Schedules that cover every shift and keep to every rule.
Shift Optimization builds a complete staff schedule in seconds — every shift covered by the right skills, every contract respected, and staff requests honored wherever the schedule allows. Your scheduling team spends its time on decisions, not spreadsheets.
- Seconds, not days
- A full schedule comes back in seconds, so you can try a few versions before choosing one.
- Every rule kept
- Contracts, rest times and shift rules are built in, not checked by hand afterwards.
- Fair to staff
- Leave is respected, and shift requests are honored wherever the schedule allows.
- Easy to change
- Lock what has been agreed and rework the rest when someone calls in sick.
Scheduling by hand costs more than time.
- Days of work every cycle Building a schedule in a spreadsheet takes experienced people days — and much of it starts again when anything changes.
- Gaps found too late Short shifts and rule breaks often surface after the schedule is out, when they are hardest to fix.
- Overtime and agency spend Shifts left uncovered end up filled at premium rates.
- Requests that slip through With dozens of people and hundreds of shifts, it is hard to see whose requests have been met and whose have not.
Anyone who staffs shifts by skill, around the clock.
- Hospital wards and departments Registered nurses, enrolled nurses and healthcare assistants across early, day, late and night shifts.
- Aged care Minimum care time and registered-nurse cover, delivered by a largely part-time workforce.
- Emergency and critical care Cover that cannot run short, with strict rest between shifts.
- Other 24/7 operations Control rooms, contact centers, security and transport crews — wherever shifts need particular skills.
From your team to a published schedule in four steps.
-
Describe your team
Each person's contract, skills, leave and shift requests.
-
Set the cover you need
How many people each shift needs, by skill, for every day.
-
Get a schedule in seconds
Everyone's shifts for the whole period, with anything short or over clearly shown.
-
Adjust and publish
Swap shifts, lock what must stay, and publish once staff have been told.
Built around the rules your organization already works to.
- Contracts Hours per week and over the schedule, the number of shifts, runs of consecutive shifts, days off in a row, weekends worked, and minimum rest between shifts.
- Shift rules Patterns that are not allowed — no early shift straight after a night — and limits on particular shifts for each contract.
- Skills and cover Every shift is staffed by skill, so a registered nurse is never counted as a healthcare assistant. Cover can be strict, or allowed to bend when you decide it may.
- Leave and requests Time someone cannot work is always respected. Requests to work, or not to work, a shift are honored wherever the rest of the schedule allows.
A sick call shouldn't mean starting again.
Schedules change after they are made. Lock the shifts that must not move — a whole person, a day, a single shift — and let Shift Optimization rework the rest.
You choose how much may change: only what is broken, the week on screen, or everything not locked. The result comes back as a proposal, person by person, for you to accept or discard — so staff who have already been told see as few changes as possible.
Tested on the standard nurse scheduling benchmark.
Researchers compare scheduling methods on a shared set of published problems, the Curtois and Qu nurse scheduling instances: 24 problems, from 8 nurses over two weeks to 150 nurses over a year, each listed with the best schedule anyone has published for it.
We scheduled the first 19 instances with the engine behind Shift Optimization, using FICO Xpress as its optimization solver. Every schedule was then checked by a separate program against the benchmark's own rules: none breaks a hard rule.
- 11 of 19
- match the best published schedule
- 18 of 19
- within 2% of the best published schedule
- 19 of 19
- keep every hard rule
Consistently good schedules, with no tuning. We used the same settings on every instance, adjusting only the time allowed. Across teams of 8 to 120 people and planning periods of two to twelve weeks, 18 of the 19 schedules came within 2% of the best published result.
