Job Dispatch Software: What to Look For and How to Test It
The short answer: judge job dispatch software by its engine, the part that decides which technician gets which job, in what order, and what moves when the day changes, and test that engine with your own jobs, your own rules, and your worst recent days before you sign anything. The mobile app, the map, and the invoice template are table stakes. The engine is what you are actually buying, and it is the part a demo is least likely to show you honestly.
Our scheduling and dispatch software guide covers the category as a whole: what the software does, the four capabilities that separate optimization from a calendar, and the questions to ask any vendor. If you are new to the terms, start with scheduling and dispatch explained. This guide goes one level down. It is about the dispatch engine itself: what it should decide, how assignment, skills matching, travel, and re-dispatch work inside it, how to run a trial that tells you the truth, and how to score what comes back.
What job dispatch software actually decides
Strip away the screens and a dispatch engine makes four kinds of decisions:
- Who. Which technician or crew is allowed to do this job, and which of the allowed ones should.
- When and in what order. Where the job sits in that person's day, given the promised window, the travel, and everything else on their route.
- What moves. When something changes (a job runs long, a technician calls in sick, a part is missing), which assignments change and which stay put.
- What happens to the new job. When an urgent call lands at 10:30, who absorbs it and what gets pushed.
To make those decisions the engine needs four inputs. Jobs: location, job type, estimated duration, the window the customer was promised, the skills and equipment required, and a priority. Resources: each technician's skills and certifications, shift, start and end location, and what is on the truck. Rules: the hard constraints that can never be broken and the soft preferences that should usually win. An objective: what a good day means for your business, whether that is less drive time, more jobs completed, more windows met, less overtime, or some balance of them.
The output is a proposed plan, and a good engine shows its reasoning. If a dispatcher cannot see why a job went to one technician instead of another, they will not trust the plan, and a plan nobody trusts gets rebuilt by hand every morning.
Three kinds of dispatch engines
Products in this category use the same words for very different engines. Before you compare features, find out which of these three you are looking at.
A board with suggestions. The dispatcher drags jobs onto technicians, and the software filters the list (only show qualified people, only show people with open time) or highlights the nearest one. The decisions are still the dispatcher's. This is the right engine for a small team with simple rules and a dispatcher who knows the territory, and it is what some products mean when they say "smart dispatch."
Rules-based assignment. The software assigns each job as it arrives, using a filter and a ranking: remove everyone who is not qualified or available, then pick the best of the rest by distance, workload, or a preference order. It is fast and easy to understand. Its weakness is that it decides one job at a time. The nearest qualified technician at 8:05 may be the only person who could have taken the harder job at 9:30, and a one-at-a-time engine cannot see that coming.
Optimization. The software looks at the whole day, or the whole week, and searches for the set of assignments that best meets the objective without breaking any hard rule. This is what dispatch optimization software should mean. It handles trade-offs that the other two cannot, such as giving a nearby job to a farther technician because it frees the specialist for the afternoon. It is also the engine most likely to produce a plan that looks strange to an experienced dispatcher, which is why explanations and easy overrides matter more here than anywhere.
None of these is the right answer for everyone. The mistake is buying one while believing you bought another.
How assignment decisions get made
Inside every engine, assignment runs in two passes, and the quality of each pass is what you are evaluating.
Matching: who is allowed
The first pass removes everyone who must not do the job. Typical hard filters:
- Skills and certifications, ideally with levels (can install, can repair, can commission) rather than a yes or no, and with expiry dates, so an expired certification stops assignments on the day it lapses rather than when someone notices.
- Site access: clearances, background checks, or customer-specific approvals that only some people hold.
- Equipment: the job needs a lift, a specific meter, or a part, and only some trucks carry it.
- Territory and licensing: the areas each person may work, including any licensing that applies by location.
- Crew size: two-person jobs that need two qualified people at the same place at the same time.
A matching pass that gets this wrong is worse than no software, because it fails with confidence, and the cost of one unqualified technician on one regulated job lands on your business, not on the software.
Ranking: who should
The second pass orders the people who are left. This is where soft preferences live: the technician already nearby, the one who was at this site last time and knows the equipment, the one with the lightest load today, the trainee who should be paired with a senior on this job type, the customer who asked for a specific person. Ranking is where your business's judgment hides, and it is the part most products let you tune least. Ask exactly which preferences you can set, how they are weighted against each other, and whether you can change the weights yourself.
Duration estimates: the hidden input
Every plan rests on how long each job will take, and this is the input most evaluations ignore. If the engine assumes every repair takes an hour and your repairs range from twenty minutes to half a day, every plan it produces is fiction by mid-morning. Ask where durations come from. The weakest answer is a fixed number per job type that someone types in once. A better answer adjusts by job type, equipment, and technician, and learns from the actual times recorded in the field. If the product cannot read your completed jobs back into its estimates, plan to maintain the estimates by hand, and budget the time for it.
