the schedule and decide to cancel. The agency may issue the notification and process pay. Either could lack information held by the other.
The legislation allows responsibilities to be divided between agencies and hirers. The consultation also considers amending agency regulations so the parties must exchange information needed to administer the rights.
The government’s intended approach is broadly that the party responsible for the short-notice decision should ultimately bear the cost. But workers should not have to identify the responsible commercial party before receiving payment.
The cleanest model would be:
The party that pays the worker must make the payment promptly, while agencies and hirers settle liability between themselves.
Anything else risks forcing the least powerful person in the arrangement to resolve a contractual dispute between two businesses.
Agency eligibility will also be assessed in relation to work for a particular hirer. An agency worker completing 30 hours each week across three businesses may not qualify for an offer from any one hirer if their hours are too dispersed.
Seasonal employers have a potential escape route—but not an unlimited one
Employers may use genuine limited-term contracts for specific tasks, events or temporary needs.
Where a reasonable fixed-term arrangement ends before the reference period finishes, no guaranteed-hours offer may be required. Examples could include fruit picking until a harvest is completed or event work until a conference ends. Repeated contracts for the same or similar work, however, may undermine the argument that the need was genuinely temporary.
This is necessary for authentic short-term work. It is also an obvious point of vulnerability.
A broad temporary-need exception could allow permanent seasonal demand to be repackaged as a sequence of temporary contracts. Hotels, retailers and event venues often experience predictable peaks every year. The demand fluctuates, but the business need is not necessarily exceptional or temporary.
Final regulations should distinguish between:
- a finite one-off task;
- a genuinely unpredictable event;
- a recurring annual peak;
- and an employer’s ordinary fluctuating demand.
Without that distinction, “seasonal” could become a label rather than an evidence-based exemption.
The reform may stabilise hours without stabilising wages
Guaranteed hours would create a minimum, not a maximum. Employers could continue offering additional work.
That produces two different forms of security.
A worker averaging 25 hours might receive a 20-hour guarantee and continue working between 20 and 30. Their income floor improves even though earnings remain variable.
Another worker might receive a 25-hour guarantee but lose access to overtime because the employer spreads additional hours across more employees to avoid future upward revisions. Their contracted security improves while total earnings fall.
The reform’s success should therefore not be judged solely by the number of guaranteed-hours offers made.
The government should monitor:
- contracted hours before and after implementation;
- actual hours worked;
- declined offers;
- additional hours offered;
- hourly pay;
- number and timing of cancellations;
- worker turnover;
- and the use of fixed-term or outsourced labour.
Otherwise an employer could technically comply while redesigning schedules in ways that offset much of the intended benefit.
Five ways employers could minimise liability
The legislation contains anti-avoidance powers, but regulations and enforcement will determine their strength.
Potential responses include:
1. Keeping contracted hours just above the threshold
A business might offer enough guaranteed hours to remove workers from scope while preserving substantial scheduling discretion.
2. Spreading work across more people
Instead of giving one worker 24 regular hours, an employer could allocate 12 hours each to two workers. This may reduce the guarantee generated by any one rota history.
3. Cancelling just outside the payment window
A seven-day payment rule could encourage decisions eight days ahead; a three-day rule could encourage cancellations on day four.
That would still improve notice, but the data may show a suspicious clustering immediately outside the legal boundary.
4. Reclassifying recurring work as temporary
Repeated limited-term arrangements could be used in sectors with regular seasonal demand.
5. Replacing cancellations with conditional shifts
Employers might describe work as provisional, standby or subject to confirmation. Regulations must determine when a communication becomes an actual shift for notice and payment purposes.
These are not arguments against the reform. They are reasons to collect implementation data from the beginning.
What our modelling suggests
Three conclusions emerge.
1. Regular workers are likely to gain the clearest security
Workers whose hours cluster around a stable average will receive similar offers under either the mean or median method. Their principal gain is the conversion of a customary schedule into a contractual floor.
2. Intermittent workers will be determined by technical definitions
For events, seasonal and fragmented agency work, eligibility may turn on how many weeks count as regular, whether separate assignments can be combined and whether zero-work weeks enter the calculation.
The difference between receiving a 29-hour offer and receiving no offer at all may be one additional worked week.
3. Cancellation payments may matter more immediately than guaranteed hours
A worker can decline a guaranteed-hours offer. They can also receive an offer below their hoped-for schedule. By contrast, a meaningful cancellation payment attaches a direct cost to an employer’s decision to withdraw promised work.
For households living week to week, that payment may be the reform’s most visible effect.
The test the final regulations must pass
The government describes the policy as a baseline of security and predictability.
To deliver that, the final framework should satisfy five tests.
Coverage: It should reach low-hours workers experiencing the same insecurity as formal zero-hours workers.
Representation: The calculation should reflect ordinary work without allowing isolated zero weeks or temporary peaks to distort the result.
Timing: Guaranteed hours should be allocated over a period short enough to assist household planning.
Deterrence: Cancellation payments should be high enough to influence employer decisions, not merely acknowledge the loss.
Enforceability: Workers should be able to prove entitlement using ordinary records such as rotas, messages and payslips without needing complex litigation.
The legislation has already established the architecture. The consultation is deciding how many people will actually be sheltered by it.
The question is not whether zero-hours contracts are banned
They are not.
The more useful question is what happens when a business repeatedly relies on a worker but refuses to guarantee the work; when a shift is treated as real until the moment it becomes inconvenient; or when labour demand is flexible only because the worker absorbs every loss.
A well-designed system could turn repeated work into a credible income floor and make last-minute cancellation a decision with a price.
A weak system could produce low guaranteed offers, modest cancellation payments and new contractual strategies designed around the thresholds.
Britain’s zero-hours reforms will therefore be won or lost not in the headline promise, but in four numbers still being decided:
The hours threshold. The reference period. The notice period. The payment percentage.
For workers, those numbers will appear not in legislation, but in next month’s payslip.
Data box: what is decided and what remains open
| Issue | Established in the Act | Still being determined |
|---|---|---|
| Guaranteed-hours offer | Qualifying workers must receive an offer reflecting reference-period work | Threshold, reference period, regularity test, mean or median |
| Worker choice | Worker does not have to accept the offer | Offer and response procedures |
| Shift notice | Qualifying workers have a right to reasonable notice | What is reasonable and possible exceptions |
| Cancellation payment | Payment can be due where a shift is cancelled, moved or curtailed | Notice window, percentage, wage base and exemptions |
| Maximum compensation | Cannot exceed expected earnings from affected hours | Whether payment is 10%, 30%, 50%, 65%, 80% or another amount |
| Agency workers | Included in the framework | Allocation of responsibility between agency and hirer |
| Seasonal work | Genuine limited-term needs can be treated differently | Definition of temporary need and anti-avoidance rules |
| Commencement | Rights created by the Employment Rights Act 2025 | Implementation date and final regulations |
Guaranteed Hours Calculator
Guaranteed Hours Calculator
Enter a rota history and cancelled shifts to compare possible guaranteed-hours offers, compensation scenarios and income volatility under Britain’s proposed reforms.
Build your rota scenario
Use actual weekly records where possible. Do not enter names, employer details or other identifying information.
| Week | Hours worked | Gross pay |
|---|