| Instance | Staff | Weeks | Shift Optimization | Best published | Difference | Time limit |
|---|---|---|---|---|---|---|
| 1 | 8 | 2 | 607 | 607 | 0% | 1 min |
| 2 | 14 | 2 | 828 | 828 | 0% | 1 min |
| 3 | 20 | 2 | 1,001 | 1,001 | 0% | 1 min |
| 4 | 10 | 4 | 1,716 | 1,716 | 0% | 5 min |
| 5 | 16 | 4 | 1,143 | 1,143 | 0% | 5 min |
| 6 | 18 | 4 | 1,952 | 1,950 | +0.1% | 5 min |
| 7 | 20 | 4 | 1,057 | 1,056 | +0.1% | 5 min |
| 8 | 30 | 4 | 1,306 | 1,300 | +0.5% | 5 min |
| 9 | 36 | 4 | 439 | 439 | 0% | 5 min |
| 10 | 40 | 4 | 4,631 | 4,631 | 0% | 5 min |
| 11 | 50 | 4 | 3,443 | 3,443 | 0% | 5 min |
| 12 | 60 | 4 | 4,040 | 4,040 | 0% | 5 min |
| 13 | 120 | 4 | 1,348 | 1,348 | 0% | 60 min |
| 14 | 32 | 6 | 1,278 | 1,278 | 0% | 10 min |
| 15 | 45 | 6 | 3,904 | 3,828 | +2.0% | 10 min |
| 16 | 20 | 8 | 3,225 | 3,225 | 0% | 15 min |
| 17 | 32 | 8 | 5,749 | 5,746 | +0.1% | 15 min |
| 18 | 22 | 12 | 4,497 | 4,459 | +0.9% | 15 min |
| 19 | 40 | 12 | 3,567 | 3,149 | +13.3% | 30 min |
Compared with published methods
Strandmark, Qu and Curtois (2020) report results on the larger instances for their own column generation method, for CPLEX run for an hour, and for earlier published methods, including a network flow model solved with Gurobi. A dash means that method returned no schedule.
| Instance | Shift Optimization | Strandmark et al. | CPLEX, 1 hour | Network flow, Gurobi |
|---|---|---|---|---|
| 13 | 1,348 60 min | 1,356 8 min | – | 1,652 5 h |
| 14 | 1,278 10 min | 1,870 17 s | 1,284 | 1,278 12 min |
| 15 | 3,904 10 min | 4,271 1 min | 7,789 | 3,853 5 h |
| 16 | 3,225 15 min | 3,309 13 s | 3,230 | 3,225 28 min |
| 17 | 5,749 15 min | 6,426 14 s | – | 5,746 24 min |
| 18 | 4,497 15 min | 4,962 1 min | 4,569 | 4,460 5 h |
| 19 | 3,567 30 min | 3,293 5 min | – | 3,204 5 h |
Built by scheduling experts, advised by industry veterans.
Amal de Silva
Founder & CTO
For 30 years Amal has helped airlines, railways, bus operators, energy companies and manufacturers plan their crews, vehicles and equipment. Many of the systems he built are still in daily use.
Before starting Gradient Optimization he was a Senior Research Scientist at Apple, and before that worked at Gurobi, IBM, GE Transportation and ILOG. He holds a PhD in Computer Science and Operations Research.
Vincent Gosselin
Advisor
Vincent has more than 30 years in AI and optimization, much of it at ILOG and IBM, leading large-scale optimization projects for companies including Samsung, Toyota, McDonald’s, Dassault Aviation, TSMC and Disney. He holds an MSc in AI from Paris-Saclay University and is now CEO and co-founder of Taipy, which turns optimization models into production-ready applications.
Desmond Delandro
Advisor
Desmond has more than 20 years of experience building enterprise software businesses, with senior roles at AWS, Automation Anywhere, IBM and ILOG. He is now Chief Revenue Officer and Chief Operating Officer at Avaiga, the company behind Taipy, where he leads commercial strategy and growth.
Set up for you, around how you work.
We configure Shift Optimization around your contracts, shift patterns and cover rules, so it fits your organization from day one. No specialist knowledge is needed on your side — your scheduling team works with people, shifts and rules they already know.
Want to know what is under the hood? Read the technical details on the Gradient Optimization site.
Shift scheduling, explained in four pages.
Why building a schedule is so hard, how Shift Optimization compares whole schedules instead of filling one shift at a time, and how your planners stay in control. Free to download, no form to fill in.
See what it would do with your schedule.
Tell us about your team — how many people, which shifts, and how you schedule today — and we will show you what Shift Optimization would do with it. Leave your details and we will be in touch to book a demo.
Thank you — your request has been sent. We will be in touch shortly.
Or email us directly at hello@gradient-optimization.com