Travel time
Travel is the second hidden input. Straight-line distance understates real driving, especially across rivers, highways, and city centers. Road-network travel times are better, and travel times that vary by time of day are better still. Ask also where each technician's day starts and ends (home, depot, or the last job), whether the engine plans the drive home, and how it treats parking or check-in time at large sites. Small errors in travel compound across a route of six or eight stops.
Same-day re-dispatch: what good looks like
Every product can build a morning plan. The day is what separates them. The events a dispatch engine has to handle are predictable: a new urgent job, a job that runs long, a customer who is not home, a cancellation that opens a gap, a part that is not on the truck, a technician who goes home sick, a vehicle that breaks down.
Good re-dispatch has five properties.
- It changes as little as possible. When one job runs long, the engine should move one or two appointments, not reshuffle the whole team. Technicians and customers both lose trust in a plan that changes every hour. Ask whether the engine can minimize disruption, not just maximize the objective.
- It respects locks. Jobs already in progress, appointments confirmed with the customer, and assignments the dispatcher pinned by hand should stay put unless a person releases them.
- It shows the ripple before it commits. Who moves, which windows slip, what drops to tomorrow, and who goes into overtime, all visible on one screen.
- It waits for a person on anything that matters. A proposed change the dispatcher confirms in one step, rather than a change that has already happened.
- It tells the people affected. The technician's phone updates, and the customer whose window moved gets a message, without the dispatcher making a call.
A product that does the first two badly will produce plans that are optimal on paper and chaotic in the field. Our scheduling and dispatch automation guide covers the order to automate the information around these decisions, which is usually where to start.
How to run a dispatch software trial with your own jobs
A vendor demo uses a clean data set designed to make the engine look good. A trial uses yours. The goal is to learn, before you commit, whether the engine makes decisions your best dispatcher would accept on a normal day and on a bad one.
Build a test pack
Before the trial, assemble three things:
- One ordinary day, with every job, its type, location, window, and requirements, and the roster of who was working.
- One bad day, the one with the breakdown, the overrun, and the emergency call, with the timeline of what happened and when.
- Two weeks of history, with actual start and finish times, so you can compare the engine's plans with what your team really did.
Strip anything the vendor should not see, but keep the shape of the work intact. A test pack with the hard jobs removed tests nothing.
Run the scenarios
Load the ordinary day and ask for a plan. Then replay the bad day as a series of events and watch the engine respond to each one. Then run the scenarios below, adapted to your operation. Write down for each one whether the engine handled it, needed a workaround, or failed.
| Scenario | What to watch |
|---|---|
| A job only two people are certified for | It never assigns anyone else, even when both are busy |
| A two-person job | Both people arrive together, and neither waits idle |
| A promised two-hour window | The plan keeps it, or flags it early when it cannot |
| A technician calls in sick at 7 AM | Their jobs move with the fewest changes, and the urgent ones go first |
| A job runs two hours long | Only the affected appointments move, and customers are told |
| An urgent contract job lands at 10:30 | It goes to a qualified person without breaking anyone else's hard rules |
| A gap opens after a cancellation | The engine offers to pull work forward, and asks before it does |
| The part is not on the truck | The job is reassigned, or rescheduled with the part reserved |
| A new hire with partial skills | They get only the work they are cleared for, or a paired job |
| End of day, one job left | It shows the overtime trade-off instead of deciding silently |
Ask why
For a sample of assignments, ask the engine (or the vendor) why this technician got this job. You are listening for an answer in terms of your rules and your objective. "The algorithm chose it" is not an answer, and a dispatcher who hears it will stop using the plan.
Watch the override
Have your dispatcher change a few assignments by hand. How many steps does it take? Does the engine respect the change for the rest of the day, or undo it on the next recalculation? Is the override recorded with who made it and why? Overrides are not a failure of the engine. They are how the engine learns where its rules are incomplete.
How to score the results
Score the trial against your own history, not against the vendor's claims. For the two weeks you supplied, compare the engine's plans with what your team actually did on the measures you care about: total drive time, jobs completed per technician, windows met, overtime hours, second visits, and any rule the engine broke. You are not looking for a dramatic number. You are looking for a plan that is at least as good as your dispatcher's on the easy days, clearly better on the hard ones, and never wrong on a hard rule.
Then count two things that matter more than any single measure. Workarounds: every scenario that needed a manual fix, a note, or a spreadsheet will need it every day you run the product. Dispatcher edits: how much of the proposed plan your dispatcher changed before accepting it. A plan that gets rewritten every morning is a suggestion engine at an optimization price. If you cannot measure these from the product's own data, ask how you will measure them after go-live, because automated reporting is how you will know whether the engine is still earning its place a quarter later.
