At 38 distribution sites in a fictional scheduling trial, the most consequential bargaining issue is not the hourly wage but who decides when those hours are worked. Publishing rotas three weeks ahead reduces last-minute shift changes by 27% in the demonstration dataset, while output per paid hour remains broadly unchanged. The result suggests a narrower proposition than the familiar promise that happier workers necessarily produce more: predictable employment can have value to staff even when it generates no measurable productivity dividend for their employer.

The price of that predictability is visible in the accounts. Employers in the model incur an additional $1.4 million in annual staffing costs to maintain cover during demand swings, equivalent to 2% of the sites’ $70 million baseline payroll. Lower recruitment spending recovers $500,000, leaving a net annual cost of $900,000, or about 1.3% of payroll. Across a modelled workforce of 3,800, that is roughly $237 per employee. The negotiation is therefore about distributing a real cost, rather than dividing an assumed windfall from greater efficiency.

The exercise is illustrative, not evidence from an actual employer or a representative survey of the distribution industry. Its figures describe a constructed, year-long pilot compared with a preceding baseline year. Before the change, schedules were normally published seven days ahead, leaving employees little time to arrange childcare, transport or a second job. Extending that window to 21 days transfers some responsibility for handling fluctuations in orders from individual workers to managers. It does not remove those fluctuations or establish that the same arrangement would work equally well elsewhere.

Read more: A fictional machinery merger turns on service contracts, not factory scale

In the model, last-minute changes fall from 10,000 in the baseline year to 7,300 during the pilot. Against an unchanged annual total of 100,000 scheduled shifts, the proportion altered at short notice declines from 10% to 7.3%, a reduction of 270 basis points. That distinction matters: the 27% headline improvement is a relative fall in disruption, not a claim that more than a quarter of all shifts become newly predictable. Nor does advance publication guarantee that every rota remains fixed. Some changes persist because absences and unexpected orders still require cover.

The output figures offer little support for treating predictability as a straightforward productivity investment. Units handled per paid hour move from 24.0 to 24.1, an increase of about 0.4% year on year that the scenario treats as broadly unchanged. Annual paid hours remain at 7.6 million across the sites. With no demonstrated step-change in efficiency, the extra staffing expenditure cannot credibly be presented as paying for itself through higher output. The pilot instead illustrates a benefit that conventional operational measures can miss: employees gain greater control over their lives without working a shorter week.

Recruitment provides a partial financial counterweight. The demonstration assumes annual replacement hiring falls from 500 workers to 400, a 20% reduction against the baseline year. At an unchanged recruitment cost of $5,000 per replacement, spending drops from $2.5 million to $2 million. Those savings cover about 36% of the additional staffing bill. They are meaningful, but insufficient to erase it. The model also cannot establish that scheduling alone would produce an equivalent retention improvement in practice, where wages, local job opportunities and management quality could influence departures.

Read more: A model compact puts public contracts at the centre of democratic renewal

For employers, the strongest objection is that customers do not place orders three weeks in advance simply because workers receive their rotas then. Preserving service during uneven demand may require spare capacity, more deliberate deployment of staff or paid cover that proves unnecessary on quieter days. Smaller operators could find that burden particularly difficult to absorb. From this perspective, the remaining $900,000 is not an accounting inconvenience but a recurring expense that must ultimately come from margins, prices or savings elsewhere in the business.

For workers, however, the baseline arrangement also carries costs; they simply do not appear in the employer’s payroll ledger. A changed shift can mean cancelled childcare, an additional journey or lost earnings from another commitment. The demonstration assigns no monetary value to those consequences, so it cannot establish whether employees’ aggregate gains exceed the employers’ net expense. But their omission from the accounts is not evidence that they are negligible. A wage-only negotiation risks overlooking a significant condition of employment: the extent to which paid work requires unpaid availability.

What comes next is the design of the bargain. A transparent rota agreement would specify the publication deadline, define permissible exceptions and establish how late changes are requested and compensated. A subsequent trial would need to test whether recruitment savings persist, whether service levels hold and whether disruption is merely shifted to temporary staff. The fictional pilot supplies no universal formula for sharing the bill. It makes the choice clearer: flexibility has a price, and an explicit agreement is a firmer basis for allocating it than an informal expectation that employees remain perpetually available.