Scheduling optimisation
Scheduling optimisation explained: how a service day gets planned
Scheduling optimisation is the use of algorithms to assign and sequence field jobs across a workforce so that travel, overtime and missed commitments are minimised, while skills, parts, appointment windows and working hours are all respected. It replaces the manual planning board with a plan that can be scored and reproduced.
For field service
Know what the job needs before the van rolls
Venta Capture, a product of VentaVid, lets the customer show you the fault first, so the engineer arrives with the right part or does not need to arrive at all.
Spelled optimisation in British English and optimization in American English, and sold under names like resource scheduling, workforce optimisation, or dynamic scheduling. The underlying problem is the same one every time.
What does scheduling optimisation mean in practice?
It means turning a planning board into a mathematical problem with a score, then searching for a better score than a human would find by hand.
Academic literature calls this the technician routing and scheduling problem, a close relative of the vehicle routing problem. Research on the dynamic multi-skill version describes it as scheduling and routing with time windows and synchronisation constraints, where new requests arrive in real time and force constant reevaluation of team plans (arXiv:2309.09321).
That last clause is the part practitioners feel. The hard problem is not building a schedule. It is rebuilding one at 11:40 when a job has overrun and an emergency has landed.
What constraints does the optimiser have to respect?
The constraints are what separate real scheduling from route planning. Most operations have more of them than they realise:
- Skills and certifications. Gas Safe, high voltage authorisation, manufacturer accreditation, security clearance for a site.
- Time windows. The customer's appointment slot and the site's access hours, which are often not the same thing.
- Contractual commitments. Response and resolution deadlines, which vary by contract, not by job type.
- Parts and tools. Van stock, stores collection, a lifting device that only two vehicles carry.
- Working time. Shift patterns, breaks, maximum drive time, start and finish locations.
- Synchronisation. Two-person lifts, a second trade attending, a shutdown window agreed with production.
- Continuity. Sending the technician who attended last time, which customers value and solvers rarely weight.
How is a schedule scored?
Every optimiser is minimising a weighted sum, and the weights are a business decision that gets treated as a technical one.
Typical terms in the objective: travel distance or drive time, overtime, contractual breaches, unassigned jobs, appointment window misses, and workload imbalance across the team. Set the travel weight high and you get a tight, cheap-looking plan that quietly breaks appointment promises. Set the breach penalty high and you get a compliant plan with vans crossing each other all day.
Neither is wrong. But the trade-off belongs to the operations manager, not to whoever configured the tool.
Scheduling optimisation: a worked example
A utility metering team runs 11 technicians and 96 jobs a day across a county, with two-hour appointment windows and three technicians certified for commercial installations. The manual board averaged 38 percent of paid time on the road and roughly four missed windows a day. Reweighting the objective to penalise window misses heavily, while accepting slightly longer routes, cut missed windows to under one a day and pushed travel time to 41 percent.
Travel got worse. The metric the customers actually notice got better. That is what a weighting decision looks like when it is made deliberately.
Why good algorithms produce bad schedules
The solver optimises the inputs it is given, faithfully, including the wrong ones. Three inputs cause most of the damage:
- Job duration estimates. If a job type is booked at 60 minutes and averages 95, every schedule built on it collapses by mid-morning and the day is run by exceptions.
- The fault description. Get the fault wrong and the skill requirement, the duration and the parts are all wrong together. Aquant's 2025 Field Service Benchmark Report found that a failed first visit adds two more visits on average and extends resolution by 14 days (reported in TechNation). No amount of routing cleverness recovers that.
- Stale van stock. The optimiser assumes a part is on the vehicle. Stores says it is. The technician knows it was used on Friday.
This is why first-time fix rate and schedule adherence rise and fall together. They are both downstream of how well the job was understood before it was planned.
Static, dynamic and real-time scheduling
- Static. The plan is built overnight and largely held. Works for planned maintenance and installations.
- Dynamic. The plan is re-optimised through the day as jobs complete, overrun or arrive. Necessary wherever reactive work is a large share of the board.
- Real-time or appointment-time. The solver runs while a customer is on the phone, offering only slots it can actually keep.
Dynamic scheduling has a human cost that is easy to underestimate. A plan that changes six times before lunch destroys a technician's ability to prepare, so most operations that make it work freeze the next two or three jobs and only re-optimise beyond that horizon.
Better inputs at the front of the process do more for schedule quality than a better solver does. Knowing the actual fault, the actual access situation and the actual part needed turns an estimate into a plan, and takes a share of jobs off the board entirely as avoidable truck rolls. That is what a proper pre-visit assessment is for. Guided capture tools like Venta Capture, a product of VentaVid, are one way teams collect that detail from the person already on site before the schedule is built, rather than discovering it on arrival.