Seven things to write down before the trial
The trial is only as good as your description of the work. Write these down first, even roughly:
- Job types and durations. Each type of job you do and how long it really takes, as a range, from your records rather than your quote templates.
- The skills matrix. Who can do what, at what level, with which certifications and their expiry dates.
- The hard rules. Everything that must never be broken: certifications, site access, equipment, licensing, crew size.
- The soft preferences. Continuity, workload balance, trainee pairing, customer requests, and which of them win when they conflict.
- Priorities and promises. Contract response times, promised windows, and what jumps the queue.
- Your objective. What a good day means, in order: fewer miles, more jobs, more windows met, less overtime.
- The systems of record. Where the customer, the job, the parts, and the invoice live today, and what the dispatch engine has to read and write.
Most companies find that writing the list is the most useful hour of the evaluation. It turns the dispatcher's memory into something you can test, and it often shows that a rule everyone assumed was essential is actually a habit.
When off-the-shelf dispatch is enough
For many field operations, a packaged product is the right answer, and you should buy one. It is enough when your rules are ordinary (a common trade, a standard skills matrix, normal windows), when the trial handles your bad day with few workarounds, when your dispatcher accepts most of the proposed plan, when the product reads and writes the systems you already use, and when the per-seat price still makes sense at the size you expect to be. If that describes you, put your energy into setting the product up well: an accurate skills matrix, real durations, and a habit of recording overrides. In discovery, we will tell you when that is the answer.
When to build your own dispatch engine
The signals that a product will not grow into your operation show up clearly in a good trial:
- Hard rules came back "mostly." Certifications, equipment, licensing, or crew rules the product cannot express exactly.
- The ranking cannot be tuned. Your soft preferences are the ones that matter, and the product's weights are fixed.
- Re-dispatch reshuffles everything, or ignores the locks your dispatcher sets.
- Durations never improve, because the engine cannot learn from your completed jobs.
- The workarounds outnumber the passes on your scenario list.
- The engine cannot reach your systems of record, so the plan is built on data someone re-typed.
If three or more of those describe your trial, building is worth a serious look. Our guide to building your own dispatch and scheduling system compares the three ways to do it.
What a custom dispatch engine looks like
A custom build does not replace your CRM, your accounting system, or your inventory system. It reads from them, runs your rulebook, and writes the plan back. In practice that is a matching pass built from your actual hard rules, a ranking pass tuned to your preferences, duration estimates learned from your own completed jobs, re-dispatch that changes as little as possible and respects every lock, plans that explain themselves, and a field view and customer messages that update on their own. The custom application work builds the engine and the screens; the integrations work connects it to the systems that hold the truth.
An illustrative example. A commercial fire protection company runs 35 inspectors and service technicians across three states. Inspections come due on fixed schedules, each state requires its own licensing, some buildings need escorts booked in advance, and two-person jobs need a licensed lead and a helper at the same site at the same time. The product they trialed assigned by distance, ignored the licensing by state, and rebuilt the whole afternoon whenever one inspection ran long, so the dispatcher planned on paper and typed the result in. A build for this company encodes the licensing, escort, and crew rules as hard constraints, plans inspections weeks ahead from their due dates, ranks by continuity so the technician who knows a building goes back to it, learns durations by building and inspection type, and moves only the affected appointments when a job overruns, with a one-step confirmation for the dispatcher. The dispatcher stops rebuilding the plan and starts managing the exceptions.
That kind of build typically ships in two to eight weeks from an approved Blueprint, with a working demo every week, and it runs inside one fixed monthly price that covers the build, hosting, maintenance, and ongoing improvements, with nothing to pay up front. Your job and customer data stays yours. How it works walks through the whole process from the first call.
A plan you can start this week
- Name your engine. Decide whether you need a board with suggestions, rules-based assignment, or optimization, based on how often your day changes and how complicated your rules are.
- Write the seven things down. Job types and durations, skills matrix, hard rules, soft preferences, priorities, objective, systems of record.
- Build the test pack. One ordinary day, one bad day, and two weeks of history with actual times.
- Run the scenario table against two products. Same pack, same scenarios, the same person scoring both.
- Count workarounds and dispatcher edits. Those two numbers tell you more about fit than any feature list.
The short version
Job dispatch software is only as good as its engine: matching that never breaks a hard rule, ranking that reflects your judgment, durations and travel times that match reality, and re-dispatch that changes as little as possible, respects locks, and asks before it acts. Find out which kind of engine you are looking at, test it with your own jobs and your worst days, and score it against your own history by counting workarounds and dispatcher edits. Buy when the engine fits your rules. Build when the rules that make your operation work are the ones the products cannot express.
If you have the seven-item list and a bad day you can replay, bring them to a free discovery call. You will get a straight answer on whether a product can carry your dispatch, and if it cannot, a free Blueprint that names the outcome, the plan, and one fixed monthly price.